Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

E-commerce Marketing › E-commerce Conversion

Product Finder Quizzes for Online Stores

How online stores build product finder quizzes that help shoppers choose: when a finder beats filters, the questions, honest results, consent and measurement.
Guideintermediate11 min readUpdated

What a product finder is for, and when filters or a guide do better

Some shoppers know the product they want; some know which attributes matter. Others know their situation (a small kitchen, a bad back, a first bicycle) but not which product suits it. A finder is for them: it asks about the situation and translates it into attributes.

The store already serves the others. The guide to category and listing pages builds filters around the attributes that decide the purchase. The product page optimisation guide sends shoppers who are still comparing types to a guide rather than to a product page, and the buying guides guide writes those guides. The guide to selling gifts online organises gift guides by recipient and budget. And the guide to answering pre-purchase questions answers a fit question by asking about the situation first, which is what a finder does for shoppers who never ask.

Where the shopper isWhat usually serves themWhy
Knows the productSite search and the product pageNothing left to choose
Knows the deciding attributesFilters on the listing pageThey can narrow the range themselves
Wants to understand the typesA buying guideThey want the reasoning, not only the answer
Knows the situation, not the attributesA product finderIt turns the situation into attributes
Shopping for someone elseA gift guide, or a finder that asks about the recipientThe starting point is a person
Has a situation the finder cannot coverA person answering the questionSome questions need judgement

As general practice, a finder is worth building when products differ on attributes shoppers cannot judge and the store has evidence of confusion: long comparisons between similar product pages, repeated fit questions, returns coded as the wrong choice, and site searches that describe a situation rather than a product. It is rarely worth it for a small range, for products chosen on taste, or where filters already work, and it does not fix weak product pages, which come first because every result ends on one. The e-commerce conversion section covers those pages.

Questions taken from what decides the purchase

The evidence for the questions is already in the business's records. The tutorial on turning customer language into page copy collects customers' words from reviews, support tickets, chat transcripts, returns reasons and site search terms, and codes them into jobs, objections and phrases. The site search analysis tutorial reads the search log, and the pre-purchase record shows which doubts come up most. Together they show which attribute decides fit and how shoppers describe their situation.

The jobs-to-be-done framework turns that evidence into questions. Ask about the job the product is bought for and the situation it will be used in, not the specification: how the shopper sleeps rather than which loft height they want, what they will carry rather than how many litres the bag holds. Shoppers can answer the first kind; only an expert can answer the second.

As general practice, each question earns its place:

  • It changes the result. If every answer leads to the same products, the question is clutter.
  • The shopper can answer it. Use their words, and offer a "not sure" answer mapped to a safe default.
  • It asks nothing the mapping does not use. A finder is not the place to collect personal details.

The answers are first-party data, research into what customers want; keep them apart from marketing permission, a separate choice covered below.

The finder is also a form. The forms guide summarises Nielsen Norman Group's recommendations: keep forms short, group related fields, use visible labels and a logical order. For a finder that means easy questions first, options shown at once, honest progress and a way back without losing answers.

Mapping answers to products from the catalogue's own attributes

The mapping is the finder's substance: rules that turn each combination of answers into a short list of products. As general practice, write it as a table people who know the range can read and check, each answer setting attribute values and each combination leading to the products that carry them.

Build it on the catalogue's own attributes, the ones that drive the filters and product cards. If the deciding attribute is missing, such as which sleeping position a pillow suits, add it to the product data rather than to the finder alone, so filters, cards and finder draw on one source and cannot disagree.

Part of the mappingWhat it recordsWho checks it
Questions and answersThe wording shown to shoppers, including "not sure"Customer service
Attribute rulesWhich attribute values each answer sets or rules outThe product owner
Priority between answersWhich answer wins when two point different waysThe product owner
FallbacksWhat a combination with no exact match showsMerchandising
Stock rulesWhat happens when the best match is unavailableMerchandising, with operations
Review triggerThe range changes that send the mapping for reviewThe finder's owner

Run through every combination before launch and fix any that return nothing, giving them the nearest match with an honest note, as the listing pages guide does for an empty result. When the best match is out of stock, show the next best that can be bought and offer an alert when it is back, built as the back-in-stock and price-drop alerts tutorial describes. Then keep the mapping in step with the range, reviewing it whenever products are added or withdrawn.

The results: a short list, the reason for each and a way out

As general practice the result is a short list, not a grid, because the point was to narrow the range: a best match and an alternative, such as a step up or a different material, is usually enough.

Each result carries the reason it was chosen, in the shopper's own answers: suits side sleepers on a firm mattress. The gifts guide makes the same point about gift guides: a short reason for each pick persuades more than a wall of pictures. Each result shows price and availability and leads to the basket or the product page.

The buying guides guide warns that shoppers take a guide that always picks the priciest product for advertising, and a finder that does the same is read the same way. Let the mapping decide the order, not the margin, and say plainly when nothing fits. Then give every shopper a way out: change an answer without starting again, see the full range with the filters set to match the answers, read the buying guide for the reasoning, or ask a person about a situation the finder cannot cover.

As general practice, show the results on screen without asking for an email address first; holding the answer back turns help into a toll.

Where the store offers to email the results, the ICO's direct marketing guidance applies as the list building and consent guide and the back-in-stock tutorial state it: a request for one thing is a request for that thing, and ongoing marketing is a separate, unticked choice beside it. So the form asks where to send the results and offers the newsletter as a box the shopper ticks only if they want it.

Keep a consent record of what each shopper agreed to: which channel it covers, the words shown to them, when they agreed and where. Whether the results email counts as marketing is for whoever owns compliance; for its alerts the back-in-stock tutorial takes the safer default of handling them as marketing with an unsubscribe. Using the answers for anything the shopper did not ask for is also a compliance question. Nothing here is legal advice.

How a store describes its finder is a claim. The guide to AI claims and consumer protection records, from the FTC's business guidance, that a claim must be capable of substantiation before it is made, and its own example softens "AI-powered skin analysis" into "a quiz that recommends products from your answers", since the plainer description is the one the evidence supports.

As general practice, describe a finder in those plain terms: it matches answers to product attributes. "Personalised", "expert analysis" or "perfect match" each make a claim, and each needs evidence on file before it is used.

The reasons beside each result are product claims too, and the pre-purchase questions guide's rule for agents applies: never add a promise the product page does not make. The return rate diagnosis traces one cause of rising returns to a claim the product does not keep, and a finder promising a fit the product cannot deliver produces exactly that.

Measuring whether the finder sells

A finder has its own funnel, from starts to purchases and returns. The forms guide notes that each step of a multi-step form is an event, so drop-off between questions shows which one costs the most.

Google's developer guide to measuring e-commerce in Analytics describes item lists that carry each item's position, and an event for choosing an item from a list; the e-commerce tracking tutorial shows how to set them up. Send the results as an item list of their own, named as a buyer would recognise it, and read it as the product and merchandising analysis guide reads any placement: views, selection rate, basket adds and purchases credited to the list, and which positions are chosen. That guide's dependable way to credit purchases to a list is to keep the list's identifier and name attached to each item when it is added to the basket and when it is bought.

The comparison that matters is with the store without the finder. Shoppers who choose to use one differ from those who do not, so comparing users with everyone else says little. The A/B testing guide sets out the alternative: split visitors at random between pages that offer the finder and pages that do not, fix the metric and guardrails before launch, and run to the planned sample. The conversion rate calculator turns each arm's counts into comparable rates (CVR), and the add-to-cart rate reads earlier in the journey.

Judge it on returns as well as orders. Compare each arm's returns once their return windows have closed, as the return rate diagnosis advises, and read coded reasons by answer combination: wrong-fit returns concentrated on one combination point to a rule.

The stated and revealed preference framework adds the last check: answers are stated preferences, and what shoppers buy and keep is revealed. Where shoppers who answer one way keep a product the finder did not recommend, the purchases are the better guide to fixing the rule.

QuestionMeasureWhat a weak result points to
Do shoppers start it?Starts over visitors to the pages that offer itThe invitation or its placement
Where do they give up?Drop-off by questionA question shoppers cannot or will not answer
Do the results earn a click?Selections over result viewsResults that do not match, or reasons that do not convince
Does it sell more than no finder?Orders per visitor in a randomised testA finder that slows shoppers who already knew
Do buyers keep what it recommends?Returns and reasons by arm, windows closedMapping rules that mismatch products and situations

A pillow maker builds a sleep-position finder (illustrative)

Illustrative example — not a real result

Somnurbe Pillows builds a sleep-position finder

Somnurbe Pillows, an invented online pillow maker, sells 12 pillows that differ by height, firmness and filling. Shoppers compare product pages for a long time, and over six months 38% of returns are coded as the wrong height. Reviews and returns notes show that shoppers describe how they sleep, not the loft they need.

It builds a four-question finder on sleeping position, mattress firmness, shoulder width and a filling preference, each with a "not sure" answer. It adds a sleeping-position attribute to the product data so the filters and cards use it too, and maps the answers to the catalogue's height and firmness attributes. The result shows two pillows, each with a reason drawn from the answers, and a link to the full range filtered by height. Beside "Email me my results" sits a separate, unticked newsletter box, and the finder is described as "a quiz that matches your answers to our pillows' height and firmness".

In a six-week test, half the visitors to the pillow category are offered the finder: 4,900 of the 21,000 visitors offered it start it and 3,700 finish, with most drop-off at the shoulder width question, which is reworded with pictures. Orders per visitor are 2.6% against 2.3% for visitors not offered it. Once return windows close, 7% of orders from the finder arm come back against 11% from the other arm, with the largest fall in wrong-height returns. Front sleepers often keep a medium pillow the finder did not recommend, so the team changes that rule.

Common mistakes

  • Building a finder to fix weak product pages. Every result ends on a product page; fix it first.
  • Asking about the specification. Ask about the shopper's situation and the job the product is for.
  • Questions that change nothing. If every answer leads to the same products, remove the question.
  • A mapping held only in the finder. Put deciding attributes in the product data so everything agrees.
  • Results with no reason and no way out. Show why each product was chosen, and link to the full range.
  • Recommending by margin. A finder that always picks the dearest product is read as an advert.
  • Treating a results request as a subscription. Keep the newsletter a separate box, left unticked.
  • Claims the rules cannot support. "Personalised" and "expert analysis" need evidence before they run.
  • Comparing finder users with everyone else. Test against the store without it, returns included.

Frequently asked questions

How many questions should a product finder ask?

As few as decide the choice. Write the mapping from answers to products first, then keep only the questions whose answers change which products are shown; a question that leads to the same results whatever the answer is clutter. Read where shoppers give up, question by question, and reword or remove the question that loses the most. There is no right number in general, only the smallest set that separates the store's own range.

Can a store add everyone who uses its finder to the newsletter?

No. As general direction from the ICO's direct marketing guidance, as the site's list building and back-in-stock guides state it, a request for one thing is a request for that thing: a shopper who asks for their results by email has asked for the results. Ongoing marketing needs its own, unticked choice beside the request, recorded with the wording the shopper saw. Settle with whoever owns compliance whether the results email itself counts as marketing. Nothing here is legal advice.

Does a finder replace filters and buying guides?

No. Each serves a different shopper. Filters serve shoppers who know which attributes matter and want to narrow the range themselves; buying guides serve shoppers researching the differences between types; a finder serves shoppers who know their situation but not which product or attribute fits it. A finder works best linked to both: its results lead to the full range with the matching filters, and to the guide for anyone who wants the reasoning.

Terms used on this page

Tools to try

Frameworks

Also related

Sources

  1. Direct marketing guidance — Information Commissioner's Office (UK)
  2. Measure ecommerce (GA4 developer guide) — Google for Developers
  3. FTC business guidance — Federal Trade Commission

Related