79 tools, sorted by the question you are stuck on. Every one carries two columns that framework lists leave out. How much proof sits under it. And the exact moment it stops working.
tools
The set of situations, needs and moments in which a buyer thinks of a category: the cues a brand must be linked to in memory to be considered at all.
Breaks when Useless when the category does not yet exist. A genuinely new product has no moment to be recalled in, and it must build one before this tool has anything to work on.
The empirical regularity that brands with smaller market share have both fewer buyers and slightly lower loyalty among the buyers they have.
Breaks when Weakens where switching is contractually locked (telco, insurance) and in heavy B2B where the buyer universe is structurally small.
A stochastic model that predicts, from category structure alone, every brand's expected penetration, purchase frequency and buyer overlap.
Breaks when Assumptions fail with few players, high involvement, or one-off purchases. It also tells you nothing about why a deviation exists.
A method that defines demand by the progress a customer is trying to make, rather than by product category or customer profile.
Breaks when In low-involvement, habitual purchases it degenerates into inventing a job that was never there. Not every purchase has a purpose behind it.
A two-factor account of brand growth: the probability of being noticed or recalled in a buying situation, and the ease of buying once you are.
Breaks when Where distribution is fixed and single-channel, the second term carries no information and everything collapses into the first.
An account of how an innovation spreads through a population over time, and of the innovation attributes that set the rate.
Breaks when Strong at explaining backwards, weak at predicting forwards. You only know who the early adopters were afterwards.
The claim that a discontinuity in buying logic separates early adopters from the early majority, crossed by dominating one narrow beachhead first.
Breaks when No empirical support for the chasm itself. Outside B2B technology the mechanism has little to stand on. The focus advice is good; the theory under it is not.
A three-layer market sizing convention: the total market, the portion a given business model can serve, and the share it could realistically win.
Breaks when Produces no decision. The number is almost always reverse-engineered from the desired conclusion, and it never identifies which assumption is the fragile one.
Dividing a market into groups expected to behave differently, then choosing which of those groups to serve.
Breaks when It breaks as a description of who buys. Competing brands within a category are repeatedly measured as having near-identical buyer profiles: the segments differ in size, not in kind. A segmentation that assigns each brand its own distinct demographic or attitudinal buyer is describing a market that measurement does not find. It survives as a targeting and media convenience, not as a theory of demand.
A composite fictional buyer with a name, age, role and frustrations, built so a team has a concrete person to design and write for.
Breaks when It breaks the moment the fiction is treated as evidence. Personas are typically built from a handful of interviews and then used to exclude: this campaign is not for her, that channel is not where he is. Because competing brands largely share the same buyers, a persona used to exclude is excluding people who already buy from you. The failure is silent: nobody audits the customers a persona quietly wrote off.
The proposition that at any given moment roughly 95% of business buyers are not in the market, so most advertising has to work by being remembered later rather than by converting now.
Breaks when The 95/5 split is a derived estimate, not a measurement. It comes from average purchase-cycle length: divide the share of buyers who transact in a year by the length of the cycle and you get the in-market proportion. In categories with short cycles, high growth, or frequent re-purchase, the in-market share is materially higher and the rule under-invests in capture. It is a reminder of proportion, not a constant.
The heuristic that the greatest advantage in any business is an audience already desperate to buy, so demand should be chosen before the product.
Breaks when Visible demand is the most contested resource there is. A crowd everyone can see is a crowd everyone is already selling to, so the heuristic routes founders straight at the categories with the highest acquisition costs and the most entrenched incumbents. It also carries no test. Nothing in it distinguishes a starving crowd from a well-fed one until the money has been spent finding out.
The four decision areas a marketing plan is expected to cover: product, price, place, promotion.
Breaks when It breaks the moment it is used as a strategy tool. It is a list of levers with no theory of which lever matters, in what order, or by how much: four boxes of equal visual weight for decisions of wildly unequal consequence. It is also entirely firm-side: all four Ps are things the company does, and the customer appears nowhere. 7 Ps, 4 Cs, SAVE: the long parade of proposed extensions is a symptom of a frame that organises without prioritising.
A survey rule of thumb: ask users how they would feel if they could no longer use the product, and treat 40% answering “very disappointed” as the threshold for fit.
Breaks when The 40% threshold has no published derivation. It is a remembered benchmark from a small set of startups, not an estimate with an interval around it. Worse, the survey only reaches people still using the product, so it measures the enthusiasm of survivors and is structurally silent on everyone who already left. A product with a small, devoted, unscalable audience passes it comfortably; that is the exact failure mode it is meant to catch.
The decision about how many brands a company runs and how they relate: one master brand, a house of separate brands, or a structure between the two.
Breaks when It gives no rule for where the line falls. The frame will describe any portfolio you already have and endorse almost any addition to it, because the case for a master brand (efficiency) and the case for a separate brand (focus) are both always available. The evidence on extensions is conditional. Similarity between the extension and the parent matters, and so does the strength of the parent. A frame that does not carry those conditions is a vocabulary, not a decision rule.
A 2×2 that sorts products into stars, cash cows, question marks and dogs by market growth and relative market share, with an investment instruction attached to each cell.
Breaks when It measurably worsens the decision it is used for. The damage comes from the cell labels rather than the axes: 'dog' carries a divestment instruction that the numbers inside the cell do not, and once a product has been placed there nobody re-reads the arithmetic. Market growth is also a poor proxy for attractiveness in any category where a shrinking market is consolidating into a profitable one.
The practice of claiming a single defensible place in the prospect's mind relative to competitors, rather than competing on product attributes.
Breaks when Fails when you do not control the category definition, or when the category does not exist yet. And unless the mental position is actually measured, it decays into an argument about taglines.
A five-step procedure that derives a market category from competitive alternatives, unique attributes, the value they produce, and the segment that cares.
Breaks when Powerful only where you can choose the category. In regulated or mature markets step five is locked, and the canvas quietly becomes a four-step exercise.
The strategy of creating and naming a new market category rather than competing inside an existing one, on the premise that whoever defines a category captures most of its value.
Breaks when Heavy survivorship bias. The often-quoted 76% of category market cap comes from the authors' own unpublished analysis with no disclosed sample, method or period. The companies that tried to create a category and failed are not in the book.
The distinction between attributes on which a brand must merely match competitors and the attribute on which it must differ.
Breaks when Let the parity list grow and the product becomes a me-too. Matching is not free: most budgets disappear into it without anyone noticing.
A spatial representation of how customers perceive competing brands, derived from survey data, used to locate unoccupied positions.
Breaks when You choose the axes, so you can manufacture whichever gap you want. Unbounded by survey data it is wishful thinking. And the empty space is usually empty because nobody wants what is there.
A plot of a category's competing factors against the level each player offers on them, used to force raise, reduce, eliminate and create decisions.
Breaks when Examples are selected after the fact and there is no evidence on the method's forward success rate. Uncontested space is frequently uncontested because there is no demand in it.
A set of asymmetric strategies for brands that are not the category leader and cannot win by matching the leader's resources.
Breaks when Backfires if you are the leader: challenger posture is the market leader's most expensive mistake. Note also that Avis “We try harder” is the pre-framework archetype Morgan draws on, not a case of his method.
An analysis of the structural pressure on profit in an industry from five directions: rivalry, new entrants, substitutes, supplier power and buyer power.
Breaks when It is the smaller half of the story, and Porter's own data says so: industry effects account for 19% of the variance in US public-company profitability, business-specific effects for 32%. The frame is also static and boundary-dependent. The industry has to be defined before the analysis starts, and the definition decides the answer. Applied to a young category, where the boundary is the thing being contested, it produces a confident reading of a structure that does not exist yet.
The argument that only products worth remarking on can be marketed, because advertising can no longer buy attention for an ordinary one.
Breaks when It is defined so that it cannot fail. A product that sold was remarkable; one that did not was a brown cow; the test only becomes available after the result. The measurement points the other way too. Buyers perceive very little differentiation between the brands they actually buy, and success does not track perceived differentiation. What does the work is distinctiveness, being instantly recognisable, which is a memory property rather than a remarkable one and is built by repetition the model has no place for.
Colours, characters, sounds and shapes: the non-name elements that trigger the brand in memory, scored on fame and uniqueness.
Breaks when A new brand has no assets to score: investment first, measurement much later. And changing an asset resets everything accumulated in it.
Six mutually incompatible theories of how advertising produces its effect, set out together because the industry uses all six while claiming only one.
Breaks when It will not tell you what to do. Useful at the start of a brief, useless at the end of one.
Six properties shared by ideas that survive being retold: simple, unexpected, concrete, credible, emotional, story.
Breaks when An evaluation tool posing as a generation tool. Chase all six at once and you achieve none.
A paired diagram mapping a customer's jobs, pains and gains onto a product's features, pain relievers and gain creators.
Breaks when Filled in around a table without talking to a customer, it becomes a register of assumptions. That is how it is usually filled in.
A seven-beat narrative template that casts the customer as hero and the brand as guide, applied to marketing copy.
Breaks when Produces the same shape in every sector: clarity bought at the cost of differentiation. Note also that no independently verified outcome case exists. Every circulating figure traces back to a certified StoryBrand guide's own marketing.
A one-page hierarchy of a single overarching message, the pillars supporting it, and the evidence beneath each pillar.
Breaks when An internal alignment device. Copied straight into public-facing copy it produces corporate boilerplate.
A character, format or structure reused across successive campaigns so that recognition accumulates instead of resetting each time.
Breaks when Nothing compounds for a brand that repositions or changes agency often. Without accumulation the device is just a mascot.
Six levers of persuasion presented as a general toolkit for changing behaviour: reciprocity, commitment, social proof, authority, liking, scarcity.
Breaks when It breaks when laboratory effects are assumed to transfer at laboratory size. Several of the underlying findings shrink sharply or fail to reproduce in field replication. It breaks harder when the principles are applied to brand advertising, where the audience is not deciding anything. Scarcity aimed at someone who was not going to buy either way is a message with no mechanism. Compliance research generalises badly to advertising, which is mostly not asking for compliance.
The model that buyers pass in sequence through attention, interest, desire and action, with the population narrowing at each step.
Breaks when It breaks as a description of behaviour, which is the thing it claims to be. No empirical work has established that buyers move through these stages in sequence; measurement finds interrupted, non-linear, re-entered paths, and finds that most category buyers are in no stage at all at any given time. Treating it as real produces the characteristic errors: chasing “interest” as if it converts downward on a schedule, and building stage-gated budgets for a population that does not queue.
The repeated finding that the advertisement itself accounts for the largest single share of the sales effect, larger than targeting, reach, or where it ran.
Breaks when The effect itself is well established. Rosengren and colleagues pool studies where judges rate creativity independently, before any outcome is known, so this is no longer a residual left over after the measurable variables are counted. What does not survive is the ranking. The claim that creative outranks reach, targeting and placement comes from one kind of study, a commercial decomposition run by measurement vendors, and the academic work never puts those four in the same model. The size of the effect also moves a great deal depending on which dimension of creativity is measured, in which medium, against which outcome. A Law here would mean a magnitude you can plan against without knowing your category. This is not that yet.
A five-step ladder of how much a prospect already knows, used to decide where a message has to start: unaware, problem aware, solution aware, product aware, most aware.
Breaks when There is no measurement attached to it. Nothing tells you which stage a visitor occupies, so the stage is assigned afterwards from what the copy did. A page that failed gets re-diagnosed as aimed at the wrong stage, and the ladder is never wrong. It also assumes one reader per page, and paid traffic delivers all five stages to the same URL at once.
A five-level scale of how many times a market has already heard a claim, used to decide whether to lead with the promise, a bigger promise, a mechanism, a better mechanism, or identification with the reader.
Breaks when The level is asserted, never measured, and it is asserted about a whole market when sophistication is a property of an individual buyer's exposure history. Someone entering a level-five category for the first time has heard none of the claims. The scale also runs one way only: it says markets get more jaded and never less, which is contradicted by the constant inflow of new category buyers that keeps the naive level populated indefinitely.
The rule that a campaign has three parts, chosen in this order: who you are talking to, what you say, and where you say it.
Breaks when The strict ordering is wrong wherever the channel constrains the message, which is most paid media: a six-second pre-roll and a long-form email cannot carry the same argument, so media is an input to message rather than a consequence of it. And the rule names the three variables without saying how to choose any of them. It is a checklist presented as a strategy.
The claim that organisations should communicate why before how and how before what, because purpose persuades and specification does not.
Breaks when Its supporting argument is a neuroanatomy claim asserted rather than sourced: the limbic system handles the why and has no capacity for language. The evidence offered is a handful of famous companies chosen after they succeeded, and it is the same company in every retelling. The model also has no procedure: it says start with why and does not say how to arrive at one, so in practice the why is reverse-engineered from a what that already worked.
The model that advertising moves a buyer up an ordered set of mental steps, each one a prerequisite for the next: awareness, knowledge, liking, preference, conviction, purchase.
Breaks when The ordering has been examined and not found. Hierarchy models neither describe the effects of advertising accurately nor conceptualise how advertising works as a market force, and forty years of reviews have failed to establish the sequence they assert. Treating the steps as prerequisites produces the standard error: money spent lifting awareness on the assumption that preference and purchase follow it on a schedule, and a tracker that reports movement at the top of the ladder as progress toward the bottom.
A two-route account of persuasion: when a person is able and motivated to think about a message they process its arguments; when they are not, they respond to the cues around it instead.
Breaks when Elaboration is not observable in advance. In the field the route gets inferred from what worked, which makes the model unfalsifiable there: a rational ad that failed is relabelled a low-elaboration situation and the model is preserved. It also presumes a decision worth thinking about, and most category purchases are low-attention repeats where neither route is engaged and availability decides the outcome.
Repeated exposure to a stimulus increases liking for it, with no argument, no reward and no recall required.
Breaks when The curve turns over. A meta-analysis of 268 curve estimates from 81 articles finds the effect is inverted-U shaped: liking rises with exposure and then falls. The schedule that built preference destroys it if it keeps running unchanged. The effect is also strongest where the exposure is not consciously recognised, which means the attention an ad wins can cost it the mechanism it was relying on.
A screening procedure that ranks nineteen acquisition channels, cheap-tests three, and concentrates effort on the one that works.
Breaks when The list froze in 2015, and the method needs test budget and traffic to run at all. Very early on you need one bet, not a survey.
A model of acquisition in which the output of each cycle becomes the input of the next, as opposed to a funnel refilled from outside.
Breaks when Not every business has a loop. Forcing one produces a north-star metric that measures the wrong thing.
A five-stage funnel used as shared vocabulary for diagnosing where users are lost: acquisition, activation, retention, revenue, referral.
Breaks when Assumes a linear one-directional flow, which flattens multi-touch B2B and pushes retention to the end where it does the least good. McClure's own funnel percentages carry his explicit “not actuals” disclaimer. Never cite them as benchmarks.
The constraint that revenue per user and purchase frequency determine which acquisition channels a business can afford at all.
Breaks when Early on you do not know real ARPU or retention, so the arithmetic runs on optimistic estimates. Balfour gives no dollar thresholds. Any you see attached to this are someone's invention.
The prescription, following from buyer-base structure, to reach as many category buyers as possible as continuously as possible rather than concentrating on heavy buyers.
Breaks when Inverts where the buyer universe is genuinely small and enumerable. A 300-account B2B list is a case where targeting really does pay.
A four-cell grid of growth options formed by crossing existing and new products with existing and new markets.
Breaks when It labels rather than directs. Knowing which box you are in says nothing about what to do inside it.
Two opposed rules for scheduling the same budget: buy enough repetition to cross a threshold, or spread continuous light coverage so an ad is present close to the purchase.
Breaks when Both break when treated as constants. The three-exposure threshold was a summary of the evidence available in 1979 and has never held as a universal number; recency assumes buying is continuous and randomly timed, which fails in seasonal and considered-purchase categories where there is a window and it is knowable. The deeper problem is that the two rules are not reconcilable. A planner citing whichever one supports the plan already chosen is using neither.
Measuring the seconds a person actually looks at an advertisement, instead of the opportunities-to-see that a media buy nominally delivers.
Breaks when It breaks on how the number is made. Attention is measured by eye-tracking on comparatively small opt-in panels, then modelled across platforms whose formats, screen sizes and viewing contexts differ enormously. So the cross-platform comparison, which is the use everyone wants, is the least supported part. Thresholds for what counts as attention are vendor-defined and not standardised, and the strongest evidence links attention to short-term sales measures rather than to long-term brand effects. Treated as currency it becomes a target, and the format that maximises measured seconds is not automatically the format that sells.
Treating the right to send someone a message as granted by them and revocable, and building an asset out of the grants rather than out of purchased attention.
Breaks when The permission is largely an artefact of the form. Consent rates move sharply with the default and the wording of the request while the offer stays identical, so a list measures the checkout design rather than anticipation, and a large permission asset can contain almost no intent. It is also capped by construction: everyone on the list already found you, which makes it a retention instrument routinely presented as a growth one.
Choosing the smallest group you could serve well enough that they would miss you if you vanished, and building for exactly them instead of for everyone.
Breaks when It contradicts the two best-evidenced entries on this site and it loses to both. Double Jeopardy says brands with fewer buyers also get lower loyalty from the buyers they have, so a small audience is penalised twice rather than compensated. Mental & Physical Availability and Reach Maximisation say growth comes from broad continuous reach into light buyers, and that the efficiency of narrow targeting is usually a growth loss. Godin's own framing is that the smallest viable audience is a stepping stone, not a compromise; held as a permanent strategy it is the mechanism by which small brands stay small.
A four-question survey technique that derives an acceptable price range from judgements of too expensive, expensive, cheap and suspiciously cheap.
Breaks when Measures stated intention, not behaviour; systematically inflates the upper bound; ignores competitor prices and context. It gives you a range, never a price.
A survey technique that asks purchase likelihood at a series of set prices and builds demand and revenue curves from the answers.
Breaks when You choose the price points, so the answer comes from inside your own range. Still stated preference, and it does not model competitor response.
A method that infers the hidden weight a buyer places on each attribute and on price by observing forced choices between whole product bundles.
Breaks when Expensive, sensitive to design error, needs real sample. Past a certain attribute count respondents fatigue and the data degrades.
The maximum a rational buyer should pay: the cost of their best alternative plus the quantified value of the difference you provide.
Breaks when Strong in B2B, meaningless for emotional or status goods. And pricing at the full calculated value leaves the customer no reason to move.
A tiered price structure with deliberate constraints, called fences, that stop customers in one tier from buying at another tier's price.
Breaks when If the difference between tiers is not perceived, everyone takes the middle one and total revenue falls. A tier without a fence is just a discount list.
An account of choice in which outcomes are judged as gains or losses against a reference point, with losses weighted more heavily than equivalent gains.
Breaks when Laboratory effect sizes shrink markedly in the field, and popular derivatives like the decoy effect replicate poorly. Trust the core asymmetry, not the ornaments.
Adding a deliberately worse third option to a choice set so that buyers move toward the option you want them to take.
Breaks when It breaks under replication. Large-sample and field attempts have repeatedly failed to reproduce the effect outside the original narrow stimulus designs, and where it does appear the size is far below what pricing-page advice implies. It also assumes buyers evaluate the set as a set, which is not how most purchases happen. The practical risk is not a null result. It is a live pricing page carrying a real option that some customers will actually buy, at a price you set to be unattractive.
The percentage change in sales produced by a 1% change in price, generalised across studies into a single magnitude.
Breaks when It is an average drawn mostly from packaged-goods research and it hides exactly the variance a single decision turns on. The estimate also moves with method rather than with the market. Correcting for price endogeneity increases measured elasticity substantially, so two studies of the same category can disagree because of the model. And an elasticity measured on promotional variation says nothing about a permanent list-price change, which is the decision most teams are actually making.
The proportion of each joining group still active, plotted against time since joining.
Breaks when You cannot read flattening before enough time has passed. A curve drawn from three months of data is drawn from hope.
A family of probability models that predict, from purchase history alone, how many purchases a customer will make next and whether they have quietly stopped buying.
Breaks when The models assume a customer's underlying purchase rate is stationary, meaning it does not change. So they break precisely when you intervene: sustained promotion, a price change, a category shift, a competitor entering. They are a forecast of what happens if nothing is done, which makes them a strong baseline and a poor evaluator of your own campaign. They also explain nothing: a customer with a 4% survival probability comes with no reason and no lever.
A four-step loop proposed as the mechanism by which a product becomes a habit: trigger, action, variable reward, investment.
Breaks when The loop's own examples break it. Several of the most habit-forming products ever built work by removing variability rather than adding it, and others run investment before any reward at all, which reverses the stated order. It also cannot separate a habit the user wants from one they resent, and it imports variable reinforcement from animal schedules without the boundary conditions that came with it. It offers no account of why a habit ends, which is the thing a retention team is actually trying to predict.
A budget split, roughly 60% to broad-reach brand building and 40% to short-term activation, associated with the strongest long-term business effect.
Breaks when The ratio moves by category. The B2B figure is about 46/54, drawn from fewer than 50 cases, and the authors themselves call it tentative and warn against following it precisely. The database skews to large, mature, well-funded brands; a product starting from zero is not in it.
The gap between a brand's share of category advertising voice and its share of market, which predicts the direction and rate of share change.
Breaks when Share-of-voice measurement gets less reliable as media fragments, and the model ignores creative quality entirely. Excess voice for a bad ad buys nothing.
The modelled carry-over of advertising effect after exposure, combined with the falling return on each additional unit of spend.
Breaks when Decay half-life and saturation point are estimated from data. With thin data they are effectively invented, and the model then confirms its own assumption.
A regression on aggregate historical data that estimates each marketing input's contribution to sales.
Breaks when Correlational. Uncalibrated by experiment it moves budget systematically to the wrong channel. And whoever builds the model can largely choose the answer.
A randomised experiment that withholds spend from matched markets or audiences in order to measure what that spend actually caused.
Breaks when Needs sample, time and patience. The statistical power to detect a small effect is absent from most budgets, so “no effect” usually means “we could not measure it”.
The percentage change in sales produced by a 1% change in advertising spend, generalised across studies into a single magnitude.
Breaks when It is an average, and it has been falling. Elasticity is higher for durables than non-durables, higher early in the life cycle than at maturity, and higher when advertising is measured in gross rating points than in money. The headline figure describes no actual brand. Used as a planning input it also silently assumes average-quality advertising, which is the variable most under a team's control and precisely the one the meta-analysis averages away.
Randomised in-market tests that deliver different advertising weight or copy to matched halves of a cable television audience and read the sales difference from the same households' purchase panel.
Breaks when It is the finding that is uncomfortable, not the method. Across 389 tests, increasing advertising budgets relative to competitors did not increase sales in general, and standard recall and persuasion copy-test measures showed no strong relationship to sales effect. The two things most advertisers optimise failed to predict the outcome. A later analysis of 241 tests run between 1989 and 2003 found the weight effect significantly above zero for established products, driven by the more recent tests, so the 1995 result is partly a finding about a period. It also only ever covered packaged goods on television.
The ratio of a customer's expected lifetime gross profit to the cost of acquiring them, with payback period as the time taken to recover that cost.
Breaks when LTV is a forecast, and extrapolating it from early cohorts is the most common error in the field. Note also that the 3:1 ratio and the twelve-month payback rule are conventions, not findings. They are routinely presented as research.
Repeated survey measurement of a brand's presence in memory, covering recall, consideration and asset attribution, reported as a time series.
Breaks when Question wording determines the answer; mix prompted and unprompted recall and the series becomes meaningless. Change methodology and the entire history is void.
Rules that assign credit for a conversion across the touchpoints that preceded it.
Breaks when Observational, not causal. Benchmarked against experiments it systematically inflates cheap and search channels. It cannot substitute for an incrementality test, and it constantly is.
A single-question metric scoring likelihood to recommend on a 0–10 scale and reporting the share of promoters minus the share of detractors.
Breaks when The growth-prediction claim has been repeatedly refuted. Keiningham et al. examined 21 firms and 15,500+ customer interviews across five industries and found no support for NPS as the best predictor of growth. Correlations were inconsistent and mostly non-significant, and in two of the three US industries Reichheld showcased, ACSI outperformed it. Useful as an internal trend; not an external truth.
The principle that an observed statistical regularity collapses once it is used as a target for control.
Breaks when It does not break. What breaks is the measurement system that forgot it.
A brand's share of category-related search volume, used as a fast and free leading indicator of market share.
Breaks when It is a proxy, and proxies drift from what they proxy. Search volume rises with newsworthiness as well as with demand, so a crisis and a successful campaign look the same. In categories where buying does not involve search, or where the brand name is also an ordinary word, the signal is mostly noise. And it is cheap to move, by brand-name bidding or PR spikes, so it degrades the moment it becomes a target. That is the standard Goodhart path.
An experiment design that sizes a test to maximise total profit across the test and the rollout, instead of sizing it to reach statistical significance.
Breaks when It optimises the profit of one decision, so it deliberately accepts a higher error rate than a scientific test would. It will sometimes roll out the worse arm, by design, because the expected cost of that is lower than the cost of the larger test. That makes it the wrong instrument when the result has to generalise, feed a model, be reused across markets, or be defended to someone outside the team. It also needs a prior on the effect size, and a badly chosen prior moves the recommended sample size a long way.
Running two versions of the same advertisement to randomly divided halves of one audience and keeping the version that produces more response.
Breaks when Most tests are underpowered for the difference they are looking for, so a large share of declared winners are noise that will not repeat. The design also only compares the variants you thought of: it optimises inside a set and never reports that the set was wrong, which is how a headline test returns a clean 12% lift on a page selling the wrong thing. And it measures response, not profit. Hopkins' keyed coupon counted replies; the modern equivalent counts clicks.
Brand equity defined as the difference in how a customer responds to marketing for a branded product versus the identical product unbranded, located in brand knowledge held in memory.
Breaks when It never specifies the weights. Nothing in the framework says how much awareness is worth against how much favourability, so every tracker built on it makes those trade-offs implicitly and reports the result as a score. Defined as a differential response, it is also close to unfalsifiable in practice: any gap between a brand and an unbranded equivalent counts as evidence for it, and the framework names no case in which the differential should be zero.
Nothing matches that. .