The Best Preference Quiz Is the One Shoppers Finish

Preference quizzes die at question nine of twenty five, taking the data and the shopper with them. Designing for completion: five questions, visible progress, and a payoff before the email field.

A preference quiz is a trade: the shopper pays attention, the store pays with a visibly better first session. Most quizzes break the trade in both directions. They ask twenty questions whose effect the shopper never sees, then gate the result behind an account creation form, so the shopper pays everything and receives nothing. Completion data from quiz tools is brutal and consistent on one point: every question past the first handful costs finishers, and the questions that cost most are the ones whose purpose the shopper cannot guess, household income being the Hall of Fame example. The quizzes that finish share a shape. Five to seven questions, each one changing something on screen as it is answered, a progress indicator that moves honestly, and a results page that lands before anyone asks for an email. The email ask then converts better than the gated version ever did, because the shopper has seen the goods.

Ask only what changes a recommendation today

Audit every question against a single test: name the recommendation that changes based on the answer. Size system changes everything downstream, keep it. Category intent changes the first rail, keep it. A brand refusal changes the filter set, keep it. Favorite color, in a catalog where color affinity emerges from two clicks of behavior, changes nothing the engine cannot learn in a minute, cut it or move it to an optional second pass. The audit usually halves the quiz, and the half that survives is the half whose answers the shopper can watch working, which is the completion mechanism, not a nice extra.

Order matters as much as count. Open with the easiest question with the most visible payoff, usually the category or the use case, and put anything sensitive, budget, sizes in systems shoppers find personal, after the shopper has seen two questions pay off. Trust in a quiz is sequential: each answered question is a small bet that the next one is also worth it. Spend the early questions proving it.

Related reading on this site: PrefEngine and Preference Sunset Rules: When to Delete Old Preferences cover the neighbouring problems in more depth.

Progress, exits, and the partial save

Show progress as questions remaining, not a percentage bar that sits at 12 percent for three questions and teaches the shopper the bar lies. And build the exit ramp: a shopper who bails at question four has still given you three answers, so save partials and run the session on them. The all or nothing quiz, where abandoning discards the answers, punishes the store for the shopper's honesty. Partial data from a finisher rate you can measure beats complete data from the minority who endure the full form.

Mobile is where quizzes go to be abandoned, so design the tap targets and the one question per screen layout for a phone on a train, not for the desktop your team built it on. If a question needs a paragraph of explanation to be answerable on mobile, the question is wrong, not the shopper.

Close the loop in the first session

The payoff page is the product. It should show a small set, visibly built from the answers, this rail exists because you chose trail over road, with an edit control beside each inference so the shopper can correct the record while they are looking at it. Corrections made on the payoff page are the highest quality preference data a store collects all year, because the shopper is reacting to a concrete recommendation rather than imagining their taste in the abstract. Log them separately from the raw quiz answers and weight them above everything else the quiz produced.

Then stop asking. The store that follows a completed quiz with a second survey request in the welcome email has misunderstood what it bought. It bought permission to watch behavior for a while. Go watch.

Frequently asked questions

How many questions should a preference quiz have?

Five to seven is the range most stores can sustain without heavy abandonment. Every question should change a recommendation the shopper can see; questions whose answers only feed a distant model are the ones to cut or defer.

Should quiz results be gated behind an email signup?

No. Showing the personalized result first and asking for the email afterward converts better and builds trust, because the shopper can judge the value before paying for it. Gating the payoff behind a form is the most common completion killer.

What should happen to answers from abandoned quizzes?

Save them and use them. Partial answers from a shopper who stopped at question four are still valid preference data, and building the session on them respects the effort already given.

What makes a good first quiz question?

An easy question with an immediately visible effect, usually the category or use case the shopper came for. Early questions earn trust for later, more personal ones, so the opener should pay off on screen at once.