> ## Documentation Index
> Fetch the complete documentation index at: https://docs.exoid.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Recipes by research type
> Twenty-five consumer study examples: the business question, the methodologies that fit and how to build each one in Exoid
Every study starts from a business question. This page collects the questions that come up most often in consumer market research and turns each one into a study you can build: which methodologies fit, which blocks and logic you use in Exoid, and what to look at in the results.
Read them as examples, not as prescriptions. The same business question can be answered by several designs, and the design you pick decides what you find out: a ranking tells you the order, a monadic comparison tells you the level, an open-ended question tells you the reason. Every recipe ends with a suggestion for building it a different way — try that too.
Every recipe names one or more methodologies. What each of them measures and how it works is in [Research methodologies](/en/methodologies).
Four use cases get the long treatment, because they carry the heaviest decisions and because the traditional way of running them costs the most.
How the brand is perceived against competitors, and whether it's improving.
Which message to put on pack or in campaign, out of many candidates.
How many groups the market splits into, and which one to focus on.
How the shopper gets to the purchase, and where you lose them.
The cost and time figures below describe the traditional research market. They come from Exoid's advisor in the July 2026 workshop and are market orders of magnitude, not Exoid prices or promises.
## Product and innovation
### Concept test
**The client's question.** "Does the idea work with my consumer target?"
**The methodologies.** [Monadic test](/en/methodologies#monadic-test), [Kano factors](/en/methodologies#penalty-reward-and-kano-factors), [MaxDiff](/en/methodologies#maxdiff).
**How you build it in Exoid.** Present the concept with a Message block, text and media together, and set **Wait before advancing** so nobody skips it. Then separate ratings on interest, uniqueness, credibility and perceived value — four distinct measures say far more than one overall score — plus a single choice on purchase intent. Close with an AI-moderated open-ended on the why: a reasoned "no" is worth more than a vague "yes".
To work out which features to invest in, the Kano exercise is a pair of questions per feature — how would you feel if it were there, how would you feel if it weren't — both on the same scale inside a matrix. You get the must-have, proportional and delighter classification by crossing the two answers in Excel or SPSS from the export.
**What to look at.** Purchase intent across [comparison clusters](/en/analyze/summary) of existing customers versus everyone else.
**Try it another way.** Run the same concepts monadically, one version per respondent with the [media variable](#showing-different-stimuli-to-different-respondents), and you measure the absolute appeal of each. Or show all of them and use Ranking: you get a sharper order but a friendlier judgement, because comparison makes people generous.
### Product test in use
**The client's question.** "Does the product hold up once people actually use it at home?"
**The methodologies.** [Multimodal qualitative analysis](/en/methodologies#multimodal-qualitative-analysis), [Key Driver Analysis](/en/methodologies#key-driver-analysis).
**How you build it in Exoid.** Two moments, two campaigns. Before use: expectations, with ratings and a purchase intent question. After a few days of use: performance on the same attributes, in a matrix with the same scale, plus a File upload block for a photo of the product in its real setting and an open-ended with [AI moderation](/en/builder/ai-moderation) on the first use. Pass the same respondent identifier as a [URL variable](/en/logic/variables) in both waves so you can join them on the export.
**What to look at.** The gap between expectation and experience, attribute by attribute. That gap is where returns and bad reviews come from.
**Try it another way.** Ask only after use, with a single interview and a recall question on what people expected. It costs half as much and you find out whether the disappointment is remembered — which is what drives word of mouth.
### Packaging test
**The client's question.** "Which pack works better on shelf and in use?"
**The methodologies.** [Monadic test](/en/methodologies#monadic-test), [MaxDiff](/en/methodologies#maxdiff), [implicit association test](/en/methodologies#implicit-association-test-single).
**How you build it in Exoid.** Image Choice is the natural block: you show the shelf with the competing packs and ask which one you'd pick up. Set **Wait before advancing** to a few seconds, because that's how long the decision takes on shelf, and always keep image randomization on — shelf position weighs as much as design.
Right after, a recognition question — which of these did you see — measures stand-out. Then ratings on legibility, premiumness and brand fit, and a pack × attributes matrix for associations. To understand the pack in use, the File upload block brings you a photo of it at home, opened and being used.
**What to look at.** The gap between "I noticed it" and "I'd buy it": two different problems, solved with different levers.
**Try it another way.** Show one pack per person with the [media variable](#showing-different-stimuli-to-different-respondents) and no competitors in sight. Stand-out disappears as a measure, and what comes forward is what the pack says on its own — often a completely different answer.
### Naming and logo test
**The client's question.** "Which name, and which mark, do we go out with?"
**The methodologies.** [MaxDiff](/en/methodologies#maxdiff), [implicit association test](/en/methodologies#implicit-association-test-single), [monadic test](/en/methodologies#monadic-test).
**How you build it in Exoid.** For names, a text-only exercise: with seven candidates or fewer a Ranking block gives you the order in one screen; with more, the most/least important sets described in the [claim test](#claim-test). Then a names × attributes matrix — easy to say, easy to remember, fits the category, feels premium — and an open-ended on what the name suggests, which is where you catch the meanings nobody in the room had thought of.
For logos, the same structure with Image Choice and **Supersize** on, so the mark is shown large enough to judge.
**What to look at.** Not the favourite, but the one with no strong rejections: names die from the people who hate them, not from the people who shrug.
**Try it another way.** Put the name on the pack and test it in context rather than on a white page. A name that wins in the abstract often loses once the brand block, the colour and the product are around it.
### Range and assortment
**The client's question.** "Which SKUs do I offer? In store and online?"
**The methodologies.** [TURF](/en/methodologies#turf), [MaxDiff](/en/methodologies#maxdiff), [Choice-Based Conjoint](/en/methodologies#choice-based-conjoint).
**How you build it in Exoid.** The input to TURF is a single question done well: a multiple choice, "which of these would you actually buy", across the whole candidate range, options in random order and no selection cap. Alongside it, a Number Input on purchase frequency and a Ranking for priority. Export the multiple choice and compute combined reach — which set of SKUs covers the most people without overlapping — in your own tool.
If the range differs by channel, ask the same thing twice, splitting store and online with [navigation rules](/en/logic/conditions), or create a comparison cluster by channel in the results.
**What to look at.** Not the top-voted SKU, but the one that adds new people to those already covered by the first choices.
**Try it another way.** Ask the same question as a shelf: Image Choice with the full range in a grid, a **Range** selection mode of one to three, and **Wait before advancing** on. You get the basket, not the wish list.
### Pricing
**The client's question.** "What price do I sell at? What price do I promote at?"
**The methodologies.** [Price Sensitivity Meter](/en/methodologies#price-sensitivity-meter), [Choice-Based Conjoint](/en/methodologies#choice-based-conjoint).
**How you build it in Exoid.** The Price Sensitivity Meter is four Number Input blocks in sequence, in the canonical order — at what price does the product feel too expensive, expensive but acceptable, cheap, so cheap you'd doubt the quality — with the same product description repeated in the instructions. Four questions, ten minutes of building. You plot the four curves in Excel from the export and read the acceptable range off the crossings.
Conjoint is the more ambitious route: each task is a single choice block between complete product profiles, and **Create in bulk** plus **Add Bulk Options** make building twelve or sixteen of them quick. You prepare the task plan with your design tool, field it in Exoid, export the dataset to SPSS and estimate utilities and price scenarios there.
**What to look at.** From the Price Sensitivity Meter, the acceptable range by segment: it's often narrowest in the segment that buys most.
**Try it another way.** Skip the stated price entirely and put the product on a shelf at three different prices, one price per respondent with the [media variable](#showing-different-stimuli-to-different-respondents). What you measure then is choice, not opinion about price — a much harder test to pass.
## Communication
### Claim test
**The client's question.** "Which claim is the most immediate? Which one convinces the consumer to pick us?" It's asked by brand managers, marketing research managers and innovation managers, with R\&D or regulatory affairs covering claim substantiation; in retail, the private label product manager. Almost always under time pressure, against a packaging or campaign launch deadline.
**How it gets solved today.** With institutes, through monadic concept and claim tests: a few weeks of work, orders of magnitude between ten and thirty thousand euros. Otherwise with a quick homemade round of questions on whatever panel is at hand. Otherwise it isn't tested and the most senior opinion wins. Time pressure weighs so heavily that people sometimes pick the fastest agency rather than the best one, and the decision can be taken before the results are even delivered.
**The methodologies.** [MaxDiff](/en/methodologies#maxdiff), [monadic test](/en/methodologies#monadic-test), [implicit association test](/en/methodologies#implicit-association-test-single).
**How you build it in Exoid.**
When there are many candidate claims — eight, twelve, twenty — the question isn't whether people like them but in what order they go. You build the MaxDiff exercise like this: prepare subsets of four or five claims, and for each subset add two single choice blocks with the same options, "which convinces you most" and "which convinces you least". **Add Bulk Options** lets you paste the claims instead of typing them one by one, and **Create in bulk** builds the repeated blocks in one go. Keep option randomization on across all of them.
Prepare the set plan before you build: each claim appears the same number of times and with different partners each time. Field it in Exoid, export the dataset and compute the utility scores in your statistical tool.
If you have seven claims or fewer, a Ranking block gives you the order with far fewer screens.
The winning claim then needs diagnostics: a claims × criteria matrix — clarity, credibility, uniqueness, relevance — and a single choice on purchase intent. If the claim lives on a pack, use Image Choice with **Wait before advancing** set, so the stimulus stays on screen for the time you decide. Close with an AI-moderated open-ended question: "what does this claim make you think?" is what surfaces unintended readings and substantiation problems before regulatory finds them.
**What to look at in the results.** Counting the "most" minus the "least" per claim already gives you a readable ranking before any model. In the [Summary](/en/analyze/summary) turn on **Show statistics** on the scales and create comparison clusters between current buyers and everyone else: a claim that only convinces existing customers won't grow you. Then open [Talk with Data](/en/analyze/talk-with-data) in **Scientific** mode and ask whether the lead of the top claim holds: it runs real statistical tests and warns you when a segment's base is too small to conclude anything.
**Try it another way.** Give each respondent a single claim with the [media variable](#showing-different-stimuli-to-different-respondents) and ask for purchase intent and diagnostics on that one alone. MaxDiff tells you which claim wins the comparison; the monadic version tells you how well each claim does on its own, which is how the consumer will actually meet it. Running both on the same claims is the fastest way to learn how much your ranking owes to the competition inside the exercise.
### Advertising test
**The client's question.** "Does the campaign convince people to buy me? What image does it leave of my brand?"
**The methodologies.** [Monadic test](/en/methodologies#monadic-test), [Pre-roll test](/en/methodologies#pre-roll-test), [implicit association test](/en/methodologies#implicit-association-test-single).
**How you build it in Exoid.** Load the video as the block's media and set **Wait before advancing** to the length of the spot. Straight after, before any aided question, an open-ended on spontaneous recall: what do you remember of what you just saw. Then brand recall with a single choice ("which brand was it for?"), judgement with ratings on clarity, relevance and likeability, and post-exposure associations with a brands × attributes matrix you can compare against the same matrix among people who didn't see the spot.
**What to look at.** Spontaneous brand recall is the harshest metric and the most useful one: plenty of campaigns are liked and leave no name behind.
**Try it another way.** Add a control cell that never sees the spot, assigned with the [media variable](#showing-different-stimuli-to-different-respondents) or with a second link, and read every metric as the difference between exposed and unexposed. Absolute scores from an ad test are hard to interpret; differences are not.
### Creative pre-test
**The client's question.** "Which of these routes do we produce?"
**The methodologies.** [MaxDiff](/en/methodologies#maxdiff), [monadic test](/en/methodologies#monadic-test), [multimodal qualitative analysis](/en/methodologies#multimodal-qualitative-analysis).
**How you build it in Exoid.** Storyboards and key visuals are images, so Image Choice with **Supersize** on shows them properly. Two closed questions per route — is it clear, is it for a brand like this — then one open-ended with [AI moderation](/en/builder/ai-moderation) asking what the route promises. At this stage the verbatims matter more than the scores: you're choosing which idea to spend production money on, not certifying a finished film.
**What to look at.** Whether the message people take out is the message on the brief. A route that communicates something else brilliantly is still the wrong route.
**Try it another way.** Skip the scores altogether and run it as a qualitative study: three open-ended blocks with AI moderation and a File upload asking people to show something in their home that reminds them of the idea. With sixty respondents you get material a focus group would take three sessions to produce.
### Post-campaign
**The client's question.** "Did the campaign work? On whom?"
**The methodologies.** [Implicit association test, multiple](/en/methodologies#implicit-association-test-multiple), [Mental Availability](/en/methodologies#mental-availability), [Key Driver Analysis](/en/methodologies#key-driver-analysis).
**How you build it in Exoid.** Duplicate the pre-campaign study with **Duplicate** and keep question codes, copy and block order identical: the comparison only works if nothing else moved. Add a campaign recognition block — the key visual in an Image Choice, "have you seen this in the last month?" — and use it to split the sample into exposed and unexposed with a [comparison cluster](/en/analyze/summary).
Add a manual [variable](/en/logic/variables) called `wave` with the wave number, so it lands in the Variables column of Responses and in the export and you can stack the waves.
**What to look at.** Awareness, associations and buying situations covered, exposed versus unexposed, before versus after. Media plans get renewed on the second of those four numbers.
**Try it another way.** Run it continuously instead of before and after: a small weekly sample, always the same questions, and you see the curve rather than two points. Two points can't tell you when the effect started to decay.
### Promotion evaluation
**The client's question.** "Does the promotion bring new buyers, or is it a discount to people who'd have bought anyway?"
**The methodologies.** [Monadic test](/en/methodologies#monadic-test), [Price Sensitivity Meter](/en/methodologies#price-sensitivity-meter), [TURF](/en/methodologies#turf).
**How you build it in Exoid.** Show one promotional mechanic per respondent with the [media variable](#showing-different-stimuli-to-different-respondents) — three for two, thirty percent off, a gift with purchase, loyalty points — then a single choice on what they'd do and a Number Input on how many units they'd buy. A screener at the top separates current buyers from non-buyers, which is the split the whole analysis rests on.
**What to look at.** Declared incremental units among non-buyers. Among current buyers, extra units mostly mean stocking up, and stocking up is borrowed from next month.
**Try it another way.** Put the mechanics side by side in a Ranking and ask which is most attractive. You'll get a different winner, and comparing the two answers tells you whether the mechanic wins on appeal or on actual behaviour change.
## Brand
### Brand equity
**The client's question.** "How is my brand perceived, how strong and distinctive is it against competitors, and is it getting better or worse?" It's asked by Heads of Consumer Insights, brand managers and CMOs in FMCG; in retail the same job belongs to banner marketing and to whoever owns the private label. It comes in two shapes: **one shot**, a picture taken before a specific decision — a launch, a repositioning, entering a category — and **tracking**, continuous monitoring that catches a drift before it turns into lost share.
**How it gets solved today.** With research institutes, through continuous brand health trackers: orders of magnitude from a few tens of thousands to over a hundred thousand euros a year. Otherwise with the occasional one-off study on a panel, or not at all, using sales as a proxy for brand health. The agency cycle — brief, proposal, selection, design, fieldwork, analysis, readout — burns weeks before the first number lands, and vendor selection alone often eats several of them.
**The methodologies.** [Implicit association test](/en/methodologies#implicit-association-test-single), single and multiple, [Mental Availability](/en/methodologies#mental-availability), [Mental Advantage](/en/methodologies#mental-advantage).
**How you build it in Exoid.**
Start with a category screener: a Yes/No or single choice block on consumption, with a [navigation rule](/en/logic/conditions) sending anyone out of target straight to the ending screen. Give every block a readable [question code](/en/builder/block-options) such as `SQ_CATEGORY` or `Q_AWARE_SPONT` — filters, quotas and comparison clusters are all built on codes.
Then, in this order:
An open-ended question, before any list: "which brands come to mind in this category?". It has to come first, or the brands you show later contaminate recall.
A multiple choice on the brand list, with option randomization on so position doesn't skew the data.
One multiple choice per buying situation you want to cover — breakfast in a rush, guests for dinner, after the gym — always with the same brand list in random order. The number of situations your brand shows up in is the indicator.
A brands × attributes matrix, or one multiple choice per attribute ("which of these brands is the most..."). Say in the instructions that people should answer from the gut, without overthinking.
An open-ended question with [AI moderation](/en/builder/ai-moderation) on the association that matters most: the verbatim sits right next to the number.
For the sample, recruit from the [panel](/en/distribute/recruiting) with targeting profiles and set [quotas](/en/distribute/quotas) on the answers to balance brand users and non-users.
For tracking, duplicate the campaign at every wave with **Duplicate** from the campaign menu and keep question codes, copy and block order identical. Add a manual [variable](/en/logic/variables) called `wave` with the wave number: it lands in the Variables column of Responses and in the export, so you stack the waves in Excel or SPSS and read the trend there.
**What to look at in the results.** The gap between spontaneous and aided awareness: if the brand is recognized but not recalled, you have a memory problem, not a liking problem. How many buying situations you cover versus competitors. The attributes where you lead and where you trail, using [comparison clusters](/en/analyze/summary) between users and non-users. Wave on wave, the direction of travel matters more than the absolute value.
**Try it another way.** Run the association part one brand per respondent, assigned with the [media variable](#showing-different-stimuli-to-different-respondents), and each person judges a single brand without seeing the competitive set. The map gets cleaner and the answers less strategic, at the cost of a larger sample. Or go the other way and build the whole thing around Category Entry Points only: fewer attributes, more occasions, and a study that reads growth potential instead of image.
### Brand health tracking
**The client's question.** "Is anything moving, and did I notice in time?"
**The methodologies.** [Mental Availability](/en/methodologies#mental-availability), [Mental Advantage](/en/methodologies#mental-advantage), [NPS](/en/methodologies#net-promoter-score).
**How you build it in Exoid.** A short interview — eight to twelve blocks — that you can repeat without fatigue: spontaneous and aided awareness, three or four Category Entry Points, the attribute matrix, one purchase intent question. Duplicate it every wave, keep the codes frozen, and carry the wave number in a variable. Short and frequent beats long and rare: a tracker you dread fielding stops being fielded.
**What to look at.** Direction, not level. And the moment when spontaneous awareness moves before anything else does — it usually moves first.
**Try it another way.** Attach one rotating open-ended question with AI moderation to each wave, on a different topic every time. The tracker keeps its comparable core and gains a qualitative rolling window that costs you a single block.
### Mental availability
**The client's question.** "In how many buying occasions do I exist?"
**The methodologies.** [Mental Availability](/en/methodologies#mental-availability).
**How you build it in Exoid.** One multiple choice per Category Entry Point, always with the same brand list and randomization on. Phrase the occasion the way a person would live it — "you get home late and there's nothing ready" — not the way a marketing plan would write it. Eight to twelve occasions is a full picture; four is enough for a first read.
**What to look at.** The count of occasions where your brand is named, against the leader's count. In this framework, share of occasions is the growth number.
**Try it another way.** Turn it inside out: instead of asking which brands fit an occasion, show the brand and ask which occasions fit it, with a multiple choice on the occasion list. You measure the same link from the other end, and the two readings rarely agree — the gap is the interesting part.
### Competitive benchmark
**The client's question.** "How do I stand against the two or three brands that actually take my sales?"
**The methodologies.** [Implicit association test, multiple](/en/methodologies#implicit-association-test-multiple), [Mental Advantage](/en/methodologies#mental-advantage), [Key Driver Analysis](/en/methodologies#key-driver-analysis).
**How you build it in Exoid.** A brands × attributes matrix with your brand and the competitive set on the same rows, randomized, plus a single choice on last brand bought and one on the next likely purchase. **Transport options** is what makes this study efficient: the follow-up questions can carry over only the brands the respondent selected earlier, so nobody rates a brand they've never met.
**What to look at.** The attributes where the gap with the leader is widest and the attribute matters most. Those two conditions together, not either alone.
**Try it another way.** Ask the same people to describe each brand in one open-ended sentence, with AI moderation. Matrices tell you where you are on your own axes; open answers tell you which axes the consumer is using, and sometimes yours aren't on the list.
## People and market
### Segmentation
**The client's question.** "How many groups does my market — or my shopper base — split into, and which one do I focus on?" It's asked by CMOs, Heads of Strategy and Insights Directors; in retail by whoever runs loyalty and CRM and by the category manager, who builds shopper segments off loyalty data. It's a strategic, infrequent, high-impact choice: positioning, communication, media and assortment all follow from it.
**How it gets solved today.** With institutes, through large segmentation studies: orders of magnitude between eighty and two hundred thousand euros and three to six months of work. Few agencies can run one, so selection is long and proposals are expensive. Otherwise internal CRM data gets clustered with no attitudinal depth, or targeting stays purely demographic. The known risk is that the segmentation, once delivered, stays on paper.
**The methodologies.** [Segmentation](/en/methodologies#segmentation), need-based approach.
**How you build it in Exoid.**
The attitudinal battery is the core of the study: twenty to forty statements about needs, occasions and attitudes towards the category, spread over one or more matrices sharing the same five- or seven-point agreement scale. Randomize the row order. If you need more granularity than a point scale, the Slider measures intensity on a continuum.
Around the battery:
* **Behaviors and usage occasions** with multiple choice blocks, so you have something to profile the segments with afterwards.
* **Frequency and spend** with Number Input blocks, so the data stays numeric and usable in the clustering.
* **Demographics at the end**, not at the start: in a need-based segmentation they describe the segments, they don't define them. Only screeners belong at the top.
The sample has to be large and balanced: recruit from the [panel](/en/distribute/recruiting) and use quotas, because clusters don't hold on small bases. Turn on [Advanced statistics mode](/en/advanced/advanced-statistics) before going live, so you get the structured dataset and the SPSS export.
When collection closes, export from Responses to SPSS (`.sav`) or Excel and run the clustering in your statistical tool: number of segments, membership rules, profiling.
Then the segments come back into Exoid in two ways. In the results you create one **comparison cluster** per segment, translating the rule into conditions on the battery questions, and from there you compare every question in the study segment by segment. In the next study, a [Script](/en/advanced/scripts) on the first block can compute the segment from the answers just given and store it in a variable such as `vars.segment`, which appears in the Variables column of Responses and in the export. That's your typing tool, inside the platform.
**What to look at in the results.** The size of each segment and how reachable it is. How genuinely distinct the segments are on the questions that matter, not only on the battery that produced them. And above all: who buys you today, which segment they sit in, and whether that segment is growing or shrinking.
**Try it another way.** Segment on behaviour instead of attitudes: occasions, frequency, spend, repertoire of brands. It's a shorter interview, the segments are addressable in media the day they exist, and you lose the motivational story. Or start with a MaxDiff on needs and cluster on the resulting scores — a middle route that keeps the battery short and the discrimination high.
### Targeting and buyer profile
**The client's question.** "Who actually buys me, and are they the people I've been talking to?"
**The methodologies.** [Segmentation](/en/methodologies#segmentation), [Key Driver Analysis](/en/methodologies#key-driver-analysis).
**How you build it in Exoid.** A short interview on the category: last brand bought, frequency, occasions, channel, spend, plus the demographics at the end. Recruit a representative sample from the [panel](/en/distribute/recruiting), then build one comparison cluster for your buyers, one for the leader's, one for the category as a whole. The whole study is one screener and eight questions, and it's the cheapest high-value thing you can field.
**What to look at.** Where your buyers differ from the category, not where they resemble it. Similarity is the base rate; difference is the target.
**Try it another way.** Ask the same questions of lapsed buyers only, recruited with a screener on past purchase. The profile of who left is more actionable than the profile of who stayed.
### Habits and consumption occasions
**The client's question.** "When, where and with whom is my product actually consumed?"
**The methodologies.** [Mental Availability](/en/methodologies#mental-availability), [multimodal qualitative analysis](/en/methodologies#multimodal-qualitative-analysis), [Segmentation](/en/methodologies#segmentation).
**How you build it in Exoid.** Anchor it to the last real occasion, not to habits in general: Date and time to place it, single choices on moment, place and company, a multiple choice on what else was consumed with it, a Number Input on quantity. Then a File upload for a photo of the setting and an AI-moderated open-ended on why that product and not another. Recall of a specific event is far more accurate than a self-report about "usually".
**What to look at.** Which occasions carry the most volume and which are growing. Occasion is more often the unit of growth than the buyer is.
**Try it another way.** Field the same interview once a week for a month, short link over email, and you have a light diary instead of a snapshot. Repetition on the same people turns declared habit into observed frequency.
### Category exploration
**The client's question.** "What's going on in the category?"
**The methodologies.** [Multimodal qualitative analysis](/en/methodologies#multimodal-qualitative-analysis).
**How you build it in Exoid.** This is the study with more open questions than closed ones. A few closed questions to place who's answering and what they consume, then open-ended questions with [AI moderation](/en/builder/ai-moderation) that follows up while the person is still answering, instead of leaving you a single line. File upload brings you the photo of the cupboard or the shelf, and the Website block asks people for the page they use, which you can then look at yourself.
In the results, [Talk with Data](/en/analyze/talk-with-data) in **Conversational** mode is the right way to explore: every answer carries the codes of the questions it cites and suggests where to dig. When you find something, send it over with **Add to report** and assemble the readout in the [AI Report](/en/analyze/ai-report).
**What to look at.** Hypotheses, not estimates. The sample can be small here, because the useful output is the right question to put in the next quantitative study.
**Try it another way.** Turn the qualitative findings straight into a MaxDiff: take the twenty things people said and make the next study rank them. Exploration that ends in a ranked list gets acted on; exploration that ends in a deck usually doesn't.
### New market entry
**The client's question.** "Does this product make sense in a country, a channel or a category where we aren't yet?"
**The methodologies.** [Monadic test](/en/methodologies#monadic-test), [Price Sensitivity Meter](/en/methodologies#price-sensitivity-meter), [multimodal qualitative analysis](/en/methodologies#multimodal-qualitative-analysis).
**How you build it in Exoid.** One study, three parts: the category as it stands there — brands used, occasions, price paid — then the concept with a Message block and a purchase intent question, then the four price questions. Recruit each market from the [panel](/en/distribute/recruiting) as a separate target group and keep the interview identical so the markets are comparable. Add a manual variable with the market code, so the export carries it.
**What to look at.** The distance between the category as you know it at home and the category as described there. Concept scores travel badly across countries; category structure travels well.
**Try it another way.** Test the local competitors' packs and claims instead of your own concept. You learn the codes of the market before spending on an entry proposition, and it costs a single Image Choice study.
## Purchase and relationship
### Shopping journey
**The client's question.** "How and why does my consumer get to the purchase — or not — in store and online, and where do I lose them?" It's asked by shopper marketing managers, trade marketing, category managers and e-commerce managers; in retail by customer experience, store management and loyalty. The trigger is always concrete: low online conversion, a launch that isn't moving on shelf, cannibalization between SKUs, or a consumption peak to capture such as Christmas or Black Friday. Sales and loyalty data tell you the what, not the why.
**How it gets solved today.** With institutes, through shopper and path-to-purchase studies, ethnography plus panel: orders of magnitude between fifty and a hundred and fifty thousand euros and long timelines. Few agencies cover both physical and online well, and stitching different sources together at analysis stage is laborious. Otherwise it's some in-store observation, or it gets inferred from sales data. It's the job least served by pure quantitative platforms, and the limitation is well known: by the time the study lands, the market has moved.
**The methodologies.** [Multimodal qualitative analysis](/en/methodologies#multimodal-qualitative-analysis) integrated with quantitative work: [implicit association](/en/methodologies#implicit-association-test-single) on the first moment at shelf, [Key Driver Analysis](/en/methodologies#key-driver-analysis) on choice drivers, [MaxDiff](/en/methodologies#maxdiff) on triggers and barriers, [Segmentation](/en/methodologies#segmentation) on shopping missions.
**How you build it in Exoid.** One study in phases, with [navigation rules](/en/logic/conditions) sending each mission and each channel down its own branch and bringing them back to the common trunk.
Single choice on the last purchase: planned stock-up, urgent need, special occasion, trial. It's the block that governs everything else — give it the code `MISSION`.
Single choice between store, online and mixed, with navigation rules to the physical or the digital branch. Each branch has its own questions and rejoins the common trunk with **Always go to**.
Date and time to place the purchase in time, a multiple choice on the touchpoints encountered, a Ranking on which one weighed most in the decision.
Image Choice with a photo of the shelf or the product page, **Wait before advancing** set to a few seconds to reproduce real exposure time, then the question on what you noticed first. With the [media variable](#showing-different-stimuli-to-different-respondents) you can show each respondent the shelf of the retailer they actually shop at.
Two lists in separate blocks. If you need them ranked, use the most/least important exercise described in the [claim test](#claim-test).
File upload for a photo of the receipt or the shelf, and an open-ended question with [AI moderation](/en/builder/ai-moderation) about the moment they gave up. This is the qualitative part, and in Exoid it lives inside the same quantitative study.
A matrix on product attributes and an outcome measure — satisfaction or repurchase intent — in the same study. Those are the two ingredients of a Key Driver Analysis: you collect both here, export the dataset and run the regression in your own tool.
**What to look at in the results.** Filter the [Summary](/en/analyze/summary) by mission: filters are built on questions, picking a code, an operator and a value. Create a store versus online comparison cluster — barriers change completely between the two. In the [Responses](/en/analyze/responses) table look at where in the flow people stop, since the detail of a single response tells you how many questions they completed. And read the barrier verbatims before the numbers: they're the part sales data will never give you.
**Try it another way.** Cut the journey into three short studies fielded in the same week — one on the trigger, one on the shelf, one on the moment after purchase — each on its own sample. You lose the individual path and you gain honest answers, because nobody is thirty questions deep by the time you ask the hard one.
### Digital experience and product page
**The client's question.** "Why doesn't my product page convert?"
**The methodologies.** [Monadic test](/en/methodologies#monadic-test), [Key Driver Analysis](/en/methodologies#key-driver-analysis), [multimodal qualitative analysis](/en/methodologies#multimodal-qualitative-analysis).
**How you build it in Exoid.** Show a screenshot of the page as the block's media, **Wait before advancing** set to the time a real visit lasts, then: what is this product, what does it cost, would you buy it. Three questions that reveal whether the page communicates at all. Follow with a matrix on clarity, trust and completeness of information, and an open-ended with AI moderation on what's missing.
**What to look at.** The share of people who can't say what the product does after looking at the page. It's usually higher than anyone in the team expects.
**Try it another way.** Give each respondent a different version of the page with the [media variable](#showing-different-stimuli-to-different-respondents) and compare comprehension across cells. Same study, one extra variable, and the diagnostic becomes a test.
### Satisfaction
**The client's question.** "Are the people who bought happy, and with what exactly?"
**The methodologies.** [Key Driver Analysis](/en/methodologies#key-driver-analysis), [Penalty Reward and Kano factors](/en/methodologies#penalty-reward-and-kano-factors).
**How you build it in Exoid.** A matrix on the experience drivers, all on the same scale, plus one overall satisfaction question: those two are the input to a Key Driver Analysis, which you estimate on the export. Distribute it with the [link](/en/distribute/share-link) over email to your own base and pass the customer ID as a [URL variable](/en/logic/variables), so answers rejoin your CRM. With a [webhook](/en/webhooks/introduction) each response reaches your endpoint while collection is still running.
**What to look at.** Not the drivers that score worst, but the ones that score badly and weigh heavily. High weight and high score is what you protect; high weight and low score is what you fix.
**Try it another way.** Ask about the last specific interaction rather than the relationship as a whole. Overall satisfaction is stable and uninformative; the last contact is volatile and tells you what actually happened.
### Loyalty and NPS
**The client's question.** "Do customers stay and come back? Do they spend more with me than elsewhere?"
**The methodologies.** [NPS](/en/methodologies#net-promoter-score), [Key Driver Analysis](/en/methodologies#key-driver-analysis).
**How you build it in Exoid.** A Rating block from 0 to 10 with **Display type** set to `Numeric` for the recommendation question, and right under it an open-ended with [AI moderation](/en/builder/ai-moderation) shown only below a threshold, using a [visibility condition](/en/logic/conditions) with the "less than" operator: ask why to the people who have a why.
Set the promoter, passive and detractor split with a [Script](/en/advanced/scripts) writing `vars.nps_group` from the rating answer, or reconstruct it with three comparison clusters on the rating. The net score — promoters minus detractors — comes out of the export in one formula.
**What to look at.** Not the score, the movement: which driver separates promoters from passives. NPS tells you how you're doing, Key Driver Analysis tells you what to act on.
**Try it another way.** Replace the recommendation question with a share-of-wallet question: out of your last ten purchases in the category, how many were us. It's harder to answer and far harder to game, and it correlates with revenue rather than with mood.
### Renewal and churn
**The client's question.** "Who's about to leave, and why?"
**The methodologies.** [Key Driver Analysis](/en/methodologies#key-driver-analysis), [MaxDiff](/en/methodologies#maxdiff), [multimodal qualitative analysis](/en/methodologies#multimodal-qualitative-analysis).
**How you build it in Exoid.** Two samples in the same study, split by a screener: current customers and people who left in the last year. Same battery of drivers for both, one single choice on renewal intent for the first group and one on the reason for leaving for the second, then a most/least important exercise on the reasons so they come out ranked rather than all equally important. An AI-moderated open-ended on the moment they decided closes it.
Distribute over the [link](/en/distribute/share-link) with the customer ID as a URL variable, so the answers join your churn data and you can check declared intent against what actually happened later.
**What to look at.** The reasons that lapsed customers rank first and current customers rank low. That gap is the warning system.
**Try it another way.** Ask only the people who stayed why they nearly left. Retention interviews get better recall than exit interviews, because the person is still willing to talk to you.
## Showing different stimuli to different respondents
Plenty of the recipes above need one respondent to see one version: a monadic concept test, a pack per person, a control cell with no ad, a shelf that matches the retailer the person actually shops at. You do all of that inside a single study, with the **Media** field.
Every block's **Media (optional)** panel has a **Media source** with two values:
* **Upload file** — the image, video or audio is fixed and everyone sees the same one.
* **Use variable** — the media comes from a [variable](/en/logic/variables), so it changes from respondent to respondent.
The usual way to feed that variable is the URL. You prepare one link per version, each carrying the parameter that identifies its stimulus, and distribute them across the sample or across panel batches. Everything else in the interview is identical: same blocks, same question codes, same order. The variable lands in the Variables column of [Responses](/en/analyze/responses) and in the export, so at analysis time you always know who saw what, and every cell is a [comparison cluster](/en/analyze/summary) inside the same campaign.
Since the cells live in one campaign, you compare them directly in the [Summary](/en/analyze/summary) without stacking exports, and the quotas apply to the whole study, so the cells stay balanced on the profile you care about.
A [Script](/en/advanced/scripts) on the first block can write the same variable, which is how you rotate versions without preparing separate links.
The alternative design is worth knowing too: give everyone all the versions and ask for an order, with Ranking, Image Choice or the most/least important exercise. Direct comparison picks up smaller differences and takes fewer respondents; one version per person is closer to how people meet a stimulus in life. Two answers to the same question — and running both is often the cheapest way to understand your own result.
## Next steps
All fifteen methodologies one by one: what they measure and how you set them up in Exoid.
The full map of capabilities, organized by goal.
Every block available in the Builder and when to use each one.
The guided end-to-end path, from creation to results.
## Related topics
- [Research methodologies](/en/methodologies.md)
- [What you can do with Exoid](/en/capabilities.md)
- [Exoid documentation](/en/index.md)