Beverage Marketing

A textbook for choosing and selling a viable beverage offer

Can Imagine | Public reading edition | 15 October 2026

Selling a beverage means winning a real choice, serving it reliably and retaining enough value to continue. This textbook helps founders, marketers, designers and operators choose what to change: product, promise, pack, price, route, campaign or operations.

The framework connects four jobs: choose the opportunity and growth problem; explain choice through rival testable mechanisms; establish trade, operating and economic feasibility; then compare interventions and make a bounded commitment. These jobs interact; they are not compulsory customer stages.

The six questions about occasion, recognition, expected value, access, experience and commercial delivery are optional prompts, not investment scores. Neither repeat percentage nor any single metric is a universal objective. Choose for the business, horizon and value after relevant costs.

How to read and use this book

Founder route: before ordering a run, start with the completed choice in chapter 13 and the decision record in chapter 17. Use chapters 7 and 8 to expose the cash, minimum and operating constraint; use chapters 3 through 6 to challenge the occasion and route. Finish with one permitted next commitment, a spending boundary, an owner and a reopening condition.

Marketer route: for an in-market offer, begin with the growth boundary in chapter 1 and the alternatives in chapter 9. Use chapter 5 for a pack decision, chapter 10 for a creative or creator brief, and chapters 11 and 12 for measurement. Finish with a filled brief and a defensible change, stop or bounded-test recommendation. Read chapter 14 for the established-business perspective.

You can use the remaining chapters as reference. Completing every worksheet is optional. Appendix D supplies model repairs for six core exercises. This is decision-focused commercial education; it is not a full curriculum in quantitative market sizing, research methods, creative production, food science or law. Specialist work is needed where the decision exceeds these boundaries.

Work on one real decision. Keep its buyer, occasion, substitute, offer, route and assumptions visible. Ask whether each chapter changes an available action; a completed worksheet that changes no decision is paperwork.

The public case window is 6 October 2024 through 6 October 2026. Publication and observation dates are distinguished in the source notes. Older comparison periods are baselines, not relabeled current events. FDA and FTC guidance are live reference material checked on the cutoff date, not recent commercial case evidence. This edition contains original synthesis and short attributed accounts of public findings; it does not reproduce private source collections or creator posts.

Start with the record you need

For a first draft, fill this small record: Decision and review date: ____. Current practice and alternative: ____. Evidence versus assumptions: ____. Cost and cash limit: ____. Missing readiness evidence: ____. Owner, next action and revisit trigger: ____. A blank is an unresolved fact, not permission to spend. Appendix D shows how to repair weak answers.

Reading formats and scenario boundaries

This public edition is available as a page-designed PDF and as reflowable HTML. The HTML provides semantic table structure and descriptions of every figure. The HTML embeds its fonts and uses the adjacent editorial image folder; keep the HTML and editorial folder together for offline reading. The PDF itself is not represented as a tagged or PDF/UA-certified edition.

The book stands alone. The completed examples can be followed without companion software. When adapting a worked example, keep its quantities, costs, dates and decision boundary together.

This commercial textbook does not replace formulation, medical, legal or investment advice. Safety, shelf life, classification and claims need qualified product/jurisdiction review. Numerical business examples are teaching assumptions, not quotes or benchmarks.

Name the scenario before copying a number. Chapter 13 uses a 1,000-can learning run and a separate 12,000-can commercial option with 10,800 saleable cans and a 6,750-can economic threshold. Chapter 7 uses a separate 600-can case. The website's separate 8,000-can teaching case is not an update to chapter 13. Their quantities and cost boundaries must not be mixed. Creative, pack and research examples below also have their own stated budgets and periods.

When you move a calculation into a deck, workbook or decision memo, reconcile the table values, plot geometry and decision text after every update. Record the source version and check date.

Four kinds of material

Reported evidence attributes a company, publisher or study finding. A company release documents its report, not an independent experiment proving its explanation. Original synthesis connects findings to decisions and remains open to correction. Fictional examples use invented assumptions to teach a calculation or choice. Proposed tests state a comparison and decision-changing result; they are not completed research.

No cited dataset validates the complete framework over a simpler product, price, availability and cost review. Its benefit must exceed the effort of using it. Match evidence to commitment: repair a broken locator directly; investigate a large production commitment more carefully.

1 Choose the strategy and define growth

For this business, product and channel, which feasible change is most likely to create additional valuable purchasing over the relevant period, and what evidence would make us choose differently?

That question contains four distinct problems: the opportunity to be chosen; the customer's choice among alternatives; the trade and operating system that supplies the choice; and the share of value the business can keep. A brand can improve one while harming another. The framework therefore cannot be a sequence of positive scores to maximize.

Make the strategic choice first

Before optimizing a campaign, choose the buyer, occasion, substitute and category boundary. A drink for an adult weekday lunch is a different proposition from a post-workout protein shake, a household soda multipack or a bartender's mixer. State what the buyer gets that makes switching worthwhile: a preferred sensory experience, less preparation, a better serving format, a credible function, acceptable total cost or meaningful social value. A list of fashionable ingredients is not a demand thesis.

Then state why the trade would carry it and why this business can deliver and capture the value. Relevant capabilities might include formulation quality, production reliability, efficient local delivery, privileged access to an occasion, recognizable brand assets, a trusted sales relationship or unusually good customer knowledge. These are candidate advantages, not automatically durable moats. Ask what competitors can copy, how fast, and what remains valuable after they match the visible idea or change price.

Write the strategy in a complete sentence: For this person in this situation, our offer should replace this alternative because of this observable advantage; this route reaches the choice; these capabilities let us keep serving it on viable terms. Underneath, list the two assumptions that would kill the idea. If the advantage is not valued, the route is unavailable or the economics require implausible volume, do not move straight to a launch campaign.

This step rules out attractive bad ideas before their outcomes arrive. An elaborate refrigerated package can be unsuitable for a route that cannot maintain its required handling. A three-flavor launch can exceed working capital before a single ad runs. A premium positioning can fail if actual buyers will not accept its price at the intended occasion. None requires a large randomized trial to recognize a binding mismatch.

Describe what is changing

Begin with an aligned growth account. For a defined household population, channel and period, retail spending can be decomposed into population size, the fraction buying, purchases per buying household, units per purchase and realized consumer price per unit. Equivalently, it can be described as buyers multiplied by average spending per buyer. These are arithmetic identities when their units and weighting agree. They are not independent causal levers.

This distinction matters because average frequency can rise when occasional buyers leave, even if no remaining person changes their behavior. Average ticket can rise through price, pack size or mix. Revenue can rise without more liquid being sold. A new flavor can gain by taking sales from the core product. The first task is to identify which quantity actually changed and which population produced it.

Keep consumer demand separate from trade movement. For a consistently defined inventory boundary, net receipts are reconciled with outward sales, losses and the change in inventory. A launch load-in or later destocking can move brand shipments without a corresponding same-period change in consumer purchases. A listing says the trade buyer authorized something; an in-stock shelf and a sold unit say different things.

Finally distinguish company, brand, SKU and portfolio. Acquisition, foreign exchange, product mix and changed distribution can make company growth unlike organic consumer demand. The current cases in chapter 15 supply direct examples of these divergences. Their role is to rule out false equivalences, not to prove why an advertisement or formula worked.

Chapter decision: identify the specific outcome to improve. For example, recruit more eligible first buyers, improve future purchases, increase profitable quantity, improve realized contribution, reduce lost transactions or reduce costly inventory. Do not default to repeat, social engagement or store count simply because that metric is convenient.

A compact map of the decision

Strategy defines the opportunity: a buyer, occasion, substitute, advantage, route and relevant capability. The growth account tells the team what moved: buyers, purchases, units, price, mix, stock or costs. Mechanism hypotheses explain why an intervention might change that account. Constraints remove infeasible options. The final comparison ranks the remaining actions by their expected value, cost, timing, risk and learning value.

Keep those roles separate. An accounting identity can be exactly true without telling you what to do. A convincing psychological explanation can be irrelevant if the product cannot reach the buyer. A profitable-looking unit can consume too much cash when its minimum run is large. A low recognition score can be less worth fixing than an expensive delivery arrangement.

Reported evidence: Celsius Holdings' Q2 2026 company revenue rose 10.6%, while CELSIUS brand revenue fell 11.7%. The same release reported a 2% decline in CELSIUS retail dollars in a specified US RTD-energy measure. [R04] Those are different entities and movement measures. The immediate task is to reconcile the boundaries, not declare the brand's marketing successful because the parent grew.

Make opportunity cost explicit

Every chosen activity uses money, inventory, people or time that could serve another option. Compare a local delivery repair with a local ad campaign, not only with doing nothing. Compare another flavor with improving availability of the core flavor. Include a delayed launch or no launch when the proposed minimum is unaffordable. A plan becomes strategic when it says which attractive option will not receive resources and why.

Chapter exercise: write the business decision in one sentence. Under it, name the strongest alternative use of the same resources and the observation that would make you switch to it.

2 Explain choice without inventing a universal journey

A purchase happens in a feasible choice set. Products can enter that set through a shelf, menu, search, household stock, another person's selection or a remembered option. Eligibility, stock, serving form, location, time and total acquisition effort can restrict it. A person need not have a prior brand relationship to choose an option that is present, understandable and acceptable.

Within that set, attention and recognition can affect what is considered and how easily the correct product is found. Expected taste, function, social meaning and credibility can affect anticipated value. Price, case commitment, waiting, search and preparation contribute to sacrifice. Defaults, prior habits and other people's decisions can affect choice without conscious deliberation. This is a set of possible mechanisms, not an assumption that everyone calculates a formal utility score.

Consumption supplies new experience. It can confirm, improve or disappoint expectations, and it can reveal that the product fits a different occasion than intended. It also changes household inventory and the timing of the next need. Future choice depends on those changes, the next opportunity and the alternatives then available. More exposure does not necessarily mean greater liking; a first sip does not establish full-serving satisfaction or a desire to pay again.

Social transmission is an optional additional path. People can help others discover an offer, convey credibility, signal taste or share entertainment. But a person may advocate for a joke or event without wanting the drink, while another quietly repurchases for years. Community and advocacy are possible assets whose commercial role must be specified and measured, not mandatory stages every brand must complete.

The causal discipline is to keep each bridge visible. An intervention may improve recognition without changing choice, understanding without changing a basket, or satisfaction without overcoming price and access. It may also improve sales but attract expensive or short-lived demand. Do not treat a favorable intermediate metric as a validated commercial proxy unless the relationship has been established for the decision being made.

Chapter decision: choose at least two rival explanations for the observed shortfall and an intervention that can distinguish them. 'The brand is weak' is not an explanation. 'People identify it as very sweet and therefore reject it in the lunch occasion' is a testable proposition if expectation, choice and real product experience are measured separately.

A mechanism has a beginning and an observable consequence

Consider a can whose front panel suggests a very sweet drink. One proposed mechanism is that a shopper rejects it because the expected sweetness conflicts with the meal. Changing the design may change that expectation. The next question is whether that change raises paid choice at the actual price. If expectation improves but purchase does not, price, availability or an incorrect diagnosis may explain the missing effect. Calling all three outcomes brand improvement would hide the failed bridge.

Some choices require little conscious brand evaluation. A guest accepts the venue's default sparkling water, a workplace buyer chooses a case that fits the budget, or a shopper buys an unfamiliar store brand. These are not exceptions to be patched into a mandatory attention-to-loyalty funnel. They show why defaults, trade selection, context and sufficient category information belong in the model.

Reported evidence on understanding and choice

In an October-November 2024 randomized simulated-grocery study of 5,636 US primary grocery shoppers, label formats produced different psychological and basket-choice outcomes. A spectrum format improved the primary basket-healthfulness outcome against positive-only labels; several other formats did not show a statistically detectable improvement on that endpoint. [R05] A second analysis of the same trial found the Nutrition Info format produced the largest gain in correctly identifying the healthier option. [R06]

A design can therefore win one measured objective and lose another comparison. These analyses are not independent replications, and neither measures ordinary beverage sales or repeat. The practical implication is to choose the decision outcome first and validate the bridge from comprehension or liking to that outcome.

Feedback and competing causes

Demand can cause a retailer to add distribution. Added distribution can generate discovery. High demand can produce stockouts; a service failure can also produce stockouts. Looking at a stockout alone cannot distinguish those causes. The same caution applies when satisfied buyers become community members: membership may be partly a result of liking rather than its cause.

A useful causal sketch uses ordinary words: this intervention changes this aspect of the offer, which should change this behavior under these conditions. Add the strongest rival cause. Specify what you can observe before interpreting the result. The sketch need not be elaborate; its purpose is to expose a claim that could be wrong.

A journey is a local description

An attention-to-proof-to-trial-to-repeat sequence can be useful for a specific campaign when those events are observable. It helps the team ask whether a changed message led to a changed action. It does not establish that every buyer must follow the same order. A shopper can discover an unfamiliar drink already available in a cooler. A guest can taste a host's choice before seeing the package. A workplace buyer can choose for many drinkers. An occasional-event product can create value without frequent personal repurchase.

Draw the actual process before assigning stages. Mark which person makes each choice and which events the data can observe. Improve access, price, service or experience whenever the evidence supports it; do not wait for an earlier awareness score to pass an invented threshold. The useful question is which feasible change creates enough additional value, not whether every box in a sequence is positive.

3 Find the occasion and the real alternative

A category is not an occasion

Energy, soda, water and protein are useful production and retail categories. They do not fully explain choice. The same person may seek stimulation before a drive, something cold with lunch, a treat after work and an adult-feeling drink at dinner. A beverage competes with the options available in that situation, including coffee, food, tap water, an existing routine or buying nothing.

An occasion description should contain a trigger, setting, desired outcome, substitute, barrier and frequency. 'Health-conscious Gen Z' contains none of those precisely. 'An adult who wants a cold, lightly sweet drink with a weekday sandwich and normally buys a conventional soda' provides a testable starting point. Age can matter, but the meal, budget and available alternatives often tell the team more about the offer.

Separate the drinker from the decision maker

Write a separate value proposition for each consequential decision maker. A consumer wants a satisfying lunch drink. A café buyer wants reliable supply, acceptable gross profit, quick service and evidence customers will order it. A distributor wants a manageable assortment and productive routes. A creator wants material their audience values and a fair, clear arrangement. One audience's enthusiasm does not automatically satisfy the next.

Find a wedge and a route to expansion

A useful starting occasion is important enough to matter, frequent enough to support the intended economics, reachable through a realistic channel and currently imperfectly served. None of those conditions can be inferred from social engagement alone. Interview people about their last actual occasion, not whether they like the idea. Ask what they bought, where, what it cost, who chose and what nearly changed the choice.

Begin with a narrow proposition, then test adjacent occasions. An afternoon decaf coffee may later serve an evening dessert ritual. But expansion can weaken clarity if caffeine status, serving temperature or expected use becomes confusing. Each extension should preserve enough shared cues to transfer recognition while making the new use explicit.

Practical recommendation: create an occasion map before a content calendar. For each candidate, write the substitute and the evidence that a problem exists. Rank by reachable opportunity, learning value and operational feasibility. Leave market-size estimates blank when they are unsupported; a confident invented total is less useful than a small observed purchase diary.

Turn an occasion into a researchable proposition

Use a recent purchase conversation rather than asking whether somebody is a target consumer. Ask the person to reconstruct the last occasion: what triggered the choice, what was available, who paid, who drank, what else was considered and why the chosen option won. A photo, receipt or actual menu can make recall more specific when the person is comfortable sharing it. An interview explains a situation; it does not estimate how prevalent the situation is.

Before collecting receipts, recordings or purchase histories, define the decision purpose and collect only what it needs. Prefer participant codes and aggregate results. Decide access, retention and deletion responsibility, tell participants how the data will be used, and obtain permissions appropriate to the research and destination. A worksheet does not authorize sharing identifiable records with an agency or AI service. [E06]

Do not infer that a broad age label explains those choices. A younger shopper may buy conventional soda on one occasion and a new sparkling drink on another. A distinctive character or cultural reference can make an offer intelligible beyond the literal demographic it depicts. Segment only where the difference changes the product, route, message or economics.

Build a narrow starting point with a credible expansion path

Specificity is valuable when it makes the first offer testable and reachable. It becomes harmful when it excludes real adjacent demand for no operational reason. A local lunch route may teach the team how to stock cold singles before it considers grocery multipacks. The next route still needs its own economics and choice evidence; success in one does not automatically validate another.

Proposed test: compare an explicit lunch proposition with the current description in comparable shops while holding the drink, price and cold availability stable. Observe paid units per eligible transaction and contribution. If people understand the occasion better but do not choose the drink more often, revise the intervention or its proposed mechanism. Do not keep narrowing the audience after results until the story appears to work.

4 Develop the product and an honest promise

Product performance is broader than ingredients

The experienced product includes flavor, aroma, texture, sweetness, carbonation, temperature, serving size, opening, portability and what happens after consumption. The same formula can feel different as a warm delivered case and an ice-cold can at a festival. Packaging and context influence expectation; the liquid then confirms or disappoints it.

Translate a benefit into an honest experience

Separate sensory pleasure, practical function and symbolic value: a drink can be chosen for one or several. The practical question is not simply what can be added to a can. It is what customer problem is solved by making the benefit drinkable, and what new trade-offs follow. Removing mixing may add freight, packaging, storage or formulation challenges. Adding multiple functions may make the promise difficult to understand and the taste harder to enjoy. These are hypotheses for development teams to test, not defects established for any named product.

A product test should isolate what is being learned. Blind tasting asks about sensory response without branding. Branded tasting asks how the total proposition affects expectation and experience. A home-use test reveals whether people finish the product, choose it at plausible times and want another at the real price. None substitutes for the others. A winning sip at a tasting table can become an unfinished can at home.

A minimum product learning program

Recommendation: test a manageable set of alternatives with intended adult users in realistic conditions. Capture first expectations, blind preference where relevant, branded evaluation, full-serving completion, intended occasion and willingness to buy at a stated price. Recruit beyond enthusiastic followers. Record disliked attributes and nonbuyers, and separate feedback by cohort instead of averaging away distinct preferences.

Set technical validation apart from marketing opinion. Have qualified specialists establish safety, stability, process and claim suitability for the intended market. In the United States, FDA distinguishes health claims, nutrient-content claims and structure/function claims; they are not interchangeable [E03]. This book is not a formulation or legal manual. The claims register below turns that responsibility into a practical workflow.

Functional credibility and actual efficacy are different

A recognizable ingredient can make an offer sound credible without demonstrating that the finished drink delivers the advertised benefit. Ingredient form, amount, the rest of the formula, processing, stability and actual use can affect which claim is supportable. A persuasive pack or endorsement cannot resolve those technical questions.

Maintain a claims register before producing labels or campaigns. Record the exact wording, what an ordinary buyer is likely to infer, the evidence supporting that inference, the actual product and serving covered, the relevant market, an accountable reviewer and a review date. Treat an implication made through images or context as seriously as a sentence. The purpose is to prevent marketing from promising more than the product can responsibly substantiate.

For US marketing, review FDA labeling classification and FTC advertising substantiation separately. Their responsibilities overlap; a permitted label category does not automatically clear an advertisement. Review the whole message, including imagery and endorsements, against evidence relevant to the finished product and use. A disclaimer cannot repair a contradictory benefit promise. [E03] [E05]

A compact claims record keeps these checks together: exact wording and likely implication; label, advertisement or both; product and serving; jurisdiction; supporting evidence and limitations; qualified technical/claims reviewer; version/date; approved execution or hold. For the fictional lunch drink, a proposed health-benefit line is held because no substantiation has been supplied. A sensory line still needs to match the actual product. Marketing approval cannot substitute for product release.

Diagnose a promise problem separately from a product problem

Fictional example: a citrus drink is judged pleasantly dry when tasted blind, but its bright confectionery-style pack makes prospective buyers expect a very sweet drink. Some appropriate buyers pass; others try it expecting candy-like sweetness and feel disappointed. A clearer front panel might improve matching between buyer and experience without changing the formula. It might also remove a cue that some buyers value. Compare options instead of assuming minimalism is the answer.

The inverse can occur. A clear pack accurately promises a sensory experience people do not enjoy at the intended price. Better communication would only recruit more people to a weak offer. Reformulation, serving changes, pack size, price or a different occasion may deserve priority. Keep those alternatives visible.

Learn across a full serving and realistic use

A short sip test favors a different task from finishing a serving with food or drinking the same product repeatedly at home. Test temperature, preparation and serving size as used, not merely as convenient at a research table. Record completion, actual use occasions, disliked attributes and abandoned servings. Ask what the person chose instead on the next relevant occasion.

Use blind testing to study sensory response and branded testing to study the whole proposition. Randomize serving order when practical, use consistent handling and avoid implying benefits the study cannot establish. Intended-user recruitment matters: enthusiastic employees and followers can help catch defects but cannot substitute for a broader learning sample. Product testing should never expose people to an unvalidated safety or eligibility risk.

Keep a blind condition genuinely blind

For a clean sensory comparison, do not reveal the package, brand or benefit story before the blind condition. Use coded samples, consistent approved serving conditions and randomized sample order. A separate group sees the package first for the branded condition; record prior familiarity and exposure. Showing a person the pack and later hiding it does not make that person an unexposed blind respondent. Technical safety and eligibility review come before recruitment or tasting.

Fictional protocol CB-TASTE-01 screens whether a pack miscommunicates sweetness; it does not establish a population preference. Recruit eligible unfamiliar adults for separate coded-sensory and package-first groups. Exact prompt: 'Describe the sweetness you expect' for the package group before tasting; then both groups answer 'Describe the sweetness you experienced, what you finished, and why you would or would not choose this serving at $3.' Record group, sample/alternative code, order, serving condition, prior exposure, descriptors, completion and stated-price choice. No universal sample size is prescribed.

Illustrative record: participant code P01, package-first, Lime, cold single, no recorded prior exposure, expected candy-sweet, experienced lightly sweet, finished serving, declined at $3 because the flavor was drier than expected. This invented row suggests checking expectation mismatch; one person's reason does not justify reformulation or estimate demand. A coded-sensory row must leave package expectation blank. Compare group patterns and real full-serving use before choosing product versus promise repair.

Separate the origin story from the product claim

Fictional sensory-origin line: 'I wanted a less-sweet drink with lunch.' It describes a motivation; any attached product description must still be true. A founder story implying that the drink reproduces a personal health improvement is held for qualified review when appropriate evidence is absent. Review the combined story, imagery, product name and invitation to buy. Founder experience, customer testimonials and creator reposts are not efficacy evidence. Add each execution to the claims register with its evidence and responsible reviewer; do not convert a sincere anecdote into an approved finished-product benefit. [E05]

Release the product through documented technical criteria

Ask the qualified development and manufacturing team which specifications, process validation, package compatibility, stability and release checks are required for this particular product. Decide who can release or reject a batch and how deviations are handled. A visually finished can is not proof of a production-ready formula. Treat sample artwork, bench prototypes and production goods as distinct states.

Chapter decision: decide whether the next useful change is to the actual experience, the promise, the serving conditions or the target occasion. Evidence of one should not be used as proof of another.

Learning exercise: write the promise a stranger would infer from your front panel. Compare it with the experience they actually describe after finishing a serving. Which misunderstanding could be fixed through communication, and which requires changing the offer itself?

5 Make the brand and package easy to choose

Give the person something they can retrieve

A brand is useful when it makes a choice easier to recognize, understand or evaluate. Distinctive visual and verbal cues can help a buyer locate an option again. Meaning supplies reasons to prefer it. These are related tasks with different tests. A distinctive can can be easy to find and undesirable; an appealing promise can be impossible to associate with the correct pack.

Distinctiveness and differentiation

Distinctiveness answers 'Which one is it?' Differentiation answers 'Why this one rather than another?' Category legibility answers 'What is it?' A beverage package must often manage all three in little space. Novelty that makes a soda look like a cleaning product may attract a camera while increasing hesitation at a cooler. The exact outcome depends on context and should be tested.

An attribute also needs a consequence. 'Low sugar' may signal a preferred nutritional profile; it may also create concern about taste. 'Premium' is an intended inference rather than a benefit on its own. Connect the attribute to a concrete choice while retaining truthful, legible information. Do not imply that an identity proposition excuses weak performance or substantiation.

Recognition and credibility answer different questions

Recognition asks whether people can identify and find the correct offer. Credibility asks whether they believe a particular promise. A familiar can can be easy to retrieve while its functional claim remains unconvincing. An unfamiliar brand can make a credible, understandable promise yet be difficult to find again. Neither condition alone establishes purchase or value.

Treat the two as separate diagnostic questions rather than literal accounts that every action must deposit into or withdraw from. Test brand attribution and correct variant selection for recognition. For credibility, identify the exact uncertainty, such as taste, ingredient content, preparation or reliable delivery, and the evidence appropriate to it. Keep a third question visible: even if the offer is recognized and believed, is it worth its total price and effort in this situation?

Repeated cues can help people learn what to look for. Preserve the cues that work, test changes where confusion persists and verify actual choice. Discounting is not automatically a breach of trust, and a consistent design is not automatically an effective one.

Identity is one source of value

The phrase 'people buy identities, not products' is too broad. People also buy a beverage because it is cold, inexpensive, familiar, tolerated by a household or the only appealing choice in stock. Identity can be potent without being universal. The operational question is when symbolic value meaningfully changes choice, for whom, at what cost, and whether it survives beyond the initial display.

Community is a stronger claim than audience. A set of followers may have no relationships with each other and no shared activity beyond watching. Look for recurrent participation, member-to-member interaction, shared norms or rituals and a reason to return without a fresh giveaway. Even then, community participation and beverage purchases require separate measurement.

A useful brand brief

Recommendation: write one page that fixes the intended occasion, substitute, promise, evidence and recognizable assets. Add what the brand does repeatedly, not merely adjectives about personality. 'Playful' is vague; a recurring way of naming flavors, hosting an event or answering a question can be observed and produced.

Test the brief with three tasks. Can people identify the brand among alternatives? Can they explain what they expect the product to be? Can they say where it would fit in their lives? Use people with different levels of familiarity and report those groups separately. A strong result among followers can coexist with confusion among first-time shoppers.

The pack has three connected jobs

A beverage package communicates, contains and travels. Its communication job includes category, flavor, variant, brand and truthful promise. Its physical job includes protecting the actual product and working with its process and intended use. Its commercial job includes surviving storage, shipment, shelf or cooler presentation and ordinary handling. A design that wins an image poll can fail either of the other jobs.

Before approving artwork, use the actual size and print conditions where possible. Inspect a shelf or menu at the distance and lighting of the intended purchase. Ask people to find a specific variant, identify the category and explain the expected taste. Then verify the production sample through the technical and retail checks relevant to its route, including legibility and scanning in the intended systems. Do not infer performance from a large screen mockup.

Reported evidence on packaging expectations

Three Chinese online tea-packaging experiments collected in December 2024, January 2025 and April 2025 examined transparency and perceptions or imagined choice. Effects varied with the tested conditions; the final study used two actual brands. [R07] This supports investigating package cues as an expectation mechanism. It does not establish retail sales, repeat or storage performance. Because two brands differ in more than familiarity, their comparison does not isolate familiarity alone.

Preserve measured assets and repair measured confusion

Consistency is useful when customers use a cue to find the right product. It is costly when it preserves a mistake. Define the specific cue being retained or changed, the people affected and the intended outcome. An established brand may need continuity for regular buyers and clearer variant information for everyone. A new brand may need category legibility before an elaborate symbolic world.

Proposed test: compare current and proposed pack families on correct brand attribution, correct variant selection, wrong-variant errors and realistic paid choice. Separate current buyers from unfamiliar shoppers. A redesign that raises masterbrand recall but increases wrong-variant selection fails that shopping objective. A visually conservative design can also lose if it leaves the drink misunderstood.

Chapter exercise: specify one fixed brand asset, one element allowed to vary and one error the new system must reduce. Explain why each choice matters commercially.

Learning exercise: remove the logo from three executions. Record who correctly attributes them, who attributes them to a competitor and what cues they used. Then put the executions beside the actual package. Recognition in a feed should help a person find something they can buy.

A pack decision with a result that can lose

Fictional pack screen CB-PACK-01: the same three citrus-drink variants, same can size and price, and two credible hierarchies. System A gives the masterbrand the largest type and flavor a smaller lower band. System B keeps the masterbrand visible but enlarges the flavor and moves it directly below the drink description. The schematic compares those cues; it is not a printer dieline, legally complete label or evidence of manufacturing fit.

Predeclare two tasks: find the Lime variant among three cans and describe expected sweetness before tasting. In this invented demonstration, 40 unfamiliar eligible adults are randomly assigned to each system, with shelf position rotated. Record all assigned participants. Correct first variant choices are 30/40 for A and 35/40 for B; wrong choices are 10 and 5. Candy-sweet expectations are 18/40 and 12/40. These small invented counts illustrate the record, not a reliable sales forecast or a proven superiority claim.

Decision: advance B to a physical legibility and branded-use check; do not authorize a full print run. The changed cue is flavor hierarchy, not an assumed universal rule that less branding is better. Check whether existing buyers still recognize the family and whether smaller masterbrand type hurts findability. Trade-off: new artwork, proofs and component inventory can cost more than the gain. Keep the current pack if the next check shows confusion or the revision cannot clear production and claims review within the budget. This is a separate pack-learning example, not the 600-can cash case.

6 Sell to the trade and make the purchase possible

The trade buyer is not merely a conduit for consumer enthusiasm. Retailers and foodservice operators choose assortment, placement, preparation, prices and supplier relationships. Their decision can depend on productivity, margin, waste, shelf or staff time, working capital, supply reliability and switching cost. A consumer-liked product can lose a listing if it fails the operator's requirements.

The brand must also decide whether it can serve the opportunity. Technical and legal feasibility are gates, not branding variables to optimize away. A persuasive function claim still requires substantiation for the actual product and market. A packaging cue that improves expectations may create physical performance or cost trade-offs. No amount of perceived trust makes an unsupported claim true.

Evaluate the incremental return to a proposed intervention over a declared horizon. Include net realized revenue, relevant product and channel costs, returns or waste, the intervention's full costs and the value of displaced sales. Keep accounting bases consistent so an allowance is not subtracted twice. Separately model the cash needed for inventory, terms and timing. Contribution, operating profit and cash survival answer different questions.

Some interventions rationally lose money early to build distribution, learn or create later value. That does not excuse unbounded losses or retrospective changes of objective. Set the purpose, allowable downside and horizon first. Conversely, a higher margin rate or adjusted earnings figure does not prove that withdrawing volume was optimal; total value, relevant costs and the counterfactual still matter.

Chapter decision: define what success is worth, what the business can afford, and which options are infeasible. Do not maximize reach, volume, repeat or margin percentage independently of the business objective.

Design the route from the choice backward

A cold single in a café, a grocery multipack, a delivered case and a prepared serve are different offers. Each changes who chooses, who pays, how inventory moves, where the margin sits and which failures matter. Choose the route that can reach the intended occasion with acceptable service and economics; do not add channels solely to make a distribution slide look impressive.

Map the actual sequence from production through distributor or direct delivery to shelf, menu or website. Identify who sets the selling price, who holds inventory, who replenishes it, who handles complaints and who bears loss or returns under the agreement. Record what data the brand will actually receive. Trade movement and final-consumer purchase must not be combined into one sales count.

Make a buyer proposal answer the buyer question

A retail or foodservice proposal should explain the customer and occasion, the relevant alternative, the product and pack, expected service requirements, evidence from real movement where available and the economics under the proposed terms. State whether the evidence comes from a small paid pilot, a different geography or another channel. A buyer should not have to infer a replenishment plan from a brand moodboard.

Prepare for objections that are commercially rational: uncertain productivity, duplication of an existing SKU, slow preparation, limited cold space, waste, unreliable delivery or an expensive minimum. A response may involve a smaller opening order, a different case size, staff instruction, a limited trial or a clearer replenishment commitment. Do not promise service capacity the business has not established.

Inspect availability rather than counting listings

A listing is authorization. A shopper still needs the correct product in a place they can identify and buy from, in an appropriate condition, at an acceptable total sacrifice. Walk the purchase path. Check the current price, variant, stock, temperature where relevant, menu wording, checkout and delivery promise. Record failures and the party able to fix each one.

Track authorized outlets separately from outlets actually carrying stock. When feasible, track in-stock time, relevant facings or menu presence, cold availability, sales per outlet and service cost. Higher average velocity after removing weak outlets does not by itself show that the pruning created value. Evaluate the total portfolio and the opportunity cost of the lost route.

Reported evidence: Unilever's April 2026 investor commentary linked moderation in its Wellbeing business partly to a change in Liquid I.V.'s assortment share at an unnamed key club retailer in Q4 2025. [R14] It provides a concrete reason to monitor concentration and renewal risk. It does not quantify the isolated brand effect or identify the retailer.

Route power is different from brand ownership

Keurig Dr Pepper's 2025 annual filing describes licensed carbonated-soft-drink distribution through Coca-Cola-affiliated, PepsiCo-affiliated and independent bottlers. [R17] A drink can compete with a supplier's owned brand while relying on an affiliated route system. More visible beverage brands therefore do not automatically mean less distribution concentration. For a founder, the practical questions are route access, incentives, territory, service, exclusivity and alternatives if the relationship ends.

Proposed test: choose a few outlets with actual purchase friction and repair the specific problem. Compare the result with unchanged or staggered outlets when practical, while recording price and promotions. If verified access improves but paid demand remains weak, investigate preference or value instead of treating distribution as an unlimited excuse.

Compare the entire route account

Fictional route comparison CB-ROUTE-01 uses 24-can cases and a $3 consumer price. No supplier, distributor or café has agreed to these terms. The direct option starts the brand at $1.95 per can; the distributor option starts it at $1.55. Product and pack cost $0.78 per can. Both are candidates, not a route recommendation for every brand.

At one 24-can direct drop, delivery is $0.50 per can; brand contribution before fixed costs is $1.95 - $0.78 - $0.50 - $0.12 - $0.05 = $0.50. At two cases per stop, it becomes $0.75. Distributor-route brand contribution is $1.55 - $0.78 - $0.10 - $0.12 - $0.05 = $0.50. The distributor's $0.40 spread is not its net return: an invented $0.18 delivery, $0.05 handling and $0.04 credit/administration leave $0.13 before fixed costs. These assumed costs are charged to different entities, not subtracted twice from the brand.

Decision: the percentage alone does not choose the route. Direct delivery can improve at denser stops but requires route labor and faster service; the distributor buys work and reach but may supply less useful data and later cash. Obtain actual service, return and termination terms before ranking either offer. A lower nominal margin need not mean a worse route.

Reconcile one store week before claiming demand

Fictional ledger CB-RETAIL-01 covers one Lime SKU, three assigned cafés and Week 1 (seven authorized days each). All numbers are cans. No returns or transfers occur. Opening + receipts - paid - samples - losses = closing.

Principal rollout velocity is 50/3 = 16.67 paid cans per assigned store-week. The separate in-stock diagnostic is 50/17 = 2.94 cans per verified in-stock store-day. Removing B would hide a route failure. Neither measure identifies unique households or repeat. Brand invoices cover 48 delivered cans, not the 50 consumer units; $93.60 due is not cash received. Check collection separately. This wholesale example differs from chapter 13's sale-linked collection assumption.

Blank movement row: store/SKU/week __; opening __; receipts __; returns/transfers __; paid __; samples __; losses __; closing __. Service row: authorized/in-stock days __; price/promotion __; invoice/collection dates __; observed failure __; responsible role and due date __. The source records folder supplies this as a CSV.

7 Price the offer and reconcile the economics

Start with the customer's total sacrifice

The full sacrifice includes shelf price, shipping, search, travel, preparation, storage, delay and case commitment. A $3 can available cold may be an easier trial than a cheaper-per-can online case. A subscription can simplify replenishment for one household and create unwanted inventory for another. Price is part of a format and route decision, not an isolated number.

Price can affect both quantity and expectation. A higher price may signal a different offer while making it unaffordable in the intended occasion. A lower price may improve access while reducing the resources available to serve it. Those effects are empirical questions. Do not infer a demand curve from a single before-and-after price change or from an aggregate company price/mix line.

A fictional unit and order account

The following is an invented teaching example, not a market benchmark. Suppose a can has a $3.00 shelf price. The retailer buys at $1.95 and therefore has $1.05, or 35% of shelf revenue, before its own operating costs. A distributor buys from the brand at $1.55 and sells at $1.95, leaving $0.40 before its costs. The brand's starting revenue is $1.55, not $3.00.

Assume the brand then incurs $0.78 for product and packaging, $0.15 freight, $0.12 trade accruals and $0.05 expected damage or returns. Contribution before marketing and fixed overhead is $0.45 per can. At $0.20 variable marketing cost per incremental can, the remaining contribution is $0.25. These simplified figures exclude taxes and other possible costs. The point is to name the deductions, not to memorize the margins.

Now suppose promotional allowances rise $0.20 while unit sales increase. Contribution before marketing falls to $0.25. More units can produce less total cash if promotion, inventory and collection terms are unfavorable. Distinguish gross margin, contribution, operating profit and cash timing. A 'high-margin drink' claim is unusable unless its price level and included costs are known.

For an invented DTC example, a 12-pack sells for $36. Product and pack cost $10.80, pick-and-pack $3.00, shipping $8.00, transaction fees $1.38 and expected refunds/damage $0.72. Contribution before acquisition and fixed overhead is $12.10. A $15 acquisition cost makes the first order negative $2.90 before overhead. If 30% of acquired customers place one identical repeat order during the chosen window with no new paid acquisition, expected contribution per acquired customer across the initial order and one possible repeat is $15.73 before the original $15 cost, leaving $0.73. Additional service, retention costs or discounting can erase it. This arithmetic is illustrative, not a forecast.

A price increase with a volume trade off

In this invented brand-level example, 10,000 cans produce $0.50 contribution each, or $5,000 before fixed costs. A change in the brand's net realized price increases contribution to $0.70 per can, assuming other variable costs remain unchanged. If volume falls to 8,000 cans, contribution becomes $5,600 before any added implementation cost. Higher contribution can therefore coexist with lower volume.

The continuous threshold is $5,000 divided by $0.70 = 7,142.86 cans. With whole cans, at least 7,143 are required to match the original contribution before added implementation cost. This is a sensitivity threshold, not a prediction or a price elasticity estimate. If the change also costs $600 in the evaluated period, the 8,000-can scenario merely matches the original $5,000 result. Trade response, future purchasing and competitive pricing may alter the answer.

A trade buyer can rationally prefer fewer units

Suppose an operator sells 40 servings a day of option A, with $1.00 gross profit per serving before preparation and waste. Daily preparation/service cost is $15 and waste is $5, leaving $20 in this simplified comparison. Option B sells 30 servings at $1.50 gross profit each, with $5 preparation/service cost and $2 waste, leaving $38. The operator can reasonably prefer B despite fewer servings.

This is not a claim that B always wins or that those are normal costs. It shows why the beverage seller must understand operator value. Taste and consumer preference still matter, but a staff-intensive mixer, slow preparation or unreliable supply can make a promising listing unattractive. The seller may improve the serve, pack or service arrangement rather than merely offering more advertising.

Format can change both demand and cost

A single can reduces trial commitment but may be expensive to ship direct. A multipack can lower fulfillment cost per serving while requiring storage and more upfront money. A concentrate can improve freight economics while adding preparation, dilution consistency and staff training. A refrigerated product may fit a premium sensory promise but narrow routes and increase waste risk. Shelf-life and handling specifications are technical facts to establish for the actual product, not marketing claims to assume.

Trade spend and cannibalization

A promotion may increase the brand's shipment and the retailer's order without creating equivalent new consumption. A new SKU may sell by displacing a productive core SKU or pantry-loading existing buyers. Include the trade allowance, retailer execution, subsequent period and displaced contribution in the comparison. A higher per-SKU average after pruning does not prove greater portfolio value.

Separate a margin rate from total value

Reported evidence: Vita Coco's 2025 company sales rose 18% to $610 million and branded coconut-water volumes rose 21%. Gross profit grew 12% while gross margin fell from 39% to 37%. [R10] The rate alone gives the wrong answer to whether gross profit grew. That does not make all volume growth attractive: inventory, receivables, costs and the counterfactual still matter.

A commercial review should show absolute contribution, the rate used to calculate it and the cash needed to obtain it. Define the period and include what the proposed action changes. If a promotion adds paid units but mainly brings forward next month's purchase, evaluate the combined period. If a new pack shifts customers from a higher-contribution pack, subtract displaced value.

Build a cash calendar alongside the contribution model

Inventory may be paid for before shipment, customers may pay later, and allowances or returns may arrive after the headline sale. Model the actual terms rather than a generic cash-conversion rule. Write a dated sequence of supplier deposits, balances, freight, storage, marketing, collections and expected credits. Mark which amounts are contracted and which are assumptions.

Cash required at the lowest balance is a different question from expected profit over the full batch. A positive contribution model can still run out of cash; a loss-making pilot can be an affordable learning investment if its purpose and cap were agreed first. Do not use optimistic repeat to conceal an immediate funding gap.

Chapter exercise: reconcile a shelf dollar through every entity to the brand's net receipts. Then calculate contribution and the maximum cash exposure on separately stated assumptions.

See different financial measures in the same account

Separate fictional account CB-FIN-01 assumes all 100 produced cans are sold in one period, no opening inventory and no tax or financing expense. Invoice revenue is $200; a $10 allowance is deducted from revenue, leaving $190 net revenue. Product cost of goods sold is $80, so gross profit is $110 and gross margin is $110/$190 = 57.89%. Variable selling/delivery cost is $20, leaving $90 contribution under this model. Fixed operating expense is $40, leaving $50 operating profit. This chosen classification is for teaching, not a prescription for every reporting policy. [E08]

If only $150 of the $190 net receivable is collected now and all $140 of costs are paid now, current cash movement is +$10; $40 remains receivable. Do not deduct the $10 allowance again, or record the $80 inventory purchase a second time as an extra contribution cost. Revenue recognition, cost classification and collection dates answer different questions.

A complete period indexed batch and cash decision

Fictional companion CB-PILOT-600-UNSOLD-v0.2 is separate from chapter 13. All amounts are USD. Day 0 is hypothetically 2 November 2026 and Day 11 is 13 November 2026. These planning dates are not evidence that a real order or payment occurred. Revenue before the separate allowance is $1.80 per paid can. Product/pack costs $0.85 per produced can; outbound $0.12, withheld allowance $0.08 and other cash cost $0.03 apply per paid can. Fixed activation is $600. The narrow contribution screen totals $1.08 variable cost and $0.72 contribution per paid can.

At a hypothetical maximum of 600 paid cans, that screen shows $432 before fixed and -$168 after it. Covering $600 requires 834 whole paid cans, exceeding 600 capacity. It allocates product only to paid cans. The owned-batch view instead charges all 600 produced cans: 12 rejected, 60 sampled, 300 paid and 228 remaining. Paid-sale capacity after rejects and samples is 528. Remaining recovery and other batch commitments are explicitly zero.

Full-batch account: $540 revenue - $24 allowance - $510 production - $45 outbound/other - $600 activation = -$639. The same 300 paid cans in the narrow screen yield -$384. The $255 difference is 300 nonpaid cans x $0.85, including rejected, sampled and remaining stock. No assumed recovery hides the cost.

Within a day, pay outflows before recognizing receipts. Peak cumulative net outflow is $1,155; the lowest balance is -$255 on Day 3, so another $255 is needed to make every assumed payment. The ending $261 does not fund the earlier gap. Negative balances expose an unfunded need; they do not assume borrowing. Withholding the $24 allowance reduces receipts once and is not another outflow.

Decision: WAIT. The fictional founder/pilot lead has a hypothetical $1,200 spending cap and $650 loss cap, but currently authorized new spending is $0. The $639 forecast loss is not a maximum loss. With zero paid sales and zero recovery, the already committed $510 production plus $600 activation lose $1,110; paid-sales outbound/other costs are then zero. That exceeds the hypothetical $650 loss cap by $460. Solving the $255 forecast funding gap therefore does not satisfy the loss ceiling. This specified stress case is not a bound on every possible loss. Reduce the commitment or otherwise resolve the risk with actual evidence and owner approval; do not silently raise the cap. Obtain written quote scope and dates, payment/collection terms, product/storage release evidence and an available funding source; reopen on Day 11 (13 November 2026) only with those facts and explicit owner approval. The arithmetic is checkable; supplier identity, demand and technical readiness remain unverified.

To reuse the record, replace opening cash, dated payments, dated collections and their evidence; recalculate the lowest balance and funding gap before committing. In the optional workbook, Cash timing C7:C11 holds peak outflow, trough, gap, ending cash and net change. Batch decision C7:C13 holds the reconciled batch account. A positive unit contribution cannot replace either check.

8 Choose suppliers and operate a reliable beverage business

Reliability is part of the promise

A campaign and a supply plan cannot be evaluated independently when one creates demand the other must fulfill. Delayed production, inaccurate inventory, damaged packaging, unavailable flavors and unresolved complaints change the experienced offer. They can make good creative look ineffective or accelerate disappointment after a successful launch.

SKU expansion has a cost beyond the formula

A new flavor adds forecasting, packaging inventory, production scheduling, quality review, sales materials and a place in the assortment. It may win an occasion; it may also displace a productive SKU or make the shelf difficult to shop. Measure incrementality and operational burden before celebrating every new listing.

For a launch, estimate a conservative demand range, minimum practical production, shelf-life exposure, reorder lead time and cash tied up before collection. Those estimates should be owned by people able to change the plan. If a launch requires an optimistic sell-through scenario merely to avoid excess inventory, creative enthusiasm is not enough to make the risk acceptable.

Use a failure pre-mortem: imagine the campaign generated attention but the business lost money. Was the offer underpriced, freight underestimated, the retailer margin unsustainable, the wrong customers recruited, the product unavailable or the batch too large? Name the first observable warning for each plausible failure and the action it triggers.

A supplier brief starts with a product and a route

A co-packer cannot quote a reliable production plan from the word soda alone. Prepare a specification of the intended beverage, ingredient and claim constraints, target sensory profile, format, serving size, required handling, intended market and route, initial volume range and desired launch timing. Identify what is established and what still needs development. The brief is a starting point for qualified technical discussion, not a substitute for it.

Distinguish the jobs you are buying. A formulator develops a product within agreed scope. A process specialist establishes or reviews the relevant technical approach. A packaging supplier provides specified components. A co-packer manufactures and packs under an agreement. A testing laboratory answers defined technical questions. One business may perform several roles, but responsibilities, evidence and approvals still need to be explicit.

Compare quotes on the same basis

Ask each prospective supplier to specify what its quoted price includes: ingredients, packaging components, setup, processing, filling, quality checks, case packing, pallets, storage and freight, where applicable. Ask which costs are fixed per run and which change with units. Record the exact minimum, the unit it applies to and whether flavors can share it. A minimum for a printed can is different from a minimum for a finished beverage run.

Also record yield assumptions, overrun or underrun terms, treatment of rejects, lead times, capacity reservations, component commitments, payment dates and quote validity. Confirm who owns unused ingredients and packaging, where they remain and who pays for their storage or disposal. These are proposed diligence questions; no standard rate or universal contract term is assumed.

An apparently lower price can lose once freight, waste, a larger minimum or cash timing is included. An expensive micro-run can be worthwhile if it avoids a much larger uncertain commitment. Compare total expected cost and exposure for the decision, rather than a per-can figure detached from volume.

Validate the production route before treating it as capacity

Ask for relevant capabilities and references, then verify the actual line, process and pack are suitable for the proposed product through qualified review. Agree how pilot samples relate to a production batch. A bench formula, a line trial and the first commercial run can reveal different problems. No current source cited here supplies a general shelf-life or stability guarantee.

Define specifications, acceptance criteria, testing responsibilities and records before production. Establish traceability and the procedure for a deviation, complaint or potential safety problem. An operational team needs a way to isolate affected stock and contact the responsible experts. This book does not prescribe a legal recall procedure or a technical release test; those depend on the product and market.

Manage the first batch as a learning system

Track produced units, samples, damage, losses, trade inventory and paid sell-through separately. Reconcile counts rather than treating every shipped can as sold to a drinker. Keep batch identity and the actual date/handling limits available to the people making fulfillment decisions. Do not run a promotion to clear stock that is unsuitable to sell.

Define a replenishment trigger from demand, lead time, uncertainty and usable inventory. A short launch spike can produce a false reorder signal. A long lead time can cause the opposite error: waiting for stock to become visibly low before ordering again. Use several scenarios and identify the cost of a stockout versus excess inventory under the actual agreement.

Keep assortment within the team's ability to deliver

A new flavor can win a buyer and still weaken the business if it divides a small production run, adds components or displaces a productive line. Before adding it, state the distinct occasion or value, incremental demand hypothesis, minimum and cash exposure, complexity cost and role within the portfolio. A limited release can have a one-time purpose, but its development and remaining stock still count.

When pruning an assortment, record the same trade-offs. Fewer SKUs can simplify operations but disappoint buyers or reduce the brand's relevance. Improved average sales per remaining SKU can be a mathematical result of removing weak items, rather than evidence of stronger total demand. Evaluate the portfolio and service result over the appropriate period.

Use automation with accountable review

Automation can help reconcile stock, prepare consistent sell sheets, organize customer feedback and produce creative variants. It can also repeat an incorrect product fact across many assets. Keep formula facts, claims, package dimensions, approved logos, rights and current availability in controlled source records. Require review before generated output becomes a label, promise, endorsement or supplier instruction.

Measure actual time saved, error rate and rework. More content is not an operating achievement if no one can learn from it. Chapter decision: choose the smallest reliable operating system that supports the next commitment, and identify the failure that would require it to change.

A filled supplier comparison and responsibility handoff

Fictional supplier screen CB-QUOTE-01 compares one 600-can development run, not the companion cash case. No actual quote or supplier exists. A's $0.90 quote covers 600 filled cans; B's $0.65 is filling only. Dates and exclusions remain fields to verify, not invented commercial assurances.

Decision: do not select B on $0.65 alone. A has lower assumed commitment, but neither is order-ready. Procurement lead requests matching written scope, terms and technical evidence before the next review; no deposit is authorized. A blank comparison row records vendor/quote ID, specification version, date/expiry, unit/minimum, inclusions, exclusions, total commitment, payment dates and evidence owner. The records CSV supplies these fields.

This is an organizational handoff, not a safety or recall procedure. In a complaint tabletop, support preserves the complaint/lot facts, reaches the designated qualified incident lead and operations follows the established stock-hold instruction. The real product-specific procedure and authority must exist first. Record responsible person, approval authority, evidence, due date, status and escalation contact separately so a commercial sign-off cannot clear a technical blocker.

9 Compare interventions and buy useful evidence

Use six optional diagnostic questions as prompts: occasion, recognition, expected value, access, experienced value and commercial delivery. Do not average them into a score or automatically invest in the lowest one. A weak metric is not the same as a responsive mechanism. The best next investment depends on the marginal effect of a feasible action relative to its cost, delay, uncertainty and risk.

For each candidate action, write:

• The exact change, person or trade account, product, channel and period.
• The change in behavior expected, and the mechanism proposed to produce it.
• The strongest rival explanation and what observation would separate them.
• The expected economic value as a range, with assumptions visible.
• Cost, cash commitment, time to learn and unacceptable downside.
• A current-practice or do-nothing comparator, plus the next-best feasible action.
• A result that would make the preferred action lose.

This process should yield a decision among actual alternatives. A team might compare better local availability with broader reach, reformulation with a clearer promise, a smaller trial pack with deeper discounting, or recruitment with replenishment support. It should not produce the same answer for every brand.

When uncertainty is too large to choose, ask whether a bounded test is likely to change the decision enough to justify its cost. Sometimes the sensible action is a small experiment; sometimes available evidence is already sufficient for a reversible change. A costly experiment that cannot distinguish the available choices has little decision value, even if it produces an impressive report.

Use evidence proportional to the commitment

A founder does not need a panel subscription or randomized trial for every action. Start with the cheapest reliable observation that could change the decision. Walk the intended stores. Attempt the purchase. Check the actual shelf price, temperature, pack, barcode and stock record. Ask buyers about their last real occasion and examine receipts or reorder records they are willing to share. Read the supplier's written minimum, lead time, terms and shelf-life specification. These checks often expose an actionable constraint more directly than a broad attitude survey.

Next use a small realistic trial when the product is technically ready. Sell through a manageable set of cafés, retailers or direct customers; record paid units, serving conditions, waste, actual availability and the cost of servicing them. Treat free samples separately. Interviews and first sales reveal useful facts, but a convenient pilot is not a market prevalence estimate. Avoid changing formula, pack, price, placement and audience simultaneously if the learning requires knowing which change mattered.

Use stronger comparisons as the cost of being wrong rises. A staggered rollout can be more feasible than a randomized market test. Comparable stores, repeated baseline observations and an unchanged offer can improve learning, although selection and concurrent events remain possible explanations. A randomized test is valuable when feasible, proportionate and capable of answering the decision. It is not a ceremonial gate before fixing a broken locator or rejecting an unaffordable production run.

The evidence ladder is about what a finding permits you to infer. Product interest is weaker than fulfilled paid choice for predicting willingness to pay. One paid order is weaker than observed replenishment for predicting a recurring business. A matched comparison is stronger than an unexplained before-and-after change for attributing effects. Do not discard low-cost evidence; use it for the decision it can actually support.

Bound the decision when precision is unavailable

Maintain an assumption register: each important assumption, its source, a plausible low/base/high value, the action sensitive to it, and the date it must be checked. A range is useful only if it is grounded; invented confidence intervals are no improvement over an invented forecast. A break-even calculation can show which assumption matters without pretending to know its true value.

Prefer reversible bets when uncertainty is high. Choose a narrow assortment, a limited geography, adjustable creative or a capped pilot where possible. Define a stop-loss in money, inventory, time or service risk. State in advance whether an inconclusive test earns more research or stops because further information costs too much. A 100-day plan can organize learning and commitments; it cannot guarantee product-market fit or successful sales.

The worked cases later in this book use invented figures to demonstrate the method. They are not industry benchmarks, recommendations for a named brand or actual client results.

Define the smallest worthwhile effect

Translate an intervention's all-in cost into the additional outcome it must plausibly create. Fictional example: a local program costs $600 and each genuinely incremental paid can contributes $0.40 before that program's cost. It requires 1,500 incremental cans merely to cover the program, before any additional fixed overhead. If the reachable market cannot plausibly supply that quantity in the relevant window, a large experiment is unlikely to rescue the plan.

This calculation is not a predicted effect or proof that the contribution will remain constant. It shows what evidence would matter. The team may accept a smaller immediate return for a specifically valued learning purpose, but that purpose and spending limit must be set beforehand.

Do not confuse a diagnostic with the intervention

People reporting poor recognition does not specify whether to buy broad reach, change the pack, improve shelf position or repair category legibility. Those actions have different response curves and costs. Similarly, low repeat does not identify whether experience, price, stock, cadence or buyer composition is the cause. A diagnostic narrows the question; it does not approve a budget.

Compare at least the current practice, the preferred action and the next-best feasible action. Include implementation fidelity: what has to happen for the action to exist as designed? A shelf sign left in a stockroom or an extra worker diverted to another task cannot test the intended intervention.

Decide what an inconclusive result earns

An imprecise result is not proof of no effect. It is also not automatic permission to keep spending. Before the test, decide which plausible result would justify rollout, another bounded learning step or stopping. If the available sample cannot distinguish economically different decisions, simplify the choice or use direct operating evidence rather than paying for false precision.

When a result contradicts the hypothesis, preserve the original record. A new explanation can be reasonable, but it needs a new prediction. Renaming the bottleneck after every outcome makes a framework impossible to learn from.

Work a complete comparison before calling a winner

Fictional research record CB-TEST-01 concerns an in-market lunch offer, separate from every production example. Sixteen eligible stocked cafés form eight pairs matched before assignment on prior movement and format. Within each pair, random assignment selects the revised menu description or current description. Hold product, $3 price, promotion and the 14-day window fixed. The unit assigned is the café; analyze one 14-day paid-can total per café, not hundreds of independent transactions. Assume pairs are independent, no customer spillover and complete POS capture for this teaching calculation.

All eight pairs remain in the result, including two revised cafés and one current café with stockouts. Totals are 200 current and 222 revised. Pair differences average 2.75 paid cans per café over 14 days. Their sample standard deviation is 4.7734; standard error is 4.7734/sqrt(8) = 1.6877. A paired t interval with 7 degrees of freedom uses the differences as its observations: 2.75 +/- 2.3646 x 1.6877, or approximately -1.24 to 6.74 cans at 95% confidence. [E07] This small-sample interval assumes the independent pair differences are reasonably compatible with a normal model; eight pairs cannot establish that assumption reliably. A methods reviewer should check design and sensitivity before consequential use. Treating individual cans as independent would invent precision.

Predeclared economics: each incremental paid can contributes $0.50 before this change; rollout implementation costs $2 per café per 14 days. The minimum worthwhile effect is 4 additional cans per café. The estimate implies $1.38 incremental contribution before $2 cost, or about -$0.63 net per café. The interval includes harmful and worthwhile effects; it does not establish no effect or justify rollout. The separate $160 research cost is already spent and remains in the whole-program account.

An operating log also records one objectively wrong purchase-location link in a revised café; a dated check and correction can establish that this defect was repaired. That does not establish the menu's demand effect. Keep the café in the assigned comparison. Decision owner: marketing lead. Day 15 action: retain current rollout, verify the link repair, and price a better-powered or better-controlled comparison only if it could change a material next commitment. Do not spend automatically to make an inconclusive result significant.

Bound what learning can be worth

Separate fictional CB-LEARNING-01: a run would earn $2,000 in a strong-demand state and lose $6,000 in a weak state. Invented probabilities are 60% and 40%. Expected result is 0.60 x $2,000 - 0.40 x $6,000 = -$1,200, so doing nothing at $0 is preferable on this expected-value objective. Perfect information would permit the run only in the strong state: expected value $1,200 before information cost. Thus $1,200 is the most information could be worth here; a $1,500 test cannot pay even if perfect. A $300 test still needs adequate accuracy and a useful decision rule.

Sensitivity: at an assumed 80% strong-demand probability, the untested run's expected result is +$400, so the preferred no-information decision changes. Perfect-information value is $1,600; its advantage over the $400 action is still $1,200. These are invented probabilities for learning the calculation, not instructions to invent confidence in a real forecast. Cash and unacceptable-loss constraints can override the expected-value ranking.

10 Build creative and creator programs that serve a decision

First choose the job of the execution

A creative asset can introduce a product, make an existing brand easy to retrieve, explain a reason to try, announce local availability, entertain an established audience or prompt replenishment. It can do more than one, but a team should specify the primary job before judging it. Otherwise a reminder ad will be criticized for failing to educate, or an awareness stunt praised as if it had acquired profitable customers.

Salience must connect to a retrievable offer

The more unusual the execution, the more important the transfer question becomes: what does the person remember afterward, and can that memory help a relevant choice? A strange collectible can be an excellent conversation object and a poor model for a daily beverage. A joke may spread because it is disgusting while depressing appetite for the liquid. High sharing is not automatically favorable consideration.

Use three tests: attention to the execution; correct brand and product attribution; change in the behavior or belief the brief names. A fourth commercial check asks whether the value of that change exceeds its cost. These measures should not be collapsed into a single engagement score.

Product as content and content around the product

A product-centered asset is useful when the object or experience contains the surprise. A surrounding-world asset is useful when the audience values a recurring story or activity. In both, the team must decide how the brand benefits. A sitcom can produce a loyal viewing audience that never becomes customers; a formula close-up can be commercially effective with modest public engagement.

The series is a practical unit of creative development. Recurring formats reduce production reinvention and make learning comparable. Vary the situation, payoff or objection while holding enough structure stable to recognize what changed. But do not run a format forever merely because it once produced a spike. Look for evidence of wear, audience saturation and diminishing contribution.

Many executions and large moments can coexist

A useful design question is how many distinct, relevant encounters a central idea can support. A physical activation can yield sampling, retail relationships, participant stories and paid creative. This does not make every byproduct valuable. Count the full production and operating cost, including content handling, staffing and logistics. Report the outcomes separately.

Practical recommendation: create a test brief containing audience, occasion, one expected change, fixed assets, variables, purchase route, permissible claims, cost and stopping rule. Test multiple executions of a clear proposition before declaring the proposition wrong; test a different proposition before infinitely polishing the same weak one.

Trust is specific to a question

A person can trust a creator's taste, doubt their health claims and still enjoy their jokes. Treating trust as a single reservoir obscures what proof is needed. A customer worried about excessive sweetness needs different information from a retailer worried about reorders. Useful proof addresses the actual uncertainty: a candid taste description, a visible preparation, relevant experience, clear ingredients or reliable service history.

Four separate ways creators can contribute

A creator can supply reach, endorsement, creative production or a transaction channel. Those functions may overlap, but they require different selection criteria. A skilled video maker may produce effective paid assets without a large audience. A local organizer may create access to a relevant occasion without making polished content. A credible reviewer may be valuable precisely because they cannot be made to repeat the brand's script.

Match compensation and evaluation to the work. Paid distribution, a production fee, a disclosed affiliate arrangement and discretionary product seeding are not interchangeable. Define usage rights, geography, duration, claim boundaries, editing permission and the intended outcome. This is a business recommendation, not a universal contractual template.

Resolve the attribution dispute by naming the decision

Use detailed tracking when it changes an actionable decision and the measurement is sufficiently valid. An affiliate code can help administer commissions; it may not measure incrementality because loyal customers can use the same code. A geographic holdout may estimate total program effects better, but small brands may lack enough comparable markets. A mixed method can combine content diagnostics, declared acquisition sources, transaction records and time-bounded experiments while documenting uncertainty.

Avoid false precision. Assigning every purchase to the last clicked creator can reward the person who captured existing demand rather than created it. The opposite error is declaring measurement impossible and spending without a budget, objective or learning plan. Choose the cheapest reliable measurement that distinguishes the decisions actually available.

Paid and organic are financing and distribution choices

A creator-style video can be a paid ad. A brand's unboosted post can require substantial production cost. Product seeding uses inventory, shipping and labor. An event that generates earned coverage still costs money. The contrast between 'authentic content' and 'ads' is therefore misleading when it hides who was compensated and how distribution occurred.

For endorsements reaching US consumers, make material brand relationships clear in the endorsement itself. Payment, free product and relevant personal or business relationships can matter. Put the disclosure where people will notice it and inspect the actual post or video, not only the contract or profile. [E04] A brand should not seek the persuasive appearance of an independent recommendation while concealing a material relationship.

Community programs deserve equally specific measurement. Count participation, attendance, return and member interaction before using the term loyalty. Then examine whether purchases are incremental and economically sensible. A free event can succeed as community service without functioning as efficient customer acquisition; decide which purpose was authorized before evaluating it.

When content and checkout share a screen

A live demonstration or creator video can put discovery, product explanation and ordering close together. For a beverage, watching is still different from tasting, and placing an order is different from receiving and consuming it. A platform may shorten the path to purchase while leaving delivery cost, case commitment, damaged goods or replenishment unresolved.

Evaluate the offer after product cost, shipping, platform and payment charges, creator compensation, paid amplification, discounts and expected refunds, counting each cost once. Record who funds each incentive. Platform-attributed sales can administer commissions without establishing incremental demand. Where practical, compare the full program with a credible alternative and follow delivered paid orders through a mature observation window.

Decide whether the program's job is profitable first orders, new-customer learning or a bounded acquisition investment. Then check the later purchase route: another platform order, a local stockist or a different pack. Channel format does not remove the need for honest disclosure or an economic account.

Turn the strategy into a production brief

A brief should name the person and their prior familiarity, the occasion, the current alternative, the one change the execution should produce and the offer they can actually obtain. Specify the necessary brand and variant cues, permitted claims, purchase route, creative freedoms, production and media budget, rights and review responsibility. Give the creative team the problem rather than only a list of mandatory phrases.

If a film is intended to make an unfamiliar taste intelligible, show or describe the relevant experience. If an asset announces local availability, verify that availability before release. If the goal is recognition at a future occasion, test attribution and memory for the right product, while acknowledging that those are intermediate outcomes. Creative can legitimately serve different horizons; it still needs a stated job.

A practical creative learning loop

First vary executions within one proposition to learn whether the expression is weak. Then compare a different proposition when there is reason to doubt the offer's promised advantage. Changing audience, message, price, landing page and pack simultaneously can tell you whether a bundle works, but usually cannot tell you which component caused the difference.

Use a small number of interpretable comparisons. Record the cost of producing each asset as well as its distribution. Watch for frequency, saturation, audience mix and channel changes before declaring a format fatigued. If the purchase path is broken, a lower conversion rate does not identify a creative failure.

Celebrity and launch attention need commercial accounts

Reported evidence: Zevia's Q2 2026 release reported sales up 1.1%, volume down 3.7%, positive adjusted EBITDA of $0.5 million and a GAAP net loss of $2.9 million. It also described stock-compensation awards associated with its Cardi B endorsement arrangement. [R11] These coexistence facts neither prove nor disprove the endorsement's effectiveness. The contribution of the program requires a comparator and its relevant total costs.

Chapter exercise: choose one proposed asset and write how it could receive high engagement while failing its business objective. Then identify a feasible measure or comparison that would reveal that failure.

Learning exercise: take the best-performing post and identify the least glamorous explanation for its result. Was it posted by a much larger account, supported with spend, driven by a giveaway or tied to a famous person? Design a test that keeps that factor from masquerading as a creative insight. Learning exercise: for each proposed creator, identify which of the four functions you are buying. Write what would make the arrangement unsuccessful even if its public engagement were high.

A completed marketer and creator brief

Fictional production brief CB-CREATIVE-01 uses the book's lightly sweet citrus lunch proposition. This is a production example with no campaign performance results, actual partner, permission or verified stockist. Its $240 concept-development cap is separate from all batch and research budgets. The fictional marketing lead can request internal concepts; publication, product release and new spending require the actual owner's approval.

Consumer execution A: open on the cold can beside a sandwich, show the front panel and pour, then show the verified café and total price. Draft voice line: 'A lightly sweet citrus drink with lunch.' That sensory description is conditional on actual product evidence. Execution B: open on the meal pairing, then show the same pack, price and purchase route. Neither says healthier, improves digestion or promises performance. The examples show what changes and what remains fixed; a real stockist name cannot be supplied until verified. Low-resource diagnostic: show one concept once to a willing eligible reader unfamiliar with it, then ask, 'What is the drink, which variant is it, what would it cost, and where could you buy it?' Record the person's answer before correcting it. A wrong product or unavailable purchase path is a repair signal. This small comprehension check cannot estimate demand lift or choose a statistically reliable winner.

Trade execution: a one-page café card leads with 'What must fit your lunch service?' It shows case quantity, storage/handling requirements from approved specifications, shelf space, ordering/replenishment contact, price basis and the operator's margin before its own costs. Unknown terms stay visibly unconfirmed. It does not claim a café partnership or offer unapproved service terms.

For a hypothetical paid creator post, the brief specifies the spoken and on-screen opening 'Ad: paid partnership with [brand]' using the actual contracting brand, and a matching prominent caption disclosure. Keep it readable and audible in the final context; do not bury it after more text or rely only on a profile. Free-product-only arrangements need truthful disclosure of that actual relationship instead. The creator and marketing reviewer inspect the exported and posted execution, including overlays and cropping. A draft contract clause is not evidence that disclosure appeared. [E04]

Listening record for the first authorized run: date/surface __; asset version __; recurring question in permitted anonymous wording __; observed purchase-path failure __; available platform measure and definition __; qualified inquiry or verified order __; revision owner/date __. No results are filled because no run occurred. If questions repeatedly concern sweetness, inspect the product/expectation match; if the locator is wrong, repair it; if attention rises without usable demand evidence, do not declare a sales effect. Useful answers can build trust over the declared horizon without each organic asset proving same-day incremental purchases.

Data handoff: share aggregate questions or participant codes where sufficient. The marketing lead controls access and a declared deletion date; raw contact details, recordings and health comments do not go into creator briefs or AI tools without an appropriate purpose and permission. [E06]

11 Use sampling and live experience to test a real choice

Sampling buys a particular kind of information

A sample lets a person experience some part of the offer at reduced risk. It can reveal taste preference, remove a misconception, demonstrate preparation or place the drink inside a memorable occasion. It does not automatically reveal willingness to pay, full-serving satisfaction or repeat. A person accepting something free is not equivalent to an acquired customer.

A good sampling program specifies who should try, what condition they should experience, what they should learn and where the next purchase can happen. Product temperature and serving conditions should resemble intended use. Record the sampled SKU and the recipient's eligibility for the research or offer without collecting unnecessary personal information. When sampling involves regulated products or age-sensitive contexts, use a qualified compliance process for the jurisdiction; do not transfer an ordinary soda activation unchanged.

Build a bridge to the next occasion

A QR code is not a strategy by itself. A bridge might be a verified nearby stockist, an on-site paid unit, a relevant café menu, a trial pack or a permission-based reminder at likely replenishment time. The best choice depends on the setting and consent. Avoid turning a pleasant sample into a long data-capture obstacle unless the information genuinely improves service or measurement.

Measurement should follow the bridge. Count eligible contacts, actual samples, meaningful conversations, paid purchases and later reorders separately. A single attendee may appear in multiple counts; report unique people where possible and explain the de-duplication method. Exclude staff and ineligible people from an explicitly defined participant denominator, and deduplicate repeat contacts. Eligible sample recipients remain in the sample-to-paid conversion denominator; free sample or giveaway units do not count as paid purchases in the numerator. A scan measures a scan, not consumption or incremental demand.

The event has several possible customers

An activation may serve attendees, creators, press, retail buyers and the brand's own distribution partners. Their outcomes differ. A spectacular installation can be justified as media production or relationship development even if on-site sales are modest, but that objective should be set before the event. Otherwise any result can be relabeled as success afterward.

A small local test can be rigorous. Match participating stores or events on relevant characteristics, randomize where feasible, and keep price and availability comparable. Measure a baseline and allow enough time for the intended purchase behavior. If a national campaign, weather change or stockout differs across groups, document it. A weak design should yield a tentative learning, not a precise return claim.

A worked sampling decision

Illustrative scenario: an adult lunch soda brand can fund either one visually ambitious event or repeated sampling outside existing stockists. The team believes unfamiliar taste is the main barrier. It should first establish that the relevant shoppers can find the drink and that the offer is acceptable. If access is broken, neither sampling option solves the next purchase.

Then pilot sampling near a subset of comparable stores, record actual availability and evaluate incremental unit contribution after product, labor and promotion costs. Gather short, specific feedback from both purchasers and nonpurchasers. If people like the sip but reject the price or do not finish a can, the next test should address the offer or product, not merely increase sampling volume.

Reconcile an activation before interpreting it

Fictional example: a team hands out 240 sample portions to 200 eligible unique adults. Forty people receive a second portion. Later, 30 of the 200 recipients make a verified paid purchase in the observed route. The recipient conversion proportion is 30 divided by 200, or 15%; it is not 30 divided by 240. Without a comparator, that is observed conversion, not the incremental effect of sampling.

If those purchases contain 42 cans, record 42 paid units separately from 30 purchasing people. Do not add the 240 free portions to paid units, and do not exclude recipients who did not buy from the denominator. If some recipients cannot be followed or purchases in other channels are invisible, disclose that measurement limit. A QR scan may help locate a person in the process but does not establish a purchase.

Filled event record CB-SAMPLE-01: 240 portions; 200 unique eligible adults; 30 verified purchasers; 42 paid cans; 15% recipient conversion. Keep a coded eligibility/receipt method and its follow-up window alongside the counts; those details are unspecified here, so the record cannot establish a causal effect or complete channel coverage. Collect no identifying information merely to make the sheet look complete. Define purpose, access, retention and sharing destination before collecting or linking records. [E06]

Decide whether the event is worth repeating

Count product, samples, staffing, travel, venue, fabrication, rights, promotions, capture and measurement. State which outcomes justify those resources: paid demand, trade relationships, useful research or a defined media asset. Several purposes can coexist, but the budget and decision should not be rewritten after the largest number arrives.

Proposed test: compare a simple sampling program near stocked outlets with the same program plus accurate purchase information. Keep the monetary offer and staff attention similar where practical. Include a no-event comparator if the question is whether sampling itself pays. More locator clicks without worthwhile incremental contribution would fail the commercial bridge, even if the activation was enjoyable.

Learning exercise: write the sentence 'After this experience, the person can...' and finish it with a real action in a real place. If the answer is only 'remember us,' decide how that memory will be tested and why it matters.

12 Measure recruitment repeat and portfolio value

Define the objective before the metric

A brand is choosing between two acquisition approaches with the same $1,000 budget and comparable assigned audiences. In this invented example, A recruits 200 first buyers and 40% buy once more within a mature 90-day window. B recruits 100 first buyers and 70% buy once more. Assume each first order yields $3 contribution before the campaign cost and each repeat yields $4, with no other orders observed in the window. For this arithmetic example, also assume a valid comparison has established these as incremental buyers and orders; a real team must not assume that attributed orders are incremental.

A produces 80 returning buyers and $920 contribution before its $1,000 campaign cost: 200 × $3 plus 80 × $4. Net campaign-period contribution is negative $80. B produces 70 returning buyers and $580 before the same cost, or negative $420. B's superior repeat percentage does not make it the better acquisition investment. A also produces more returning buyers in absolute terms.

The correct choice is not automatically to scale A. If the declared objective requires nonnegative 90-day contribution, both approaches fail it. The next decision could be to improve the offer, reduce acquisition cost, change the target audience or stop. If later purchases are expected to justify the investment, that forecast and cash constraint must be stated rather than silently added after an unfavorable result.

What the current evidence changes

Zevia's May 2026 presentation reports higher purchase frequency alongside slightly lower household penetration in its annual panel comparison. [R02] That does not identify retention within a fixed starting cohort. The decision rule is to identify whether recruitment, frequency, retention, quantity or value realization is limiting the declared opportunity, then compare interventions.

The two experimental approaches in the teaching example may also recruit different kinds of people. That is part of their total effect, not a nuisance to remove by comparing only the buyers who happen to look alike after treatment. Report per-assigned-audience or market results alongside purchaser-cohort diagnostics.

Timing and portfolio checks

A single cold can and a 24-pack have different replenishment opportunities. Late entrants need enough observation time. Stockouts and invisible sales in another channel affect measured return. A limited flavor can introduce buyers to the core line while never repeating itself. Define the relevant brand or SKU boundary and report what the data cannot see.

Practical founder version: maintain a simple paid-customer or outlet register where permissible, record dates and quantities, and ask about actual use. Do not buy a large panel merely to label the first small cohort. Escalate measurement when the decision stakes and expected value justify it.

Repeat is a behavior with a denominator

A repeat-purchase rate is uninterpretable without a unit, starting event, eligibility rule and time window. Is it people who bought twice, orders from returning customers, households that repurchased a SKU or retail locations that reordered? Did all first-time buyers have the same opportunity to repeat? Were free samples included? Was the product continuously available?

Consider an invented cohort of 100 first-time buyers. Forty buy again within eight weeks. The raw repeat rate is 40%. Suppose 20 of the 100 encountered a verified local stockout during most of the period and none repurchased through observed channels; the remaining 80 had verified continuous availability. It may be useful to report 40/80, or 50%, for the continuously available subgroup, but do not replace the original 40/100. Removing failed-access customers changes the question and can introduce selection bias. Report both and explain who is excluded.

Equal time matters. A customer acquired yesterday has not had the same opportunity to reorder as one acquired eight weeks ago. Compare mature cohorts, or use an appropriate time-to-event analysis with clear assumptions. A large case, occasional use or a seasonal product can have a longer natural interval than a single daily drink. Low short-window repeat is not automatically poor product fit.

Novelty and the base business

For a limited release, ask whether it recruits incremental customers, introduces them to the core range or profitably serves existing enthusiasts. Track core-product purchasing among the new cohort, accounting for other exposures where possible. A stunt SKU with poor own-SKU repeat might still succeed as an acquisition or media investment. That success must be demonstrated rather than assumed from shares.

Make returning easier without coercion

A sensible repeat offer may be a convenient multipack, a reliable local stockist, useful preparation advice or an optional replenishment reminder. A subscription should match actual consumption and make control clear. Forcing a recurring commitment before customers understand their rate of use may inflate near-term orders while creating cancellation, waste and resentment.

Recommendation: distinguish first-to-second purchase, second-to-third purchase and long-term retention by cohort, occasion and channel. Examine the reasons people stop: disliked experience, price, unavailable SKU, changed routine, too much inventory, changed needs or preference for alternatives. A different remedy follows from each reason. Do not use a single motivational email sequence to fix all of them.

Build a cohort table that answers a decision

For each cohort, record its first paid event, dates of eligibility, acquisition method, channel visibility, initial pack and offer, number of buyers, subsequent paid buyers and orders, contribution and follow-up period. Keep free trial in a separate record. Report a mature 90-day cohort separately from one observed for only 14 days. A new subscription order and a successfully delivered, retained order are different states.

A brand-level view may count a switch from the launch flavor to the core flavor as retained demand; a SKU-level view will not. Choose the boundary before interpreting the result. Retailer reorders can indicate useful movement but are not individual consumer retention. Explain whether a repeat measure represents people, households, orders or outlets.

Learn from people who do not return

Ask about real use and the next relevant choice, rather than assuming every nonreturner disliked the drink. Possible explanations include poor taste, a changed routine, too much inventory, high price, inaccessible stock, a preferred substitute or an infrequent occasion. Different remedies follow. Where possible, check the purchase path and actual records rather than depending entirely on recall.

Do not remove inconvenient people from the main result to reveal the perfect customer. Diagnostic subgroup views can be useful if clearly labeled and paired with the full assigned population. An acquisition program that changes buyer composition may lower conditional repeat while producing more total profitable returning buyers.

Chapter decision: identify whether the next opportunity is recruitment, return, frequency, pack quantity, realized value, access or stopping an unproductive offer. Choose the intervention using its likely incremental value, not the attractiveness of a single rate.

Learning exercise: choose one repeat metric you currently report. Write its exact denominator and eligibility period. Then identify which important customers or purchases the data cannot see.

13 Work the founder case from pilot to production

The strategic proposition

A fictional founder wants to sell a lightly sweet sparkling soda with weekday lunch. The initial drinker is an adult choosing a cold single at a sandwich café. The main substitute is the café's conventional soda, rather than every functional beverage online. The proposed advantage is a preferred meal pairing and a lighter sweetness experience at an acceptable price. No medical benefit is claimed.

The founder's possible capability is local service: visiting a compact group of cafés, helping them keep the right cans cold and learning directly from staff and shoppers. The business cannot initially match national distribution or sustain a large content operation. The strategic bet is therefore local meal choice and dependable service, not mass fame.

Assume for this teaching example that the pilot product has passed the relevant technical and claims review. This is a condition of the example, not something a real founder should take for granted.

The offer and constraints

For the first commercial batch, require nonnegative batch contribution over the 36-week selling window and enough cash to meet commitments. This is a minimum screening condition, not a complete return-on-investment target. The café shelf price is $3.00 per can. It pays the brand $1.80, leaving $1.20 before the café's own labor, rent, waste and other costs. A quoted full-scale run costs the brand $0.80 per produced can for product and pack. Local delivery and service cost $0.20 per paid can. The first commercial batch also requires $1,200 of fixed account-acquisition and launch support. Reserve 10% of the run for samples and expected losses. These figures are assumptions to verify, not standard margins.

The full-scale minimum is 12,000 cans per flavor. Cash available for the project is $20,000. Three flavors would require $28,800 for product and pack alone, before other commitments. That attractive launch is ruled out now by a cash constraint. A good-looking campaign cannot fix that arithmetic.

One flavor uses $9,600 for product and pack. After the 1,200-can reserve, 10,800 cans remain available for paid sale. Allow 36 sellable weeks after delivery, handling and shelf-life buffers. Clearing that paid inventory requires 300 cans a week. That is an inventory-clearance target, not the economic break-even point.

The rival explanations

The founder's first theory is that people want the taste but have not encountered it. Another is that interest disappears at a $3.00 paid choice. A third is that the café will not allocate cooler space or staff attention at the proposed economics. Social enthusiasm cannot distinguish those explanations.

The available next steps are a full single-flavor run, a smaller paid learning run or stopping until the offer changes. For this example, the supplier also offers a technically compliant 1,000-can micro-run at $2.40 per can. Add $600 for pilot delivery and staff support, giving a complete assumed pilot-spend cap of $3,000 before sales receipts. No additional pilot cost is assumed outside that cap. The pilot has worse unit economics, but buys information before a larger, less reversible commitment.

The decision before the result

Choose the capped pilot, subject to obtaining the agreed café placements and confirming the product can be served as intended. Set its purpose as learning whether the paid lunch offer and trade route justify scaling. Do not disguise the pilot as profitable steady-state production. Its stop-loss is the $3,000 gross outlay; no second run is automatic.

Use five cafés over three weeks. Record each site's actual paid units, cold-stock availability, menu price, samples, waste, reorders and service time. Observe the alternative customers choose. Ask short questions after an actual paid or rejected offer, avoiding promises about health or future purchase. This convenience sample does not establish a national rate, but it can invalidate an unrealistic initial route.

A promising result that still does not justify the full run

Suppose verified café point-of-sale records show 350 paid one-can consumer purchases across 15 café-weeks, alongside 100 free samples. Stock delivered to a café is not counted as a consumer purchase. Reconcile the remaining 550 cans with actual waste and stock on hand. Within the remaining pilot budget and verified use-by/handling limits, use usable stock to test additional cafés before ordering more. Paid velocity is 23.3 cans per café-week; samples are excluded from the paid numerator. The denominator for any sample-to-paid measure would be eligible unique sample recipients, which this store-total data alone does not identify.

All five cafés request another order and customer comments are positive. Those are useful signals, but not consumer cohort retention. The founder has measured outlet-level movement and retailer interest, not how many individual people bought twice.

At the observed average, five cafés would move roughly 4,200 paid cans over 36 weeks if the rate held. Sale-linked collection teaching assumption: for this batch model, assume the brand collects $1.80 for each paid consumer can and gives no credit for unsold café stock; actual wholesale terms and timing require a separate cash check. Receipts after variable delivery/service are $1.60 per paid can. With no recovery value for leftovers, batch contribution before other overhead is $1.60 times paid cans minus $9,600 production and $1,200 fixed support. Break-even is 6,750 paid cans. Five cafés at the pilot rate would leave a $4,080 batch shortfall.

Ten equally productive cafés would imply 8,400 paid cans and $2,640 batch contribution before other overhead, despite not clearing all 10,800 saleable cans. Fifteen cafés would need 20 paid cans per café-week to clear that inventory. Thus a founder may rationally accept some leftovers when the total economics work. But the extra cafés are not yet secured, their productivity may differ, and positive modeled contribution does not eliminate cash-timing risk.

These scenarios retain the case's $1.60 per paid can, $10,800 committed batch cost and zero recovery value assumptions. Café rates are extrapolations to test, not confirmed demand. The reserve is outside saleable inventory.

Decision: the observed five-café route does not justify this full run under the stated financial objective. Do not place it merely because the pilot tasted good and every café reordered. Compare three concrete repairs: negotiate a smaller commercial minimum, secure and validate more suitable outlets, or alter the offer/route to obtain enough contribution and throughput. If none is credible within the cash and shelf-life window, stop or redesign. The framework has ruled out a larger bad bet before that bet's outcome is known.

What would change the decision

Evidence of stronger paid velocity in comparable cafés, credible additional placements, a smaller minimum or longer verified selling window could support a run. Merely adding followers would not resolve this particular inventory constraint. Conversely, poor full-serving experience, unacceptable retailer waste, delivery labor beyond the estimate or repeated cold-stock failures would push toward product/format/service repair before broader recruitment.

Seasonality matters. A warm-weather launch week cannot automatically forecast winter lunch demand. Competitors may cut price or increase placement. Use a conservative scenario and revisit the decision when those conditions change. A pilot can support a bounded next commitment without proving the whole business.

Keep three boundaries visible

This case has a 1,000-can learning run and a separate 12,000-can commercial production option. Pilot unit costs do not become commercial unit costs, and the pilot outlay does not disappear from a complete project return. The batch calculation evaluates whether the next commercial commitment can cover its specified costs. A full project account must also include the earlier pilot's actual receipts and costs.

The production boundary is 12,000 cans produced, 1,200 reserved for samples and expected losses, and 10,800 available for paid sale. The economic model is batch contribution = $1.60 × Q - $10,800, where Q is paid cans under the stated collection assumption and no recovery value is assigned to leftovers. The break-even quantity is 6,750. The inventory-clearance quantity is 10,800. These answer different questions.

The cash boundary is the dated money paid and collected. The case cannot establish that cash survives merely by showing a positive batch contribution. Wholesale payment terms, unsold stock, service costs and collection timing need explicit checking. Its recommendation remains conditional on those real facts.

Other batch exercises require their own quantities, costs, reserves, collection assumptions and decision purpose. Do not transfer this case's break-even number to a different run merely because both use cans.

14 Evaluate an established business investment

A current contradiction changes the rule

Starbucks reported North American comparable transactions up 4.4% in its quarter ended March 29, 2026, while segment operating income fell 9%; management discussed investment and input costs. [R01] This does not prove that service investment failed. It does reject the simple rule that improved customer transactions necessarily mean improved current operating profit.

The changed framework decision is explicit: do not approve an experience program solely because traffic rises, and do not stop it solely because early margin falls. Evaluate the program against its declared cost, ramp period, contribution and longer-term purpose, using a credible comparator where possible. The observation changes the decision rule; it does not supply the missing causal estimate.

Apply the distinction to a beverage operator

Consider a chain deciding whether to add staffing and a second cold-drink pickup point. The objective is to reduce lost orders and improve useful customer experience without creating an unacceptable operating burden. One hypothesis is that peak queues suppress beverage purchases. A rival is that price or product appeal, rather than service, limits demand. Another possibility is that the new service mainly shifts customers from another daypart or nearby outlet.

Before a large rollout, inspect queue times, abandoned orders, daypart sales and current staffing. Pilot the service in a feasible set of comparable locations, with a staggered or randomized comparison if the scale permits. Record implementation fidelity: a nominal extra worker who is repeatedly diverted elsewhere is not the planned treatment.

Primary decision measures are incremental contribution after the actual labor and operating cost over the agreed window, together with service and safety guardrails. Diagnostics include transactions, ticket, order mix, waiting and repeat visits. Unique customers and visits are different; a visit increase is not proof of broader penetration.

Failure and transfer boundaries

The intervention loses if it improves the intended service measure but does not create enough value for its cost within the declared purpose and horizon. It may remain inconclusive if the pilot is too short to observe a plausible return or implementation was incomplete. Any extension requires a reason and another bounded commitment, not an indefinite promise that brand value will eventually appear.

A listed multinational's segment result is not a forecast for a five-café founder. It contains multiple simultaneous changes and cost effects. The founder can transfer the logical distinction between customer behavior and operating return, not the company's effect sizes, budgets or spending priorities.

Distinguish a strategic investment from a permanent exception

A team can accept current costs for later benefit if the rationale, horizon, funding and milestones are explicit. Useful milestones should provide evidence about the future return, not merely show that activity occurred. If service improves but the proposed downstream behavior repeatedly fails to follow, revise the mechanism or the offer. An indefinite statement that brand investment takes time cannot rank this action against the next-best use of money.

Chapter exercise: define the early behavioral evidence and the later economic result for a proposed experience investment. Write the maximum loss and the observation that would stop or redesign it.

15 Use recent cases to challenge industry stories

Public evidence changes the model when it rules out an inference

These cases are selected boundary tests, not a representative industry sample. Each asks which inference the reported evidence rejects, what remains unresolved and how the decision changes.

Revenue growth is not the same as more beverage volume

Reported evidence: Coca-Cola's 2025 global unit-case volume was flat while net revenue rose 2% and organic revenue rose 5%. Price/mix contributed 4%, and the company described timing in the concentrate-sales comparison. Coca-Cola Zero Sugar unit-case volume grew 14% while total sparkling soft drinks were flat. [R08]

Changed decision: separate quantity, price and mix before choosing a growth intervention. A strong variant can coexist with a flat larger category. These figures cannot identify a consumer-level elasticity or prove that a campaign caused the variant's growth.

Acquisition can dominate a portfolio growth headline

Reported evidence: Keurig Dr Pepper's 2025 US Refreshment Beverages sales rose 11.9%; volume/mix rose 9%, with 6.2 percentage points attributed to acquiring GHOST. US Coffee sales rose 0.6% while volume/mix fell 4.2% and net price rose 4.8%. [R09]

Changed decision: separate acquired scale from organic demand and retain the source's combined volume/mix definition. The coffee segment includes brewers as well as consumables. Do not treat either bridge as a clean experiment showing what beverage buyers prefer.

A new proposition is not a demonstrated success

Reported evidence: Coca-Cola announced Simply Pop in February 2025, including a prebiotic-soda offer, flavor and format details and a phased US route. Its launch explanation invoked the existing Simply brand and product-development work. [R13]

Changed decision: treat familiar-brand trust, the benefit promise and the route as hypotheses at launch. The announcement establishes the offer and rationale, not functional efficacy or profitable demand. Compare equally available, truthful offers and measure actual choice before attributing success to inherited trust.

A lower earnout value is a warning with a precise meaning

Reported evidence: PepsiCo's Q2 2026 filing reported Poppi's contingent-payment liability at $117 million on 13 June, down $161 million during the first half from $278 million at year-end 2025. The potential payment is $300 million if specified milestones are achieved by Q3 2027. Its valuation uses inputs including revenue and margin forecasts; the filing does not isolate the cause of this period's decrease. [R15]

Changed decision: investigate acquisition expectations and cash returns. This is an earnout remeasurement, not a reported brand impairment, operating loss or proof of consumer backlash. Finalized purchase consideration was $2.120 billion. [R15] That differs from the announced $1.95 billion transaction price and $1.65 billion after anticipated tax benefits. [R19]

An acquisition's return requires a transaction-specific account of operating cash flows, integration costs, remaining payments and alternatives. Retail sales are not manufacturer revenue. No consumer repeat rate, campaign effect or health benefit follows from a purchase price.

Ownership appeal and route power need different evidence

Proposed test: first measure whether relevant buyers know the owner. Compare truthful ownership information with neutral information while keeping taste, price, availability and presentation stable. Observe paid choice and later behavior, distinguishing prior buyers from newly reached shoppers. Online complaints alone cannot establish a commercially material loss, and growth alone cannot rule out an offsetting loss among some buyers.

Separately examine brand ownership, bottler affiliation, territorial rights and physical delivery. KDP's licensed bottling arrangements demonstrate that brand competition and route cooperation can coexist. [R17] They do not supply an industry-wide distributor concentration statistic. A claim that most beverages travel on two companies' trucks needs a defined geography, category, channel, denominator and actual route data.

Younger shoppers are not one rejection story

Reported evidence: an NIQ-authored February 2026 presentation reports that Gen Z dollars per buyer rose for soft drinks by 3% and energy by 16% in the 52 weeks ending 29 November 2025, using its total-US omnishopper panel measure. [R16]

Changed decision: investigate a changing repertoire rather than assume wholesale rejection of established beverage categories. Spending per buyer is not volume, penetration or personal consumption. The broad soft-drink group does not separate conventional from modern soda, and per-buyer spending can rise while fewer people buy. This is purchase-based evidence with limits, not a causal explanation of a generation.

A variant recovery does not settle the original product

Reported evidence: PepsiCo's 9 July 2026 remarks reported that Mountain Dew Zero Sugar gained both dollar and volume share in Q2, while PBNA organic volume fell 4%. [R18] The variant and segment boundaries are different, and neither isolates original Mountain Dew.

Changed decision: request original-SKU velocity at comparable availability, price and promotion; examine blind and branded experience, switching and trademark composition. New flavors can add demand, transfer buyers or coexist with erosion in the original. A visual judgment about a brand refresh is not a measured product diagnosis.

Founder influence and a continuing business can coexist

Reported evidence: Celsius's February 2025 acquisition announcement describes Alani Nu founder Katy Hearn as an entrepreneur and influencer. [R20] The company's August 2026 results report Alani retail sales growth of 55.7% year over year in US tracked RTD-energy channels for the 13 weeks ended 28 June. Q2 brand revenue was approximately $364.4 million; management cited consumer demand, increased orders from its largest customer during the distribution transition and a limited-time flavor among the factors affecting it. [R04]

Changed decision: a founder-linked brand should not be treated as a clean comparison against creator businesses. Evaluate product, distribution, price, acquisition and continued purchasing on their own evidence. These results show commercial growth in specified measures; they do not disclose the cohorts needed to prove that creator attention caused trial or that repeat caused the growth. Several mechanisms can operate together, and one company's outcome does not establish a general winner formula.

Stop where the evidence stops

These cases reject shortcuts and define useful next evidence; they do not establish universal investment rules or validate the full framework. Results after 6 October 2026 are outside this edition's cutoff.

16 Adapt the decision to category and channel

The framework is useful only when its variables fit the actual choice and operating system. The playbooks below are original recommendations, not measured category laws. Use them to identify which assumptions need checking, and replace them when a simpler channel-specific model makes better decisions.

Water and sparkling water

Refreshment may be sufficient as a starting need. The offer can vary through cold convenience, carbonation, flavor, meal pairing, price or expressive packaging. Identify the real substitute and the sacrifice of choosing the proposed offer repeatedly. An entertainment-led identity may add value for some buyers; a quiet, convenient product can also fit the occasion.

Compare blind and branded response, then test a realistic purchase. If the premium draws interest but paid choice is weak, investigate price, pack and expectation before increasing spectacle. In grocery, household stock and multipack economics may dominate; in a venue, the default and cold service may dominate.

Soda and lower sugar soda

The occasion may be pleasure with food, an afternoon treat or a familiar flavor. A functional promise can support choice and also create taste questions or substantiation requirements. Test the actual drink against the alternative people select in that setting, including the regular price and full serving. A favorable isolated sip does not settle a meal choice.

Do not assume one-way migration out of conventional soda. The recent Coca-Cola figures distinguish a growing no-sugar variant from a flat sparkling portfolio. [R08] A startup needs evidence of its own occasion and reachable demand, rather than a broad category narrative standing in for the business case.

Energy and performance beverages

Separate desired stimulation, taste, occasion, format, identity and the relevant eligibility and safety constraints. Verify product and claim suitability with qualified specialists. A first-time category buyer and a habitual category switcher can have different reference points. Celebrity exposure can attract attention without resolving either person's offer evaluation.

Track paid purchase, actual availability, experienced use and economics at the correct brand and channel boundary. Current CELSIUS reporting shows why a growing corporate portfolio, a flagship's shipments and retail-dollar movement need separate accounts. [R04] Those figures do not diagnose the cause of a particular formula's acceptance.

Hydration powders and ready to drink formats

A powder changes portability, storage, mixing effort and freight compared with shipping a ready-to-drink product. The ready-to-drink form removes preparation but can add pack, storage and route requirements. Taste and preparation consistency belong in the test. A recognized ingredient does not prove that the format solves a sufficiently valuable problem.

Measure the actual use cadence and basket. A large sachet box replenishes differently from a single cold bottle. Concentrated exposure to one large account can change access abruptly, as the Liquid I.V. assortment commentary illustrates, without revealing the complete demand picture. [R14]

Protein and nutrition drinks

The use may concern a snack, convenience, exercise or a meal component. The marketing task is to establish the intended benefit honestly and test texture, flavor, serving satisfaction, tolerability within appropriate review and willingness to choose the offer again. Do not extrapolate from enthusiastic gym sampling to every adult household.

Test realistic storage and serving conditions, the cost of the intended repeat basket and the trade's handling requirements. Qualified technical and claims review remains separate from a favorable preference test. The book provides no clinical benefit claim or standard formulation requirement.

Coffee tea and adult social alternatives

Routine, preparation, temperature and serving ritual can matter as much as novelty. A decaffeinated coffee may fit a different daypart; a tea can be valued for a particular sensory ritual; an adult social alternative can depend on the venue, glassware, food and group occasion. Treat each as a distinct proposition to test.

Keep ordinary soft drinks, non-alcoholic beer, non-alcoholic spirits and intoxicating products analytically separate. Their formulations, eligibility, handling and regulatory requirements differ. Do not transfer an ordinary soda activation into a regulated category without qualified local review. No trend claim about one of these categories establishes the others' demand.

Foodservice and on premise use

The guest buys a serve while the operator buys a workable menu and service system. Test preparation speed, staff understanding, waste, consistency, storage, margin and guest choice. A concentrate can reduce freight yet create dilution or training work; an individual pack can simplify service while changing cost and presentation. Neither format wins by definition.

Oatly's reported North American Q1 2025 volume and adjusted earnings moved in different directions. [R03] That supports a route-specific economic review, with the report's non-beverage and segment limitations retained. It does not tell a new brand which foodservice agreement to accept.

Direct to consumer and marketplaces

The shopper encounters a page, an offer and a delivered pack rather than an immediate cold single. Explain the product, variant, quantity, total delivered price, expected handling and purchase commitment clearly. Inspect fulfillment and damage as part of the experience. Separate attributed platform return from incremental contribution.

A case can improve shipping economics while raising the trial barrier. A sample pack may reduce commitment while increasing cost per serving. Compare the total acquisition and replenishment account, including observed repeat and the cash needed to hold stock. Do not fund a weak first-order offer with an untested lifetime-value story.

Formulation and manufacturing services

A service buyer commissions a consequential project. They need clarity about fit, scope, deliverables, technical responsibility, timelines, dependencies, evidence and handoff. Helpful content can make an unfamiliar process understandable; view counts are not evidence of qualified demand or successful delivery.

Measure qualified inquiries, fit discussions, proposals, wins, delivery quality and referrals with permission and appropriate data handling. Separate exploratory advice, paid development and production commitments. This playbook assumes no named firm's client relationships, costs or performance.

17 Run a launch and a weekly decision review

The one page decision record

A useful framework leaves a clear decision record. Complete these fields in ordinary language before a consequential commitment.

• Decision: what must be chosen now, by whom and by when?
• Strategy: which buyer, occasion, substitute, advantage and route are being pursued?
• Unit and outcome: person, household, outlet, SKU, brand or portfolio; what changes over which period?
• Evidence: what has actually been observed, what is reported by someone else and what remains an assumption?
• Rivals: what are the strongest competing explanations?
• Options: current practice, the preferred intervention and the next-best feasible alternative.
• Constraints: product readiness, claims, route permissions, service, supplier minimums, cash and time.
• Economics: expected incremental value, relevant cost, inventory exposure and sensitivity threshold.
• Learning: the cheapest credible check or test that would change the choice.
• Failure and revisit: what would make the preferred option lose, how much can be lost and when is the decision reopened?

Use the six diagnostic questions to discover omissions, not to substitute for this record. If the final recommendation is identical regardless of the evidence, the framework is not being used to make a decision.

A finished founder record

Fictional CB-FOUNDER-12000, reviewed after the chapter 13 pilot: WAIT on the 12,000-can commercial run. Five cafés produced 350 paid cans across 15 café-weeks; that is observed movement within the fictional case, not an actual customer result. At the extrapolated five-café pace, 4,200 paid cans yield -$4,080 before other overhead versus the required nonnegative batch result. Ten equally productive cafés are a scenario, not secured demand.

Owner: founder/pilot lead. Next action: within the next seven days, request a smaller written minimum and verify additional café terms, paid movement and service capacity. New commercial production commitment permitted by this record: $0. Existing usable pilot stock may be used only within the earlier $3,000 cap and verified handling/use-by limits. Keep that earlier learning spend in the whole-project account, outside the next-batch comparison.

Reopen only when a feasible route and production option can meet the declared 36-week economics, product/claims release and dated-cash requirements. Supplier and qualified release owners must supply their own evidence. Stop or redesign if no credible option does so. A hypothetical positive contribution scenario cannot approve production by itself.

A usable test and data record

A finished test record names decision/comparator, assignment unit, eligibility, intervention, period, primary measure, full cost, worthwhile effect, implementation failures, uncertainty and decision. Chapter 9 supplies a complete numerical example. For a real research file also specify purpose, minimum fields, participant information/permission, access owner, retention/deletion date and any agency or AI sharing destination. A blank permission field blocks that sharing. Prefer aggregate reporting; link identifiable purchase histories only where justified and permitted. [E06]

A practical evidence ladder

Level one is direct operating observation: a wrong locator, an actual failed transaction, a written supplier quote, a measured preparation time or a documented stock discrepancy. It can justify a repair or rule out an infeasible option without proving a general marketing law.

Level two is bounded real behavior: a fulfilled paid pilot, actual consumption feedback, retailer orders, repeat dates, waste and service cost. It establishes what happened in that setting. It may not establish the counterfactual or support a market-wide forecast.

Level three adds a comparison: a staggered rollout, matched stores, a stable baseline, a price/pack comparison or a randomized treatment where feasible. Stronger design improves attribution, but its value depends on implementation, sample, spillovers and the relevance of the outcome.

Level four is replication and transfer: the mechanism survives different periods, outlets or buyer groups with its boundaries specified. A famous brand's result is not replication of a founder's pilot. Recent publication is not evidence that the underlying behavior happened recently.

These levels do not prescribe endless research. A reversible $200 creative variation and a $50,000 production commitment deserve different evidence. Escalate the standard with the downside of being wrong, and stop researching when the expected improvement in the decision no longer justifies the cost or delay.

A popular teardown is a starting point for research. Likes describe a platform response to a post; they do not validate its explanation of a business. Trace consequential claims to the original report and identify the outcome, entity, period, comparison and strongest alternative explanation. Repeated stories about the same campaign are not independent replications.

A stated limit improves a claim by showing where it might fail, but it does not make an unsupported number reliable. Conversely, a directly observed purchase-path failure can justify a repair without an elaborate numerical case study. Match the evidence to the action and its downside, rather than requiring every useful observation to resemble a famous growth story.

Keep a dated decision context

Keep a dated context record. Capture promotions, major competitor moves, seasonality, weather when relevant, service disruptions and changes in audience or assortment. Do not use these as an unlimited collection of excuses. State which changes are large enough to invalidate a comparison or require a new test.

Consumption cadence also changes interpretation. Daily coffee, occasional mixers, event drinks and bulk pantry purchases produce different observation windows. Set the window from the relevant behavior and business decision rather than a universal 30-day rule. A seasonal return the next year is different from a failure to repurchase next week.

Competitive advantage must survive a response

Before scaling, ask what happens if a competitor matches the flavor, borrows the visual code, discounts a substitute or wins the same distributor. If the strategy depends entirely on a visible feature that is easy to copy, the plan needs another defensible source of value or a clear reason why speed is sufficient for the intended return.

Do not assume that more beverage brands necessarily mean less route-to-market power for incumbent suppliers, or that an acquisition price establishes an excellent or disastrous investment. Product choice, distribution control, customer demand and investment return are different questions. Test the relevant competitive thesis using matched evidence instead of treating a broad industry narrative as the business plan.

A weekly review that ends in action

Review the strategic assumption most at risk, the commercial account, the latest test or operating observation and the next commitment. Decide to proceed, repair, change the hypothesis, learn more or stop. Give each decision an owner and a trigger for reconsideration. Avoid a standing requirement that every metric improve or every uncertainty disappear.

A result that contradicts the preferred explanation is productive. It saves the business from repeating an expensive story. The point is not to defend the framework; it is to sell and serve a beverage more effectively while making the next commitment more intelligently.

Turn community feedback into an owned next action

Fictional weekly log CB-FEEDBACK-01 illustrates handling, not real customer findings. Three of twelve comments in one monitored thread question sweetness: marketing records the exact permitted anonymous wording and asks product research to check expectation versus experience. Two comments report a missing listed café: the route lead verifies availability and repairs a confirmed locator error. One comment alleges a reaction to a lot: support promptly escalates through the qualified incident procedure; it is not treated as another preference vote. The incidence in this thread does not estimate market prevalence or an adverse-event rate.

Log each comment type, observed-channel denominator, permitted product/lot context, responsible role, next check, response and decision date. For this teaching week, the route lead checks the locator next business day; product research proposes a bounded expectation check at the weekly review; the qualified incident lead handles the complaint under the actual approved procedure. Do not delay that escalation to finish a marketing report. No public response, technical conclusion or product change is authorized by this example.

A staged launch plan

First establish a viable promise. Define the choice, compare the product experience and resolve technical and claims requirements. The decision is whether the proposed offer merits the next bounded commitment, not whether it has acquired enough followers.

Next make a real transaction possible. Verify product and pack readiness, terms, inventory, service and the purchase route. Calculate contribution and cash exposure on conservative demand assumptions. A channel listing is a useful step, but it is not completed consumer access or proven demand.

Then recruit a learnable group through the method suited to the occasion: direct selling, partners, paid media, creators or sampling. Record paid behavior separately from free participation. Keep enough of the offer stable to learn, unless the explicit decision concerns a bundle of changes.

Finally observe the relevant next choice and decide. The window should reflect serving size, consumption cadence, season and commercial deadline. Expand, repair, learn more or stop using the agreed evidence. These stages organize a project; they do not describe a compulsory sequence of consumer psychology.

A concise launch workbook

Choice sheet: record product and variant, shopper and drinker, occasion and trigger, current substitute, advantage, real pack and price, serving conditions and route. Attach dated evidence of actual behavior and name the assumption most likely to be wrong.

Commercial sheet: record net realized revenue, product and packaging, freight and fulfillment, trade deductions, fees, returns, waste, acquisition and retention costs, minimums, lead times, selling window and cash dates. Identify which costs are already deducted from revenue so they are not subtracted twice.

Test sheet: record decision owner and deadline, hypothesis and rival, intervention and comparator, population, product, geography, primary outcome and denominator, observation window, full cost, guardrails and continue/change/stop criteria. State what a high intermediate score would still fail to prove.

Creative sheet: record the asset's job, buyer familiarity, reason to notice, reason to care, brand and variant cues, honest proof, actual purchase route, production cost, distribution cost, rights and claim/disclosure review owner.

Cohort sheet: record first paid event, buyer unit, cohort dates, SKU or brand boundary, mature window, observed and invisible channels, availability and promotion context, later orders and relevant contribution. Include nonreturners and explain exclusions.

Supplier sheet: record specification version, roles, quote basis, minimum unit, yield, component commitments, lead time, payment terms, release responsibility, storage, traceability and response to deviations. Do not infer these terms from an informal per-can estimate.

Use only the sheets that change the next decision. Combine them into a one-page record for a small reversible action; keep supporting schedules for consequential production or route commitments.

18 Test the framework and revise it when it loses

A useful model must change a choice before the outcome

A framework can describe every success afterward and predict nothing. To prevent that, record the diagnosis, rival explanation, intervention, forecast range and decision rule before seeing results. Measure the proposed mechanism separately from the business outcome. Do not call the model confirmed merely because every observation can be assigned to a heading.

The following tests are proposed research designs, not completed experiments. They are options to adapt when their learning value justifies the cost. Direct observation or a small operational repair may be enough for a lower-stakes decision.

Eight tests with outcomes that can disappoint us

Test 1 concerns occasion clarity. Among unfamiliar adult buyers in one stocked catchment, compare an explicit occasion proposition with equally delivered identity-led creative at the same price. Primary outcome: incremental paid purchases per assigned eligible household, with contribution through a declared replenishment window. If clarity improves but adequately precise purchase results do not, the proposed occasion-message priority fails in that setting.

Test 2 concerns recognition. Compare alternative packaging systems in a shelf-choice task, separating existing from unfamiliar buyers. Measure correct SKU selection, wrong-variant errors, time to find and masterbrand recognition. A system that improves masterbrand recall while materially increasing wrong-variant selection fails the shopping objective. Test real packaging constraints before wider rollout.

Test 3 concerns proof and fame. Where feasible, vary a specific product demonstration and recognizable endorser independently, holding offer and access stable. Measure paid acquisition and subsequent contribution per assigned person. A fame-led treatment producing greater profitable demand without the added proof refutes a blanket claim that product proof always deserves priority. Conditional repeat among purchasers is secondary because treatments may recruit different buyers.

Test 4 concerns local access. Compare store-linked creative with equally funded broader creative in comparable markets. Specify both an immediate window and a longer distribution-development window. Include stockouts and trade costs in total assigned-market results. Broad awareness that creates later profitable listings can outperform a local-first policy. Do not control away post-campaign distribution changes when estimating the campaign's total effect.

Test 5 concerns the sampling bridge. Compare sampling alone with sampling plus accurate local purchase information or an easy return route, holding monetary offers and staff contact as similar as possible. Measure incremental paid contribution, not locator clicks. Add a no-event comparator if the question is whether the event itself pays. A precise absence of commercial lift despite more clicks rejects that bridge, not all access interventions.

Test 6 concerns novelty. Compare a limited release with a normal launch or no drop using a feasible phased or randomized design. Predeclare whether the objective is portfolio contribution, priced learning or another explicit return. Include development, licensing, production minimums, waste and cannibalization. Low novelty-SKU repeat with positive profitable core-line spillover can pass; viral conversation without the declared return can fail.

Test 7 concerns creator operating models. Compare a lightly instrumented seeding process with a more managed affiliate process at similar all-in resources and recruitment standards. Measure incremental contribution, staff time, launch delay and risk incidents. This is a bundled program comparison: incentives, compensation and control may differ along with measurement. To isolate added attribution, hold those other features constant and vary the extra measurement and decisions it enables. Attribution fails its own operating case when those gains are smaller than its cost; either whole program fails when its incremental value cannot justify its resources.

Test 8 concerns the framework itself. Before future outcomes are known, compare its diagnostic choices with a simpler baseline using availability, price, prior sales and satisfaction. Hold out whole markets or brands where possible. Assess incremental results under fixed budgets after measurement cost. If the elaborate framework does not improve decisions, simplify it. Vocabulary is not a reason to retain unnecessary complexity.

Interpret an inconclusive result honestly

An imprecise estimate around zero is not proof of no effect. Before running a test, define the smallest worthwhile effect, consider sample size and variability, and decide whether the test can answer the question at an acceptable cost. Record changes in treatment delivery, contamination, concurrent promotions and missing data. Report a range of plausible effects rather than manufacturing certainty from a small sample.

Where randomization is impractical, compare like with like, examine pre-period trends and identify plausible confounders. A before-and-after case can support a decision under uncertainty, but describe its strength honestly. Never substitute a long list of metrics for a credible causal design.

Evaluate the cost of the framework itself

Compare decisions made with this method against a simpler review using the same information, budget and available actions. Before the outcomes are known, record recommended interventions, forecast ranges, information requested and staff time. Assess the economic result, quality of error detection and burden of decision-making over the relevant horizon.

Keep brands or markets held out where feasible. Repeated examples from one company are not independent confirmation that the method transfers. A forecast need not be perfect to be useful, but a more elaborate method should improve decisions enough to justify its extra cost. If it does not, simplify it.

Closing lesson

Choose a real opportunity, explain the choice without assuming a universal journey, and make the offer feasible for the drinker, the trade and the business. Compare interventions at the margin, use evidence that can change the commitment and keep the cash and inventory arithmetic honest. A useful result may be a launch, a repair, a smaller run or a decision to stop. The goal is to sell and serve beverages more effectively, not to protect a favorite framework.

Appendix A Optional diagnostic questions

Occasion: What real situation is being served, which substitute is being displaced and how often can that situation recur? Do not invent an occasion after observing a purchase and call it explanatory proof.

Recognition: Can people notice, retrieve and choose the correct brand and variant? Do not assume strong recognition is necessary for every unfamiliar or default purchase.

Expected value: What do people predict about taste, function, social value and total sacrifice? Do not treat perceived credibility as proof of actual efficacy or actual willingness to pay.

Access: What puts the correct offer into a feasible choice set? Do not confuse listing, stock, shipment, retail movement and consumption, or ignore the trade buyer's decision.

Experience and return: What happens across a full serving and the next relevant opportunity? Do not make repeat percentage the universal objective or treat satisfaction as the only possible cause.

Commercial delivery: Can the business and partners supply the offer and retain value under the relevant constraints? Do not use eventual profit to retrospectively score 'good economics' and then claim the score predicted profit.

These optional questions help discover a problem. They do not rank investments. The ranking comes from a strategic choice, feasible interventions, credible response estimates, costs, risks and the value of learning.

Appendix B Measurement and commercial terms

An occasion is a use situation. A category is a way of grouping alternatives or managing an aisle. A shopper buys, a drinker consumes, and a trade buyer decides whether and how to carry the offer. One person can fill several roles; the incentives can still differ.

Penetration is the fraction of the defined population buying in a period. Frequency is purchases per defined buyer in a period. Cohort repeat is a subsequent purchase within a specified starting cohort and window. These can move in different directions.

Sell-in is movement into the trade network. Sell-through is onward sale to the customer under the source's stated definition. Depletions generally describe another trade movement measure and must not be relabeled household consumption. Each requires its own boundary and period.

Gross profit is net revenue less costs classified as cost of goods sold; gross margin percentage is gross profit divided by net revenue. Contribution subtracts the variable or decision-relevant costs specified in the model and is not a standardized substitute for operating profit. Operating profit also reflects operating expenses under the reporting basis. Cash movement records receipts and payments, which can occur in different periods from revenue and expense. Chapter 7 supplies one consistent numerical account. [E08]

Attribution assigns observed credit under a rule. Incrementality asks what changed because of the intervention compared with what otherwise would have happened. A code used on an order can support administration without proving that the code created the order.

A constraint can make an option infeasible. A weak metric can merely describe current performance. The lowest metric is not automatically the most responsive or valuable place to invest.

Weighted distribution reflects coverage in proportion to a defined sales opportunity. Outlet count counts locations. Distribution points may count brand or SKU placements under the source's convention. Neither should be silently relabeled as the other.

A benchmark describes a reference distribution or standard under stated conditions. A target states what the business needs. A forecast predicts what will happen. A break-even threshold calculates the outcome required under assumptions. These are different objects, even when each is expressed as a number.

A proxy is an available measure used in place of a harder outcome. A diagnostic helps explain an observed result. A guardrail marks an unacceptable downside. None becomes the primary objective because it is easy to measure.

Cannibalization is displacement of existing business by another offer within the chosen boundary. Pantry loading shifts purchase timing and inventory. Neither can be inferred perfectly from launch-week volume alone.

An experiment estimates an intervention's effect under its design and implementation. A causal mechanism explains how that effect might arise. A measured association with a proposed mediator does not automatically establish that mechanism.

Appendix C Sources and evidence boundaries

References were reviewed for this edition as of 6 October 2026. Each code links to the source used for the attributed statement. Reported company figures are not causal marketing estimates. Comparative prior periods do not extend the current case window. Study publication dates and observation dates are recorded separately. No post-cutoff result is assumed.

R01 Starbucks transactions and operating income

Starbucks, Q2 Fiscal Year 2026 Results, published 28 April 2026. Quarter ended 29 March 2026. North America comparable company-operated store transactions and segment operating income. Primary company report; simultaneous investments and costs prevent campaign or service-effect attribution.

https://investor.starbucks.com/news/financial-releases/news-details/2026/Starbucks-Reports-Q2-Fiscal-Year-2026-Results/default.aspx

R02 Zevia penetration and frequency

Zevia, 1Q 2026 Financial Results presentation, 6 May 2026, slide 7. Numerator comparison for the twelve months ended 28 February 2026; prior period metrics restated. Geography, channel coverage, sample size and cohort rules are not fully specified on the public slide. A buyer-panel summary, not fixed-cohort retention.

https://s25.q4cdn.com/748406335/files/doc_financials/2026/q1/Zevia-1Q26-Earnings-Call-Presentation-5-6-26.pdf

R03 Oatly volume and segment earnings

Oatly, First Quarter 2025 Financial Results, published 30 April 2025. Quarter ended 31 March 2025, North America. Includes foodservice change and some frozen-product discontinuations, so it is not a beverage-only consumer measure. Adjusted EBITDA is a segment measure with exclusions.

https://investors.oatly.com/news-releases/news-release-details/oatly-reports-first-quarter-2025-financial-results/

R04 Celsius company brand and retail boundaries

Celsius Holdings, Second Quarter 2026 Financial Results, published 6 August 2026. Financial quarter ended 30 June 2026; retail Circana US MULO+ with Convenience RTD energy, 13 weeks ended 28 June 2026. The reported 16% dollars-per-distribution-point improvement is versus Q1, not year over year or fixed-store growth.

https://ir.celsiusholdingsinc.com/news/news-details/2026/Celsius-Holdings-Reports-Second-Quarter-2026-Financial-Results/default.aspx

R05 Label formats and simulated basket choice

Grummon and colleagues, JAMA Network Open, published 17 October 2025. Fieldwork 31 October to 21 November 2024, 5,636 US primary grocery shoppers. Randomized simulated food-and-beverage grocery task; product and price inputs drawn from September 2023. Not ordinary retail purchasing or repeat.

https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2840300

R06 Label formats and healthier option identification

Huang and colleagues, American Journal of Preventive Medicine, first online 13 December 2025, May 2026 issue. DOI 10.1016/j.amepre.2025.108222. Analysis of the same 5,636-person trial as R05; not an independent replication. Outcome is identification of a healthier option, not sales.

https://pubmed.ncbi.nlm.nih.gov/41397534/

R07 Tea packaging expectations and imagined choice

Yu and colleagues, Foods 14(22), 3893, published 14 November 2025. Data collection reported as December 2024, January 2025 and April 2025. Chinese online experiments; perceptions and imagined choice or intention. Study 3 uses two real brands, not a manipulation isolating familiarity alone.

https://www.mdpi.com/2304-8158/14/22/3893

R08 Coca Cola revenue volume and variant growth

The Coca-Cola Company, Fourth Quarter and Full Year 2025 Results, published 10 February 2026. Global company/system metrics for the year ended 31 December 2025. Prior-year comparisons are baselines. Revenue, price/mix, concentrate sales and unit-case volume have different definitions.

https://investors.coca-colacompany.com/news-events/press-releases/detail/1151/coca-cola-reports-fourth-quarter-and-full-year-2025-results

R09 Keurig Dr Pepper acquisition and segment bridges

Keurig Dr Pepper, Q4 and Full Year 2025 Results and 2026 Outlook, published 24 February 2026. Year ended 31 December 2025. US Refreshment Beverages and US Coffee segments. Coffee includes brewers and consumables; volume/mix is not pure beverage volume.

https://investors.keurigdrpepper.com/2026-02-24-Keurig-Dr-Pepper-Reports-Q4-and-Full-Year-2025-Results-and-Provides-2026-Outlook

R10 Vita Coco growth and gross margin

The Vita Coco Company, Fourth Quarter and Full Year 2025 Financial Results, published 18 February 2026. Year ended 31 December 2025. Company sales/gross profit and branded coconut-water volumes have distinct boundaries. Company movements do not reveal consumer retention.

https://investors.thevitacococompany.com/news-releases/news-release-details/vita-coco-company-reports-fourth-quarter-and-full-year-2025

R11 Zevia company results and endorsement costs

Zevia, Second Quarter 2026 Results, published 5 August 2026. Quarter ended 30 June 2026. Company net sales, volume, adjusted EBITDA and GAAP net income are separate measures. The report identifies endorsement-associated compensation, not the endorsement causal effect.

https://investors.zevia.com/news/news-details/2026/Zevia-Announces-Second-Quarter-2026-Results/default.aspx

R13 Simply Pop announced proposition

The Coca-Cola Company, Simply Brings a Juicy Pop to Booming Prebiotic Soda Category, published 18 February 2025. Announced February 2025 phased US launch. Primary statement of offer and managerial rationale; no demonstrated effect on sales, repeat or health.

https://www.coca-colacompany.com/media-center/simply-brings-a-juicy-pop-to-booming-prebiotic-soda-category

R14 Liquid IV and retailer assortment

Unilever, Pre-close aide-memoire Q1 2026, published 9 April 2026, page 2. Retrospective company commentary on Q4 2025. The key club retailer is not named; no isolated brand-effect estimate is disclosed.

https://www.unilever.com/files/unilever-pre-close-aide-memoire-q1-2026.pdf

R15 Poppi contingent consideration and acquisition accounting

PepsiCo, Form 10-Q filed 9 July 2026 for the quarter ended 13 June 2026, contingent consideration and acquisition notes. Accounting fair values and purchase-price measures require their stated definitions.

https://www.sec.gov/Archives/edgar/data/77476/000007747626000035/pep-20260613.htm

R16 NIQ reported Gen Z purchasing

Chris Costagli, NIQ, Today's Trends Unlock Tomorrow, February 2026, publicly hosted by ECRM/MarketGate, page 34. NIQ Expanded Omnishopper Panel On Demand, total US, total outlets, 52 weeks ended 29 November 2025. Dollars per buyer, not volume, penetration or individual consumption. The non-alcoholic chart has a different scope from the separate age-21-plus alcohol chart.

https://ecrm.marketgate.com/MarketingFiles/Media/20260316/NIQ_ECRM_Chris_Costagli_Presentation.pdf#page=34

R17 Keurig Dr Pepper licensed route arrangements

Keurig Dr Pepper, 2025 Form 10-K filed 24 February 2026. Bottlers and Distributors, Licensing Arrangements and third-party risk disclosures. Current reported arrangements distinguish brand rights, bottler affiliation and distribution. They do not estimate industry-wide truck shares or concentration trends.

https://www.sec.gov/Archives/edgar/data/1418135/000141813526000016/kdp-20251231.htm

R18 Mountain Dew variant and PepsiCo segment results

PepsiCo, Q2 2026 Prepared Management Remarks, 9 July 2026, pages 5 and 7. Reports Q2 Mountain Dew Zero Sugar dollar/volume share gains and PBNA organic volume decline. Variant and segment findings do not isolate original Mountain Dew.

https://investors.pepsico.com/docs/pepsico-5v9wci20/media/Files/investors/q2-2026-prepared-management-remarks.pdf

R19 Poppi transaction announcement

PepsiCo, PepsiCo completes acquisition of poppi, published 19 May 2025. Completion announcement gives the $1.95 billion price, $300 million anticipated cash tax benefits and a separate performance-based earnout. These are not the finalized acquisition-accounting consideration in R15.

https://www.pepsico.com/newsroom/press-releases/2025/pepsico-completes-acquisition-of-poppi-accelerating-strategic-portfolio-transformation

R20 Alani Nu founder context

Celsius Holdings, Celsius Holdings to Acquire Alani Nu, published 20 February 2025. Primary company announcement identifies entrepreneur and influencer Katy Hearn as founder and lists the acquisition counterparties. Used here for founder context, not for a causal estimate of influencer marketing, repeat purchase or acquisition value.

https://ir.celsiusholdingsinc.com/news/news-details/2025/Celsius-Holdings-to-Acquire-Alani-Nu-Creating-a-Leading-Better-For-You-Functional-Lifestyle-Platform/default.aspx

E03 FDA claims reference

US Food and Drug Administration, Label Claims for Conventional Foods and Dietary Supplements, live guidance checked 6 October 2026. General US reference, not a new commercial study or a complete jurisdiction-specific review. It distinguishes health, nutrient-content and structure/function claims.

https://www.fda.gov/food/nutrition-food-labeling-and-critical-foods/label-claims-conventional-foods-and-dietary-supplements

E04 FTC endorsement disclosure reference

US Federal Trade Commission, Disclosures 101 for Social Media Influencers, live guidance checked 6 October 2026. General US reference on material connections, including free or discounted products; not a substitute for specific legal review.

https://www.ftc.gov/business-guidance/resources/disclosures-101-social-media-influencers

E05 FTC health advertising substantiation

US Federal Trade Commission, Health Products Compliance Guidance, December 2022 guidance checked 6 October 2026. Explains whole-message interpretation, substantiation and the FDA/FTC relationship. General US guidance, not product-specific clearance or a legal safe harbor.

https://www.ftc.gov/business-guidance/resources/health-products-compliance-guidance

E06 FTC business data handling

US Federal Trade Commission, Protecting Personal Information A Guide for Business, live guidance checked 6 October 2026. Supports minimizing retained information, limiting access and secure disposal. It does not resolve every research, privacy or international data-transfer requirement.

https://www.ftc.gov/business-guidance/resources/protecting-personal-information-guide-business

E07 NIST comparison of means

NIST/SEMATECH e-Handbook of Statistical Methods, Two-Sample t-Test for Equal Means, section 1.3.5.3, and Confidence Limits for the Mean, section 1.3.5.2, checked 6 October 2026. Together these support analyzing paired differences with a t interval under the stated assumptions. The book's invented paired-café interval uses the one-sample t method on eight independent pair differences; it does not treat cans as independent experimental units.

https://www.itl.nist.gov/div898/handbook/eda/section3/eda353.htm

https://www.itl.nist.gov/div898/handbook/eda/section3/eda352.htm

E08 SEC financial statement definitions

US Securities and Exchange Commission, Beginners Guide to Financial Statements, 4 February 2007, checked 6 October 2026. General income/cash-flow statement explanation, not a beverage cost-classification rule. Chapter 7 explicitly states its hypothetical classifications.

https://www.sec.gov/investor/pubs/begfinstmtguide.htm

For R07, the authors' methods-date clarification is available at https://www.mdpi.com/2304-8158/14/22/3893/review_report . It supplements the published study rather than constituting another experiment.

The worked financial, production, acquisition and sampling examples are fictional original teaching material. Their assumptions are stated at the point of use. Proposed tests in the book have not been conducted by this edition.

Appendix D Practice and repair six decisions

A usable decision record names the decision and comparator; separates evidence from assumptions; reconciles units, costs and cash; proposes a check that could change the choice; and identifies readiness, owner, limit and revisit trigger. It fails if it invents evidence, mixes measures or approves an infeasible commitment. This rubric grades reasoning and completeness, not a brand's attractiveness by averaging diagnostic scores.

1 Occasion and substitute

Inadequate: 'Our audience is health-conscious young people.' This does not identify a choice. Repaired: 'Test an adult weekday sandwich occasion, replacing the currently chosen soda with a cold lightly sweet citrus single at the actual café price. Verify the last purchase and the reason to switch.' The occasion, alternative, route and unproven advantage are visible. Do not turn a few interviews into prevalence.

2 Growth boundary

Inadequate: 'Forty-eight cans shipped, so forty-eight customers tried it.' Repaired using chapter 6: 'Three stores received 48 cans, consumers bought 50 cans, and closing stock is 20 after 10 samples and 4 losses. Unique buyers and repeat are unknown.' Opening inventory makes sales above receipts possible. Invoices, consumer units and people are different measures.

3 Commercial waterfall

Inadequate: '$3 shelf price minus $0.78 product cost leaves $2.22 for the brand.' Repaired using chapter 7: 'Brand starting revenue is $1.55. Subtract $0.78 product/pack, $0.15 freight, $0.12 trade accruals and $0.05 damage/returns to get $0.45; subtract $0.20 variable marketing to get $0.25 before fixed overhead.' Costs and entity match. Neither result establishes current cash.

4 Supplier comparison

Inadequate: 'B is cheaper at $0.65 versus $0.90.' Repaired using chapter 8: 'B's $390 filling plus $300 components, $80 freight and $100 setup commits $870 versus A's $600. B leaves 600 unused components whose ownership/storage require agreement. Both need real dated quotes and technical review.' No unsupported residual value or supplier commitment is inserted.

5 Experiment and uncertainty

Inadequate: 'The revised description sold 22 more cans, so roll it out.' Repaired using chapter 9: 'Eight independent paired differences average 2.75; the approximate 95% interval is -1.24 to 6.74. The worthwhile threshold is 4. Keep assigned stockouts, retain the current rollout, fix the documented link defect and assess whether another bounded test can improve the decision enough to pay.' Positive, inconclusive and profitable are not synonyms.

6 Next commitment

Inadequate: 'The 600-can case is within a $650 loss cap, so go.' Repaired using chapter 7: 'WAIT: the -$639 forecast is within a hypothetical $650 cap, but the zero-sales/no-recovery stress loss is $1,110, exceeding it by $460; forecast cash also falls $255 below zero before receipts. Current spending authorization is $0; quote, release, route and funding evidence are absent. Founder/pilot lead reviews on Day 11 only after evidence and explicit approval.' A loss ceiling is a boundary to approve, not proof approval exists.

Use the rubric on your own record

Cover the repaired answers and fix one planted error yourself: omit unsold production, remove stockout stores, substitute a creator code for incrementality, or treat a draft permission as granted. Then check whether the repair changes the actual commitment. A good answer may be WAIT, a small repair or no further research. This edition makes no claim that a real cold-reader study or field test of these exercises has been completed.

Can Imagine free-reading note

Free public edition

This public reading edition is free to read and share by linking to the Can Imagine website. It contains the book in page-designed and accessible reading formats.

Copyright and reuse

Copyright 2026 Can Imagine. You may quote short passages with clear attribution to Can Imagine and a link to the public edition. Do not sell, repackage or distribute the full book as your own product.

Company names and trademarks mentioned in reported examples belong to their respective owners. Their inclusion is for analysis and education, not endorsement.

Privacy and responsible use

Do not put confidential client information, customer records or third-party source material into a public tool or shared workspace without the necessary permission. The examples in this book are educational records, not permission to collect or disclose identifiable data.

Educational scope

The materials support learning and business decisions. They do not guarantee sales, profit or other business results. Worked examples use hypothetical assumptions. Models are visualization references, not manufacturing approval or production-ready CAD. Use qualified specialists for the actual product's technical and legal requirements.

Editorial imagery credit: commissioned illustrative scenes of generic non-alcoholic beverages, ingredients, packaging, retail and production. These images do not depict Can Imagine clients, products or facilities.