MARKETING ANALYTICS A PRACTITIONERS GUIDE TO MARKETING ANALYTICS AND RESEARCH METHODS

1. 1Separate product from brand before measuring anything. Write down what the product delivers (functional benefits) and what lives only in consumers' minds (memories, feelings, associations). Equity is the gap between the two, so a product-only test such as taste or performance cannot measure it.
2. 2Define equity as incremental value: what the brand earns from the thoughts and feelings it evokes, over and above what the product earns. Decide up front the outcomes you will read it from: willingness to recommend, willingness to pay a premium, and repeat behaviour.
3. 3Map consumer knowledge and attitude along the route the chapter shows. Cover awareness, interest and consideration; the benefits and relevance people see; the persona the brand exudes; and whether it is seen as popular, fashionable or prestigious. Give each layer at least one survey question per brand.
4. 4Measure attitudinal loyalty with the chapter's tools: the loyalty pyramid, the Net Promoter Index and the Net Advocacy Index. Field them for your brand and the main competitors in the same sample, so the scores can be compared.
5. 5Combine the measures into a Brand Equity Index using one of the brand equity models in the chapter. Fix the components and weights before you see the data, and compute the index the same way for every brand and every wave.
6. 6Explain the score with the drivers of equity, especially brand image. Profile the brand on functional, rational, emotional, personality and brand-consumer relationship attributes. Choose the attribute set by the study's objective: functional attributes for a new brand's relevance, emotive and personality attributes for advertising work.
7. 7Add unsolicited evidence from consumer generated media (CGM). Use it to see which categories, attributes, themes and competitors are linked to your brand, and how far consumers advocate it. Take its vocabulary into your questionnaires, probe its new ideas in qual, and validate and size them in quant. CGM alone cannot support conclusions about the whole market.
8. 8Report equity relative to competitors and over time, and tie each driver gap to a marketing action (target segment, differentiation, positioning). Equity is rare: for most brands it is relatively low, so expect modest numbers rather than Coca-Cola-like ones.

Before launching New Coke in 1985, Coca-Cola ran 190,000 taste tests, and consumers preferred the New Coke formulation. A product-only measure would have said the change was safe. What actually mattered was consumers' extraordinary affinity for Coca-Cola, the brand. American consumers responded intensely, and that outpouring of affection is the brand's source of equity. Applied with the steps above: the taste tests measured the product (functional benefit), while the equity lived in memories, feelings and associations that no taste test captured. A brand equity read taken beforehand, covering advocacy, emotional and relationship attributes and loyalty, would have registered that attachment. Interbrand ranks Coca-Cola among the world's most valuable brands, and the book calls it an exception, since most brands have relatively low equity. The lesson is to measure the brand and the product separately and never let product preference stand in for brand equity.

• If your research shows product preference but you have not measured advocacy, loyalty and emotional or relationship attributes, then do not conclude the brand is safe to change; the product score does not capture equity.
• If the study's goal is to judge a new brand's relevance, then profile it on functional attributes; if the goal is understanding personality for advertising, then use emotive, personality and brand-consumer relationship attributes.
• If CGM surfaces new themes, vocabulary or competitor associations, then take them into qual to explore and quant to size them before acting; do not treat online chatter as a market-wide conclusion.
• If equity scores are low for your brand, then treat that as the norm for most brands and focus on the drivers (brand image and positioning) rather than expecting Coca-Cola-level affection.
• If you compare equity across brands or waves, then hold the index components, weights and sample definition constant, or the movement reflects method rather than the brand.

• Equating product preference with equity. The tell: the evidence is taste, performance or blind-test results with no advocacy, loyalty or perception measures, as in the 190,000 New Coke taste tests that missed the affection for Coca-Cola.
• Measuring only awareness and ignoring the rest of the route. The tell: strong awareness numbers, but no data on relevance, persona, prestige, recommendation or premium willingness.
• Using an attribute set that does not fit the objective. The tell: functional attribute ratings used to explain a personality or advertising problem, or the reverse.
• Over-reading CGM. The tell: a conclusion about the whole market drawn from online conversations, which lack the rigour to support it, with no quant validation.
• Reading the equity score without competitors or a time baseline. The tell: a single standalone number with no competitor comparison, no earlier wave and no named driver gaps.

• ▸Brand and competitors: list the brand and the competitors measured in the same sample.
• ▸Product-only strengths: record functional benefits that a product test would capture, kept separate from equity.
• ▸Knowledge and attitude funnel: score awareness, interest, consideration, relevance, and popular/fashionable/prestigious perception.
• ▸Advocacy and loyalty: record the Net Promoter Index, Net Advocacy Index and loyalty pyramid position.
• ▸Behavioural outcomes: record willingness to recommend, pay a premium and repeat purchase.
• ▸Brand image profile: rate the brand on functional, rational, emotional, personality and relationship attributes, choosing attributes by the study objective.
• ▸Brand Equity Index: record the model used, the components and weights fixed in advance, and the resulting score by brand.
• ▸CGM findings: record associated categories, attributes, themes, competitors and consumer vocabulary, and mark which items need qual or quant follow-up.
• ▸Driver gaps and actions: list the biggest gap versus competitors or the prior wave, and the positioning or marketing action it triggers.

1. 1Define the four behaviours before you pull any data: trial (bought at least once), repeat (bought again), frequency (how often), volume (how much). Use actual purchase records, not stated intent. The source treats behavioural loyalty as 'specifically a behavioural outcome', separate from attitude.
2. 2Build the pyramid as nested subsets: awareness, then trial, then consideration, then regulars, then dependables. Each level is a subset of the one above (trial within awareness, consideration within trial, regulars within consideration, dependables within regulars). Report each level as a percentage of the one before so you can see where consumers drop out.
3. 3Compute behavioural loyalty per individual or household as brand purchases divided by total category purchases over a fixed period. This is brand share among brand buyers. State the period every time, because the number is only meaningful 'over that period of time'.
4. 4Measure attitudinal loyalty separately: a positive, preferential attitude, the brand named as favourite, willingness to recommend. Put it next to the behavioural figure so you can see the gap, rather than inferring one from the other.
5. 5Diagnose any gap between the two loyalties. Where attitude is high but behaviour is low, check availability, affordability and convenience (for example, the preferred brand is not stocked at the regular store). Where behaviour is high but attitude is weak, check switching inconvenience and inertia, as in retail banking.
6. 6Track shifts between adjoining pyramid levels across periods. Movement up or down shows the brand's interaction with consumers strengthening or weakening. 'Fatter is better', so look for which level is thinning, not only the headline total.
7. 7Compare against competing brands on the same pyramid. Differences in shape show relative strengths and weaknesses.
8. 8Take the deeper cuts from consumer-panel analysis, which the book organises as: width and depth of purchase, buyer groups, trial and repeat purchase, overlap, basket, and gain-loss. Use these to explain why the pyramid moved.

The source's own case: a consumer drinks 40 cups of Nescafe and 10 cups of other coffee brands in a period. Her category total is 50 cups, so her behavioural loyalty to Nescafe is 40/50 = 80% for that period. That is a purchase fact, and it says nothing about whether she likes the brand. The source gives two reasons not to read it as attitude. If Nescafe is her favourite but her regular store does not stock it, her buying understates her attachment. In services such as retail banking, a customer may keep buying even after affinity has eroded, because switching is inconvenient and inertia holds her in place. So you record 80% as the behavioural reading, record her attitude from a separate question, and treat any mismatch as a lead to chase (availability, price, convenience, inertia) before concluding anything about the brand. The New Coke episode is the same warning at larger scale: taste tests favoured the new formula, yet consumers' actual response was not what Coca-Cola anticipated.

• If behavioural loyalty is high but attitudinal loyalty is weak, then treat the base as exposed to inertia or switching barriers, and expect it to erode once the barrier lifts.
• If attitudinal loyalty is high but behavioural loyalty is low, then check availability, affordability and convenience before touching brand messaging.
• If a pyramid level thins relative to the one above it, then focus diagnosis on that transition (for example consideration to regulars) rather than on the whole funnel.
• If you compare behavioural loyalty across brands or periods, then confirm the same category definition and time window first, because the metric is a share of total category purchases over a stated period.
• If test or survey preference contradicts marketplace behaviour, then trust observed purchase, as the New Coke taste tests versus consumer response show.

• Treating behavioural loyalty as brand loyalty. The tell: a high repeat share is read as 'consumers love us' with no attitude data, while the brand sits in a service with high switching cost.
• Reading a low purchase share as weak affinity. The tell: you never checked whether the brand was available at the consumer's regular store or affordable for her.
• Computing share against the wrong denominator. The tell: loyalty is calculated on total spend or all products instead of total category purchases, so figures cannot be compared with the 40-of-50-cups logic.
• Reporting a single pyramid snapshot. The tell: no movement between adjoining levels and no competitor comparison, so you cannot say whether the brand is strengthening or weakening.
• Relying on stated preference as a proxy for buying. The tell: a positive concept or taste-test result is used to forecast sales without checking actual trial and repeat, the gap that surprised Coca-Cola with New Coke.

• ▸Brand and category: name the brand and define the category boundary used for every denominator.
• ▸Time period: state the start and end dates of the measurement window.
• ▸Pyramid counts: enter awareness, trial, consideration, regulars and dependables, each as a subset of the level above.
• ▸Step-down percentages: for each level, divide by the level above to find where consumers drop out.
• ▸Behavioural loyalty: brand purchases divided by total category purchases per individual or household, then averaged for brand buyers.
• ▸Attitudinal loyalty: share naming the brand as favourite and willing to recommend.
• ▸Gap flag: note whether behaviour is above or below attitude, and by how much.
• ▸Constraint check: availability, affordability, convenience, switching inconvenience, inertia; mark which apply.
• ▸Competitor comparison: the same pyramid and loyalty figures for the main rival brand(s).
• ▸Period-on-period shift: which adjoining level strengthened or weakened, and the panel analysis to run next (trial and repeat, overlap, gain-loss).

1. 1Define performance in three measurable outcomes before touching any data: revenue, market share, profitability. The passages name these as the result of effective marketing, so each metric needs a baseline and a competitor comparison.
2. 2Build the loyalty pyramid for your brand. Measure the levels in order: awareness (unaided and aided), ever use (trial), would consider, regulars, dependables. Each level is a subset of the one below, so read the drop-off between adjacent levels to find where the brand loses people.
3. 3Isolate the dependables. The source says these behaviourally loyal consumers drive brand performance. Size them, then profile them by segment (age, income) as the store equity exhibit does.
4. 4Compute behavioural loyalty separately from attitudinal loyalty. Behavioural loyalty is brand share amongst brand buyers: brand purchases divided by total category purchases for the buyer. Example: 40 Nescafe cups and 10 other cups gives 80%. Compare it with the share of attitudinally loyal consumers.
5. 5Track brand equity against competitors with an equity index (Nielsen, Voltage, Equity Engine or Equity Builder style). Answer four questions: how does equity compare with competitors, what are the strengths and weaknesses, is equity improving or deteriorating and in which segments, and what is driving the change.
6. 6Check the price gap and substitutability. The Ipsos model says rising price gaps erode behavioural loyalty and high substitutability vitiates the benefits of equity. Compare your price against rivals alongside your equity score.
7. 7Link marketing initiatives to the movement in equity and loyalty, then to sales. The book positions market mix modelling as the method for assessing the impact of marketing activities on sales from historical data.
8. 8Re-run the pyramid and equity index on a fixed cadence (the shopper trends study is annual, with the previous year shown beside the current year). Use the same sampling and weighting each wave so changes are comparable.

Take Nielsen's Retail Shopper Trends study of Singapore supermarkets (2006 edition: 1,300 face-to-face interviews with household grocery shoppers aged 15-65, quota sampled by age, gender and race, weighted to national main-grocery-shopper statistics). The store equity index runs on a 10-point scale. Only Fujimart crosses 3, and most brands hover below 1. That is the market's normal shape, so a score of 3 is strong and not a failure against 10. Next to the equity index sits a loyalty pyramid for each chain, which shows whether the chain's shoppers stop at awareness or climb to regulars and dependables. Shop n Save is the performance contrast. Dairy Farm bought it for its high-traffic locations and grew it from 35 to 46 stores by re-branding G-Value, yet its stores were less profitable than DFI's other chains. Store count and location do not equal brand performance. You diagnose it by comparing its equity, its pyramid and its drivers (location, price and value for money, variety and quality) with competitors, and by asking which segments are moving the index up or down year on year.

• If awareness is high but trial or consideration drops sharply, then fix relevance or trial drivers before spending on more awareness.
• If attitudinal loyalty exceeds behavioural loyalty, then look for availability, affordability or convenience constraints, not an attitude problem.
• If equity is stable but sales fall, then check the price gap and substitutability first, since both erode behavioural loyalty.
• If an equity index is low in absolute terms, then benchmark against competitors on the same scale before judging, because most brands score below 1 and above 3 is exceptional.
• If a change in equity appears, then break it down by segment and by driver before attributing it to a campaign.

• Equating physical footprint or distribution with performance. The tell: store or outlet count grows while profitability lags, as with Shop n Save.
• Treating loyalty as one number. The tell: stated preference is strong but share among buyers is low, because attitudinal and behavioural loyalty were conflated.
• Reading the equity index in isolation. The tell: a score is called weak without competitor comparison, though most brands score below 1.
• Ignoring price gap and substitutability. The tell: equity holds yet loyalty and share fall after a price increase relative to rivals.
• Trusting a single research result over brand meaning. The tell: taste tests favoured New Coke, but consumers rejected the change because the brand carried heritage the test did not measure.

• ▸Brand and competitor set: list your brand and 2-4 direct competitors measured.
• ▸Outcome baselines: current revenue, market share and profit margin, with the prior-period value.
• ▸Loyalty pyramid percentages: awareness (unaided/aided), ever use, would consider, regulars, dependables.
• ▸Largest pyramid drop-off: name the two adjacent levels with the steepest fall and the likely cause.
• ▸Behavioural loyalty: brand purchases divided by total category purchases among buyers, as a percentage.
• ▸Attitudinal loyalty share: percentage who hold a preferential attitude, and the gap to behavioural loyalty.
• ▸Equity index vs competitors: your score, rival scores, and last year's score on the same scale.
• ▸Price gap and substitutability: your price relative to rivals, and how easily buyers can switch.
• ▸Segment and driver movement: which age or income segments and which drivers explain the change, and the action decided.

1. 1Map your decisions to the book's six domains before choosing any method: brand (image, equity), consumer (segmentation, qual, quant, satisfaction, panels, big data), product (NPD, design, validation), advertising (how it works, new media, copy testing), price and promotion (promotion evaluation, market mix modelling), and retail (tracking, analytics, distribution, category management). Mark each domain 'uses a method' or 'runs on judgement'. The blanks are your application gap.
2. 2Write the decision each method must inform before you commission it. The passages show the chain: brand equity research feeds brand and marketing strategy, and market mix modelling assesses the impact of marketing activities on sales. If you cannot name the decision a study feeds, do not run it.
3. 3Pair qualitative and quantitative methods deliberately. Use qual for exploratory, explanatory/diagnostic and evaluative work, and in product, creative and packaging development. Use quant, with defined problem, research design, questionnaire, sampling and analysis, to size and confirm what qual surfaces. Run qual first to form hypotheses and quant second to measure them.
4. 4Measure both attitudes and behaviour. Attitudinal tools include the loyalty pyramid, Net Promoter and Net Advocacy indices, brand equity models and customer satisfaction surveys. Behavioural tools include consumer panels (width and depth of purchase, trial and repeat, overlap, gain-loss) and retail tracking. Where the two disagree, investigate before acting.
5. 5Match the method to the stage of the product lifecycle. Use sensory research, conjoint analysis and the Kano model in design. Use simulated test markets, controlled store tests, the Parfitt-Collins model and the Bass diffusion model in validation. Use copy testing and ad tracking once advertising runs. Use market mix modelling on historical data to separate baseline from incremental volume.
6. 6Add the new data sources where they sharpen a decision: online qual (discussion boards, video-based) for geographically dispersed respondents, and big data, social media sourcing and machine learning for consumer analytics. Treat these as additions to the toolkit, not replacements. Keep experienced moderators on online qual, or boards degenerate into 'glorified open-ended Q&A surveys'.
7. 7Close the loop with interpretation and recommendation. Each method in the book ends with a step that converts findings into action. Require every study to end with a written recommendation, an owner and a date. After the action, check the result with tracking or panel data.
8. 8Audit coverage annually. Re-score each domain: methods in use, decisions informed, results checked. Fund the gaps in order of decision value.

In 2005 Creative Technology spent $100 million on an advertising campaign for its ZEN media player that stressed functional superiority with a feature-by-feature comparison against the iPod. The passage notes that Creative 'was marketing a B2C product in a B2B style': ordinary consumers, unlike business buyers, usually do not weigh features and price this way. A brand equity and image study would have tested this before the spend. Brand image tracking and perceptual mapping would show where ZEN and iPod sit in teenagers' and young adults' minds. A loyalty pyramid and Net Promoter reading would show how deep the iPod's affinity ran. Qualitative work with the target segment would show whether features mattered at all, and copy testing would check how the message landed. Coca-Cola in 1985 is the same lesson: 190,000 taste tests confirmed that consumers preferred New Coke, yet what mattered was the affinity for the brand. Product-level data alone did not capture that equity.

• If the decision is about how the brand is perceived, use brand image or equity research (tracking, perceptual maps, loyalty pyramid) and not product-feature tests alone.
• If you need to know why consumers behave as they do, run qualitative work first; if you need to know how many, run quantitative research next. Never size an unexplored hypothesis.
• If you want the true effect of advertising or promotions on sales, use market mix modelling on historical data, with baseline and incremental volume separated, not raw sales comparisons.
• If a launch is about to go to market, validate with a simulated test market or controlled store test before committing full spend.
• If your respondents are dispersed or hard to reach, use online or video qual, but only with an experienced moderator.

• Product-centric thinking: you test functional superiority and ignore brand affinity. Tell: your pitch is a feature-by-feature comparison and your research never measured brand perception (ZEN versus iPod; New Coke).
• Method-first, decision-last: studies run with no named decision. Tell: reports end at findings with no recommendation, owner or date.
• Single-lens measurement: you rely on surveys, or on sales data, but not both. Tell: attitudes and panel or tracking behaviour are never compared.
• Unmoderated online qual: discussion boards drift into survey-style Q&A. Tell: short, shallow answers with no probing and no non-verbal reads.
• Coverage gaps hidden by strong areas: advertising is tested but pricing, promotion or distribution runs on opinion. Tell: no method is named for a whole marketing function.

• ▸Marketing domain: brand, consumer, product, advertising, price and promotion, or retail.
• ▸Decision to inform: the specific choice at stake in this domain.
• ▸Method in use: the technique currently applied, or 'none'.
• ▸Attitudinal evidence: survey, qual or equity measure supporting the decision.
• ▸Behavioural evidence: panel, tracking, transaction or sales-model data supporting it.
• ▸Qual-quant link: how qual hypotheses were sized, or quant findings explained.
• ▸Recommendation and owner: the action taken and who is accountable.
• ▸Outcome check: the tracking or panel measure that confirms the result and its date.
• ▸Gap rating and next step: High, Medium or Low, with the method to add.

1. 1Separate the product from the brand before you analyse anything. Per the passages, a brand is 'a consumer's idea of a product' and lives in the consumer's mind as memories, feelings and associations. Write down what the product delivers (functional, rational, emotional, personality, relationship benefits) and what the consumer perceives on each of those benefits. The gap between the two is your first insight target.
2. 2Track brand image with a battery of attribute statements covering perceived benefits, imagery, symbolism and attitudes. The supermarket example uses 22 attributes. Run factor analysis to collapse them into a manageable set (22 became 5 factors: Variety & Quality, Location, Price and Value for Money, Store Experience, Loyalty Program).
3. 3Measure equity, then explain it. Knowing how much equity you hold does not reveal what is raising or lowering it. Regress the factors, plus awareness and consideration, against a brand equity index to get each driver's importance. In the source's FMCG shopping example, drivers include Consideration 18%, Awareness 15%, Location 19%, and Price and Value for Money 18%.
4. 4Cut every reading by demographic or target segment. Equity varies across groups (Exhibit 2.10) and reveals brand strengths and weaknesses that the total sample hides. Define the segment your strategy targets and do the final analysis on that base only.
5. 5Overlay your image profile on driver importance. List where you are strong and where you are weak, then check whether your strengths sit on high-importance attributes. Do this for competitors on the same base so you see relative position, not absolute scores.
6. 6Add behavioural evidence to attitudinal evidence. Read loyalty pyramid movements alongside equity trends (the source tells you to observe how store equity trends are reflected in the loyalty pyramid). Use consumer panel and transaction data to confirm that what people say matches what they buy.
7. 7Test each recommendation against strategy and feasibility before you write it. Ask whether the fix fits the brand's target and positioning, whether perceptions can realistically move, and what the move costs elsewhere (margin, service quality).
8. 8Listen for emotional signals, not just scores. The New Coke episode shows taste tests said consumers preferred the sweeter formula, yet over 400,000 calls and letters showed a depth of attachment the test did not capture. Add an attachment check to any test of a change to an established brand.

Inulas, a supermarket chain, looks misaligned at first glance. Exhibit 2.17 shows it strong mainly on low-importance attributes and weak on price and location, the attributes of greatest importance to shoppers. An analyst stopping there would recommend a price repositioning. The deeper read starts from Exhibit 2.10: Inulas' equity index is 4.3 among high income homes against 0.8 among low income homes, so the banner is targeting upmarket shoppers. The analysis is therefore rerun on the high income base (Exhibit 2.18), where Inulas' positioning is better aligned with the factors that drive those shoppers, and it outperforms market leader Fujimart on a number of attributes. 'Low price' and 'value for money' remain weak spots, but the source advises against shifting focus there: it is hard to alter perceptions 180 degrees, and lower prices would erode the margins that fund superior services. The insight is that the right base and the strategy's target change the conclusion entirely.

• If a brand's equity index differs sharply across demographic groups, then run the driver and image-profile analysis on the target segment only before drawing conclusions.
• If you know your equity level but not what drives it, then do not recommend action; regress awareness, consideration and image factors on equity first.
• If your weakness sits on a high-importance attribute but your strategy targets a segment that values something else, then check the target-segment profile before correcting it.
• If closing a weakness requires reversing a deeply held perception or sacrificing the margin that funds your differentiator, then reinforce your strengths instead of chasing the gap.
• If you are changing a long-established brand, then treat attitudinal and emotional attachment data as equal to preference-test results, because the Coca-Cola taste tests favoured New Coke but consumers still forced the original back.

• Reading the total-sample image profile for a brand that targets a niche. Tell: the conclusions say you are weak where your actual target shoppers do not care.
• Reporting equity levels without drivers. Tell: the deck states scores and trends but cannot say what action would raise them.
• Treating attribute scores as equal in weight. Tell: you list strengths and weaknesses without ranking them by importance, so a strength on a low-importance attribute is celebrated.
• Trusting a product test as proof of brand acceptance. Tell: a preference test result is the sole justification for changing an iconic brand (the New Coke pattern), and no attachment or heritage measure exists.
• Prescribing a fix that damages the strategy. Tell: the recommendation to cut prices on a premium-service banner has no margin or positioning check attached.

• ▸Target segment: define the consumers your strategy aims at and the sample base used for all analysis.
• ▸Equity by segment: record the equity index for each demographic group (e.g., high vs low income) and note where the target stands.
• ▸Attribute battery: list the image statements tracked (benefits, imagery, symbolism, attitudes) and the factors they reduce to.
• ▸Driver importance: record each factor's contribution to equity, including awareness and consideration, from the regression.
• ▸Own profile vs competitor: for each factor, mark strong or weak for you and for the market leader on the target base.
• ▸Alignment check: state whether your strengths fall on high-importance drivers, and name any weak spot on a top driver.
• ▸Behavioural cross-check: note what loyalty pyramid or panel/transaction data shows and whether it agrees with the attitudes.
• ▸Feasibility and trade-offs: for each proposed fix, record whether perceptions can shift and what it costs in margin or service.
• ▸Recommendation: one action tied to the strategy, with the evidence line that supports it.

1. 1Anchor every marketing decision to a segmentation base before allocating budget. Use the McDonald and Dunbar definition operationally: split customers into groups within which they share a similar level of interest in the same set of needs, so each segment maps to a distinct marketing proposition.
2. 2Score customers with RFM (recency, frequency, monetary) to grade them and prioritize spend. Because roughly 20% of customers provide 80% of income (Shin and Sohn, 2004), rank cells by CLV and direct retention and acquisition budget toward the top-value cells rather than spreading evenly.
3. 3Sharpen weak RFM cells by extending the model: add criteria from literature and expert opinion, weight the KPIs (e.g., via FGSWARA), then cluster (e.g., Fuzzy C-Means) so segments reflect purchase behavior rather than three raw variables alone.
4. 4Before launching a new product, simulate which existing product it will cannibalize using acceptability indices and central weight vectors (SMAA). Compare the new product's buyer profile to the average buyers of existing products to predict competitive overlap.
5. 5Validate your data-to-decision pipeline against known outcomes. Run the model against products with known market shares and compute MAPE; only trust the decision engine for forecasting once error is demonstrably low (the SMAA case hit 0.18%).
6. 6Feed qualitative signal into decisions with sentiment/opinion mining on reviews to catch orientation of opinion the RFM numbers cannot see, then adjust messaging and product positioning per segment.
7. 7Institutionalize a periodic CLV evaluation plan so decisions stay timely; RFM is a backward-looking metric, so re-score on a cycle to catch early chain signals of churn or value shift.

In the SMAA consumer-preference study, two new products N1 and N2 are evaluated before launch. Analysis of central weight vectors shows the N1 buyer profile is closest to the average customer at large, but the acceptability index for N1 is low — meaning that average-like group is actually very small, so allocating heavy launch spend to N1 would chase a thin market. N2's buyers instead resemble the average buyer of existing product B, signaling N2 will compete mainly with B and cannibalize it. Comparing N1 and N2 profiles shows they are mutually very competitive, but N2's situation is much better, so the effective decision is to prioritize N2. The team also learns a positioning nuance: when N2 is available, N1's remaining buyers are those who focus on price much more than on fulfillment time. To trust these decisions, the method was validated by comparing acceptability indices to actual market shares for products A through E, yielding a MAPE of only 0.18%, which confirms the model is accurate enough to allocate real budget on.

• If a segment holds top RFM/CLV scores, then concentrate retention and premium marketing spend there first — the 20/80 income concentration makes even allocation wasteful.
• If a candidate product's buyer profile matches the average buyer of an existing product, then treat that existing product as the cannibalization target and forecast net (not gross) incremental gain.
• If a new product's profile is closest to the average customer but its acceptability index is low, then do not scale spend — the addressable group is small despite looking mainstream.
• If model acceptability indices reproduce known market shares within a low MAPE, then approve the model for forecasting decisions; if error is high, then refine criteria/weights before acting.
• If a segment's buyers weight price far above fulfillment time (or vice versa), then differentiate the proposition on that dominant attribute rather than a generic offer.

• Deciding off raw three-variable RFM alone. Tell: segments feel generic and don't map to distinct propositions — fix by adding weighted KPIs and clustering.
• Treating RFM as forward-looking. Tell: you're surprised by churn the score never flagged; it's a backward-looking metric requiring periodic re-scoring.
• Ignoring cannibalization when launching. Tell: total category revenue barely moves after launch because the new product ate an existing one's buyers (N2-competes-with-B pattern).
• Trusting model output without validation. Tell: no MAPE-against-known-shares check exists, so allocations rest on an unverified engine.
• Picking a product because its buyer looks like the 'average customer.' Tell: high apparent representativeness but a low acceptability index means the real segment is tiny.

• ▸Segment name: the distinct group sharing a similar need set
• ▸RFM cell / CLV rank: recency-frequency-monetary score and value tier
• ▸Dominant decision attribute: what this segment weights most (e.g., price vs. fulfillment time)
• ▸Distinct proposition: the one message/offer matched to this segment
• ▸Cannibalization target: existing product whose buyer profile this offer overlaps
• ▸Acceptability vs. size check: is apparent representativeness backed by a real (non-tiny) segment? yes/no
• ▸Model validation MAPE: forecast error against known market shares before acting
• ▸Budget allocated: spend directed here and its justification by tier
• ▸Re-score date: next scheduled CLV/RFM refresh to keep the decision timely

1. 1Define the perception set before you collect anything. Perceptions have three layers: functional benefits, emotional imagery, and symbols or attitudes. Write attribute statements for all three. The source's supermarket study used 22 statements covering perceived benefits, imagery, symbolism and attitudes. A list of only functional claims misses what Argyle Pink shows, namely that intrigue, fascination and fame are what sustain the premium.
2. 2Rate every brand, yours and competitors', on the same statements. Use a 5-point agree scale: 1 Strongly disagree, 2 Disagree, 3 Neither, 4 Agree, 5 Strongly agree. Perceptions are only meaningful relative to other brands, so a single-brand survey cannot show differentiation.
3. 3Build the image profile: compute an Expected Score for each brand-attribute cell, then Profile = Actual Score − Expected Score. A positive profile means the brand owns the attribute more than its overall share of mentions would predict. A negative profile means it under-owns it.
4. 4Standardise the gaps across cells so they are comparable. Chi-square value = Profile² ÷ Expected Score. Then apply the sign of the Profile to get a signed association measure. Without this step, large attributes dominate simply because they are large.
5. 5Reduce the attributes to a few dimensions with correspondence analysis, and plot brands and attributes together on the perceptual map. Read it as the source prescribes: closer brands are perceived as more similar. A longer attribute line means that attribute differentiates the market more. Drop a perpendicular from a brand to an attribute line, and the farther along that line the brand sits from the origin, the higher its rating on that attribute.
6. 6Condense the attributes into factors (factor analysis took the 22 supermarket attributes down to 5). Regress the factors, plus awareness and consideration, on the brand equity index. This shows which perceptions drive equity, not just which exist. Knowing how much equity you have does not tell you what to fix.
7. 7Track on a repeating schedule and compare to the prior wave. Check that movements in perception line up with movements in loyalty and equity measures. The source notes that store equity trends were reflected in the loyalty pyramid.
8. 8Treat measured perception as a stated reading, not the full picture. Check it against behavioural and emotional reaction. In the New Coke episode, taste tests showed a preference for the new formula, yet consumers rejected it.

The source's shampoo image profile compares two brands (Zhi Zi and Tare in the exhibit title; the column order is not legible in the extract). Take 'Strengthens hair'. Actual scores are 23 and 43, and expected scores are 27.7 and 21.4. Profiles are 23 − 27.7 = −4.7 and 43 − 21.4 = +21.6. Chi-square values are 4.7² ÷ 27.7 ≈ 0.8 and 21.6² ÷ 21.4 ≈ 21.8. With signs applied, these are −0.8 and +21.8, so one brand strongly owns 'strengthens hair' and the other barely registers. Compare 'Value for money': actuals 20 and 15 against expected 27.2 and 21.0 give profiles of −7.2 and −6.0, and signed chi-squares of about −1.9 and −1.7. Both brands under-own value, so it does not separate them. Correspondence analysis then turns these signed values into the map, where an attribute like Hair Expert or Prevents dandruff becomes a line and each brand is read by its perpendicular position. Action follows from the numbers: the brand with +21.8 defends that ownership, and the rival decides whether to contest it or claim a different attribute.

• If a brand's profile on an attribute is positive and its chi-square is large, treat the attribute as owned and protect it. If the chi-square is near zero for every brand, as with 'Value for money' in the example, the attribute does not differentiate, so do not build positioning on it.
• If two brands plot close together on the perceptual map, consumers see them as similar. Either differentiate on an attribute with a long line where you sit farther out, or accept direct competition.
• If equity falls and you do not know why, run factor analysis and regression on the attributes plus awareness and consideration before recommending action. Do not act on the equity score alone.
• If a perception metric moves but loyalty-pyramid or equity trends do not, suspect the attribute set or the sample before concluding the brand has changed.
• If a research result (such as a taste test) conflicts with the symbolic or emotional meaning the brand carries, weigh the emotional layer heavily before changing the product. New Coke won taste tests and still met consumer rejection.

• Measuring only functional attributes. The tell: your attribute list has no imagery, symbol or attitude statements, and you cannot explain a premium the way Argyle Pink's exclusivity explains its price.
• Reading raw scores instead of profiles. The tell: the brand with the biggest base 'wins' every attribute, and the data show no differentiation.
• Mistaking proximity on the map for equality of strength. The tell: you call two nearby brands 'strong' without checking the perpendicular distance from the origin along the attribute line.
• Reporting equity without its drivers. The tell: the deck says equity moved by X points but cannot say which factor, awareness or consideration contributed.
• Trusting a rational test over the brand's meaning. The tell: a preference result is used to justify a change, with no probe of the associations consumers hold.

• ▸Brand and competitor set: list each brand rated; all must use the same statements.
• ▸Attribute statements (functional / emotional / symbolic): list each statement and tag its layer.
• ▸Rating scale and wave date: confirm 1–5 agree scale and the survey period.
• ▸Actual score per brand-attribute cell: enter the observed score.
• ▸Expected score per cell: enter the expected value using the image profiling method.
• ▸Profile (Actual − Expected): compute and record with sign.
• ▸Signed chi-square (Profile² ÷ Expected, with sign): compute for every cell and mark the largest positives and negatives.
• ▸Perceptual map reading: note nearest competitor, the longest attribute lines, and each brand's perpendicular position.
• ▸Driver ranking: list factor importances from the regression, including awareness and consideration, and name the top two to act on.

1. 1Pick the survey type first. The source chapter splits satisfaction research into a transaction survey and a relationship survey. Decide whether the decision concerns a specific interaction or the cumulative relationship, because your definition of satisfaction is cumulative experience and each type measures a different slice.
2. 2Run qualitative work before the quantitative survey. Use depth interviews or group discussions to find the issues worth tracking, then carry them into the quant questionnaire. Probe satisfaction and dissatisfaction with the category and the issues customers meet and how they are resolved. These are listed as standard usage-and-attitude content.
3. 3Build the questionnaire in the standard order: screener, main topic, demographics. Run from broad to specific and put sensitive questions last.
4. 4Put the overall satisfaction rating before any attribute ratings. Asking it after rating colour, smell, sugar and the like lets the attribute feelings bias the overall score (order effect). Also watch for halo effects across the attribute list.
5. 5Add outcome questions beside satisfaction: willingness to recommend (the net promoter question), favourite brand (which may differ from the one bought most often), and willingness to pay or to endure inconvenience. The source says these statements are highly correlated, so three or four are enough and a longer battery adds little.
6. 6Read attribute scores against an expected score. Expected = AVG(brand score) × AVG(attribute score) ÷ AVG(all scores). Profile = actual − expected. This removes brand size and attribute size, so a satisfaction gap is not just a generous-rater or popular-brand artefact. The source builds this for brand image; applying it to a satisfaction attribute grid is an extension.
7. 7Segment by loyalty. Combine behavioural loyalty (purchase data) with attitudinal data from satisfaction or equity studies, as the loyalty matrix does, and tailor actions to each segment. Check that a segment can be targeted and tracked through media habits, outlets or pack size.

The source's Volvo example shows the profile calculation. Volvo's average brand score is 39.7, the average score for the attribute safety is 53, and the overall average is 48. Expected safety rating = 39.7 × 53 / 48 = 43.8. Volvo's actual rating is 52, so its profile is +8.2 (52 − 43.8). The raw 52 overstates the lead only if you ignore that safety scores high for every brand. Applied to a satisfaction grid, you would run the same arithmetic per attribute and read only the deviations, which sum to zero across rows and columns. Separately, the source's drink-test sequence (colour, smell, cocoa, sugar, chocolate chips, then overall rating) is the named mistake. You would move the overall question to the top of the same questionnaire.

• If the decision concerns a specific interaction, use a transaction survey. If it concerns the overall relationship and loyalty, use a relationship survey.
• If you do not yet know which issues matter to customers, run qual first and only then fix the quant attribute list.
• If an overall rating question exists, it goes before all attribute questions.
• If raw attribute scores look high across all brands, read the profile (actual − expected) rather than the raw score.
• If you segment by satisfaction or loyalty, proceed only if the segment can be reached through distinct media, outlets or pack sizes. Otherwise treat it as a description, not a target.

• Overall rating placed last. The tell: overall satisfaction tracks the attribute scores almost perfectly, because the attributes primed it.
• Reading raw attribute scores as strengths. The tell: every brand "wins" on the same popular attribute, because brand size and attribute size are inflating the score.
• Substituting cheap focus groups for a quant survey, or reading groups as measurement. The tell: conclusions rest on a handful of voices with no base sizes.
• Building an attribute list from internal assumptions. The tell: scores look fine while retention falls, because the questionnaire never covered the issues customers actually raise.
• Segments that cannot be targeted. The tell: a tidy loyalty segmentation that no media or outlet plan can act on.

• ▸Decision to inform: state the choice the study supports.
• ▸Survey type: transaction or relationship, with one line on why.
• ▸Qual findings: list the issues and problems customers raised that must appear in the quant survey.
• ▸Question order: confirm the overall satisfaction rating precedes the attribute ratings, and note the position of the screener, demographics and sensitive questions.
• ▸Outcome questions: recommend, favourite brand, price premium or inconvenience statements chosen.
• ▸Attribute grid: for each attribute record the brand score, attribute average and overall average.
• ▸Expected score and profile: compute AVG(brand) × AVG(attribute) ÷ AVG(all), then actual − expected.
• ▸Loyalty segments: combine behavioural loyalty with attitudinal satisfaction and name each segment.
• ▸Targeting feasibility: for each segment, the media, outlet or pack-size route to reach it.