TASTE INTELLIGENCE
Attribute → Perception → Emotion → Value taste structure, measured from search-behavior data.
Quotes, search terms, purchase-situation names and knowledge-graph items are shown in Korean, as used in this market.
Audience Taste — the taste structure of consumers searching '바나나우유'. Based on measured query voice and searcher demographics.
Every figure carries an evidence label — measured = raw data; derived = measured values mapped onto a validated framework.
Means-End Chain
A Hierarchical Value Map (means-end chain). Node size = the card's weight (per the weight basis above; solid part = consumer words); solid links = measured in the same sentence, dashed = convention where no evidence.
Do not compare across categories — taste labels reuse the same names but point at different things. ‘Sweet’ in whisky means sherry-cask dried fruit and vanilla; in wine it means residual sugar. Never compare the same label's figures between categories.
In each bar, the solid part is consumer words (queries, community, social, blogs) and the pale part is words the market applies (press, commerce, brands, AI Overviews).
The ladder redrawn from consumers' own words only (queries, community, social, blogs): card size, order and solid links all come from those sentences. Write copy from this ladder; read market positioning from the full one.
Attribute: a feature you can see or measure · Perception: words that judge the product · Emotion: how the person feels · Value: the life goal it serves. Tap any card to open its evidence and reading. Selecting a card keeps only the attributes, impressions, emotions and values connected to it; the rest fade. Card size and numbers are measured (search volume). Solid lines are measured links (the two appeared in the same sentence; thicker = more often); dashed lines are first-order inference from convention where no evidence was found. Hover a line to see its basis. Faded cards rest on a single source sentence. Fainter dashes are links that skip a step (emotion or impression) that was not measured in this capture.
Clean · Unsweetened — clean, unsweetened or low-fat
This node exists because expressions like '무가당' were actually found. Sources: Organic snippet 9 · AI Overview (AIO) 1 · AIO source / PAA answer 1 · Related query 1.
Value region Conservation · Brand personality (Aaker) Competence
Sweet — rich, satisfying sweetness
This node exists because expressions like '꿀맛' were actually found. Sources: Organic snippet 5 · AI Overview (AIO) 2 · Related query 1 · Result title 1.
Co-occurring terms (sentences): Specialty · Rare 2
Value region Hedonism · Brand personality (Aaker) Sincerity
Organic · Farm — organic, farm-fresh
This node exists because expressions like '유기농' were actually found. Sources: AIO source / PAA answer 4 · Organic snippet 3.
Value region Self-transcendence · Brand personality (Aaker) Sincerity
Smooth · Creamy — creamy, smooth
This node exists because expressions like '부드럽고' were actually found. Sources: Organic snippet 1.
Value region Conservation · Brand personality (Aaker) Sincerity
Premium — premium ingredients and finish
This node exists because expressions like '고급' were actually found. Sources: Related query 1 · Organic snippet 1.
The impression Premium leads to the emotion Special · Pampered, and on to the value Self-enhancement — the upward path of a means-end chain.
Value region Self-enhancement · Brand personality (Aaker) Sophistication
Specialty · Rare — rare origin, variety or craft
This node exists because expressions like '독특한' were actually found. Sources: AI Overview (AIO) 1 · Organic snippet 1.
Co-occurring terms (sentences): Sweet 2
Value region Openness · Brand personality (Aaker) Excitement
Joyful · Fun — the emotional payoff consumers expect from this taste.
Detected words: 즐거 · 재미있 · 재밌 · 웃음이 · 행복해지 · 행복 · 기분좋아지 · 기분업 · 기쁨 · 기뻐 · 신이나 · 흥겨 · 빠져드 · 도파민 · 고소해서계속
Satisfied — the emotional payoff consumers expect from this taste.
Detected words: 만족 · 후회없 · 후회하지 · 실패없 · 실패하지 · 잘샀 · 돈이아깝지 · 아깝지않 · 값어치 · 본전
Curious · Intrigued — the emotional payoff consumers expect from this taste.
Detected words: 호기심 · 신기 · 새로운자극 · 색다른경험 · 처음느껴 · 놀라운 · 반전매력 · 궁금증
The impression Specialty · Rare feeds into Curious · Intrigued.
Special · Pampered — the emotional payoff consumers expect from this taste.
Detected words: 특별해지 · 대접받는 · 나를위한 · 플렉스 · 과시 · 격이다르 · 호사스러 · 호사누리 · 사치스러운기분 · 특별한날기분
The impression Premium feeds into Special · Pampered.
The measured value is far below the derived one because people rarely type emotions into a search box — they search 'Yamazaki 12 price', not 'Yamazaki makes me happy'. Emotion language lives mostly in reviews and AI Overview text, so read the measured figure as a lower bound.
Power · Status: status and prestige. It sits in the ‘Self-enhancement’ group of the Schwartz value circle; the opposite group is ‘Self-transcendence’.
Premium group into the value Power · Status.
Self-direction: choosing and expressing in one's own way. It sits in the ‘Openness’ group of the Schwartz value circle; the opposite group is ‘Conservation’.
Specialty · Rare group into the value Self-direction.
Conformity: fitting in, not standing out. It sits in the ‘Conservation’ group of the Schwartz value circle; the opposite group is ‘Openness’.
Clean · Unsweetened group into the value Conformity.
Hedonism: sensory pleasure. It sits in the ‘Hedonism’ group of the Schwartz value circle; the opposite group is ‘Conservation’.
Sweet group into the value Hedonism.
Universalism: living in harmony with nature and others. It sits in the ‘Self-transcendence’ group of the Schwartz value circle; the opposite group is ‘Self-enhancement’.
Organic · Farm group into the value Universalism.
Security: safety and stability. It sits in the ‘Conservation’ group of the Schwartz value circle; the opposite group is ‘Openness’.
Smooth · Creamy group into the value Security.
For people searching '바나나우유', 6 taste signals were detected, and the strongest is Clean · Unsweetened (7,210). That single family accounts for 44% of all taste weight and is 1.9x the runner-up Sweet (3,710). In consumer language, the people behind this search respond to "clean, unsweetened or low-fat".
The top three hold 85% of the total, which is a concentrated structure. Consumers largely agree on what counts as good in this category. That consensus is both a barrier and an opportunity: proposals outside the axis carry a heavy explanation cost, while fine differentiation inside it is recognised cheaply.
Read vertically, the attribute layer (sensory fact) peaks at Clean · Unsweetened (7,210) and the perception layer (social meaning) at Premium (1,050). The heavier attribute layer means consumers are still exploring what it actually tastes, smells or feels like. At that stage concrete product description outperforms abstract brand narrative.
The ladder converges on the value 'Conservation', with Joyful · Fun strongest in the emotion layer just below it. That is what consumers ultimately want confirmed through this category, so closing a campaign in the language of that value, rather than of an attribute, keeps the message aligned with the ladder.
Practically, make Clean · Unsweetened the spine of the message and place Sweet and Organic · Farm as supporting appeals. Product pages that start at the attribute layer (Clean · Unsweetened), rise to perception (Premium) and close on value (Conservation) follow the path consumers actually walk.
Two cautions. These figures measure how many people put a taste into words, not how strongly they prefer it. Tastes that never reach a search box stay invisible, so absence of signal is not evidence of absence of taste. And these are liking-side figures only; whether the same attribute also repels is settled in the next section.
Who said it
Measured classification of search-result URLs and AI Overview citation URLs. Communities, social and personal blogs are words consumers wrote themselves; press, commerce, brand, reference sites and AI Overview text are the market's words. The same sentence carries different weight depending on where it came from.
74% of this analysis came from consumer voice (queries + community, social, blogs).
To know whether a word is what consumers actually say, look at where it came from. Queries show intent but carry almost no sensory or emotional vocabulary; that language lives in community posts and personal blog reviews. If this mix leans toward commerce and press, the taste words in this report are closer to the market's language than to consumers'.
Watch what the AI Overview cites, too. If it only cites commerce pages, generated answers in this category are filled with selling language — which means there is room for a brand to get in with content.
Drivers of Liking
The same attribute can be wanted or avoided (sentiment ≠ preference). Right = liking volume, left = measured negation/JAR patterns ('not too …').
No repelling taste surfaced in this capture. Lead with the top liked taste, Clean · Unsweetened.
Read the avoidance side as a floor, not a measurement. People search for what they want, rarely for what they dislike, so this axis is structurally under-counted. A small number here is not evidence that avoidance is absent. If any part of this report warrants a quantitative follow-up study, it is this axis.
Preference Segments
Pick gender/age to see how strongly that group over-indexes on each taste (lift).
This section crosses taste with demographics. Pick a gender and an age band and each card re-ranks to show how many times more that group leans into each taste (the preference index). Above 1.15 is over-indexed, below 0.85 under-indexed.
Demographic-only segments are easy to reach but explain nothing about why people buy; taste-only segments explain well but cannot be reached. The numbers in this cross-tab are the bridge. Use them to check that the demographic target of the media plan and the taste target of the creative are not pointing in different directions.
Measured demographics for this run are 37.7% male / 62.3% female, concentrated in 40-49. Any taste whose own distribution departs sharply from this baseline is where segmentation has something to offer.
The four lines on each card answer different questions. Wants is what they like; Acts is where in the purchase journey that taste sits; Avoids is what causes drop-off; Spreads is which adjacent category the taste extends into. For cross-sell and collaboration candidates, read Spreads.
Tastes marked as having no demographic sample come from SERP text only. They are valid as tastes but should not carry demographic interpretation. The preference index is also approximated from the product of marginals, so genuine interactions such as a spike specific to women in their twenties need separate verification.
Context × Taste
Taste is situational, not fixed (Person × CEP × Attribute). Cells = measured co-occurrence of context and taste words in the same query/snippet.
| Clean · Un | Sweet | Organic · | Premium | Specialty | Smooth · C | |
|---|---|---|---|---|---|---|
| Home · Solo | 210 |
In the "Home · Solo" context, Sweet switches on most strongly. Switching on here means the two vocabularies co-occur in the same search text with the greatest weight.
Treat taste as a fixed trait and each person belongs to one segment; treat it as something context switches on and off and the same person belongs to several. The latter predicts better, which in practice means when to speak is a bigger lever than whom to speak to.
This matrix feeds a seasonal calendar directly. If different contexts switch on different tastes, the same product needs different copy by period; if the signal concentrates in one context, concentrating budget there is the efficient move.
Co-occurrence is not causation, and low-volume cells are statistically unstable. Read the top cells.
Perceptual Map
The horizontal axis asks “toward others or toward myself”, the vertical asks “seeks the new or keeps the familiar”. These are plain-language versions of Schwartz's value circle (self-transcendence↔self-enhancement · conservation↔openness). Bubble size = search volume.
Hover a bubble for its share, search volume, layer, value region and the expression actually found. Only the six largest are labelled by default; press ‘Show all labels’ above the map to see every name.
The centre of gravity sits in "Conservation". The two axes come from the Schwartz value circumplex: self-transcendence versus self-enhancement horizontally, conservation versus openness vertically. Adjacent regions are compatible; opposing regions conflict.
A centre of gravity in "Conservation" means that is the value consumers express through taste in this category. Openness lies diagonally opposite, so claiming both at once cancels the message. It is worth checking whether the brand is currently saying both.
For extension decisions this map prices the move. Shifting to an adjacent region is cheap because it does not break existing perception; a diagonal move is repositioning in all but name. Where you see empty space, check bubble sizes to tell an opportunity apart from a space that is empty because nobody wants it.
Coordinates are theoretical mapping, not measurement, so read relative arrangement rather than absolute position. Schwartz values are individual-level theory, so applying them to category taste involves analogy. The map shows the size of a taste, never its profitability.
Preference → Choice
Horizontal: whether the source text is exploring or buying (measured). Vertical: whether consumers wrote the word or the market applied it (measured). Top-right = words consumers use right before buying, i.e. ad-copy material. Bottom-left = words the market applies during exploration, i.e. content and SEO material.
Reading: ‘Clean · Unsweetened’ sits furthest top-right at 28% consumer voice and 88% funnel. Consumers actually use it and close to the buying stage, so it is the first candidate for ad copy. Closest to the buying stage: Clean · Unsweetened · Organic · Farm. 2 were left unplaced for lack of any funnel-stage signal. Hover a bubble for its share, search volume, layer, value region and the expression actually found. Only the six largest are labelled by default; press ‘Show all labels’ above the map to see every name.
Brand/Keyword Symbol — what taste '바나나우유' stands for in the public mind. Based on measured market description language (AIO·snippets) and reference patterns ('바나나우유-style', 'like 바나나우유').
Every figure carries an evidence label — measured = raw data; derived = measured values mapped onto a validated framework.
Symbol Profile
Taste words extracted from market description language (AIO bodies · organic snippets), not from queries — 'what this keyword is described as', not 'what people want'.
No searches using '바나나우유' as a taste benchmark were found in this capture — it may not yet be a taste symbol, or the signal is in the long tail.
The market describes '바나나우유' as Clean · Unsweetened·Sweet. However, no trace was found in this capture of consumers using the keyword as a benchmark for something else. A described taste has formed, but it has not yet reached the stage of a symbol.
At this stage the work is to make the description language consistent. If the vocabulary the market uses to describe this brand differs from the vocabulary the brand uses about itself, that gap is likely what blocks symbolisation. Put the description list above next to your own messaging guide and compare.
Note that the audience taste measured in Part A and the described taste here in Part B are different data. Where they agree, brand perception and consumer expectation are aligned; where they diverge, the gap itself is the positioning task.
Reference patterns live mostly in other categories' search corpora, so a single-keyword capture under-detects them. Zero references is not proof that no symbolism exists. AI Overview text is a generative summary, so the market's language and the model's language may be mixed.
Brand × Taste
Which taste words co-occur with each KG brand in the corpus (Brand-Concept-Map style, measured co-occurrence).
| 빙그레 | Clean · UnsweetenedSweet | 5,526 |
| 푸르밀 | PremiumClean · Unsweetened | 500 |
Knowledge Graph
Decision structure extracted & quote-verified by the Ontology Builder (8070) from search, SERP and AI-Overview bodies. LLM-assisted (source kept distinct).
Statistics
Taste × demographic association via Cramér's V. weak<0.1 · med 0.1–0.3 · clear 0.3–0.5 · strong>0.5.
This index is not computed on a probability sample; it converts search-volume weights into cell counts. The conventional Cramér's V thresholds (0.1/0.3/0.5) are therefore not applied. Use it only to compare axes within this one report.
Gender association measures 0.048 (little difference) and age association 0.156 (small), with Taste × Age differences relatively larger. Cramér's V normalises the chi-square statistic by sample size and degrees of freedom, so it does not inflate with sample size.
A weak value means demographic targeting has little to offer here. Taste is splitting within the same age band, so targeting by taste and context will probably outperform demographic targeting on the same budget. This is not bad news; it is a budget allocation instruction.
This is association strength, not causation. The contingency tables convert search-volume weights into cell frequencies rather than being a strict probability sample, which is why effect size is reported and p-values are not.
| Gutman (1982) · Reynolds & Gutman | Means-End Chain / Laddering → Hierarchical Value Map. | L2 |
| Schwartz (1992) | Basic human values — circular structure. | L3 |
| Aaker (1997) | Dimensions of Brand Personality, JMR. | L3 |
| Li et al. (2015) J. Food Science | Drivers of Liking — sensory attributes → acceptance. | L4c |
| Nagamachi — Kansei Engineering | Consumer feeling → design specification. | L3·L5 |
| John, Loken et al. (2006) JMR | Brand Concept Maps — brand association networks. | L4d |
| ABSA (IJHM 2026 · MDPI 2025) | Aspect × Sentiment — sentiment ≠ preference. | L1b |
| Hur, Etcoff & Silva (2023) | Preference structure of everyday aesthetic choices. | L4 |
| Bourdieu (1984) · Peterson (1992) | Distinction / Cultural omnivore — taste homology. | L4 |
| Cramér (1946) · Newman (2006) | Effect size · community structure. | L5 |
Taste Finder v2.0 · Taste Intelligence Framework v5
Give a keyword (category, product or brand) and it extracts sensory and emotional words from related searches, SERPs, AI Overviews and People Also Ask, groups them into taste tribes and saves a report. Product planners and marketers use it to choose product language and messages. Data comes from ListeningMind DaaS and Maistro lmpipe for Korea, the US and Japan. Reports can be compared.
Key ideasTaste tribePeople Also Ask (PAA)AI OverviewSearch result page (SERP)
Works well withCEP FinderBrand FinderBrand Naming Agent