Quiz funnel

What a quiz is trading

A personal answer for an email address. That is a better trade than any popup can offer, and it is the whole reason the format exists. What most brands then do is throw the answer away: every taker lands in the same generic welcome flow, which spends the quiz's entire advantage on the capture and none of it on the relationship.

The one rule: spend the answer. The taker just told you which email to send. If the flows do not use that, you have bought a popup at quiz prices and added three weeks of build time.

How to think about it

Sketch the machine, then write the questions, then the buckets, then the flows. Writing questions first is the common order and it produces a quiz whose scoring gets reverse-engineered afterwards, which is how every path ends up leading to the bestseller.

• A quiz earns its funnel when the product genuinely benefits from diagnosis. Shades, routines, sizes, stacks, plans. If every customer should buy the same thing, the quiz is theatre, and takers can tell within three screens.
• Every question changes the recommendation, or it dies. Run the test on each one: which bucket does each option move, and by how much? A question that moves nothing is stolen time, however much you would like to know the answer.
• One self-declared priority question is the exception worth keeping. "What matters most to you?" rarely changes the scoring, but it is the single most useful line for the flow copy, and the taker knows exactly why you asked.
• Sequence for completion. Easy and visual first, sensitive last. Budget and contact details never open the show.
• The result page diagnoses before it recommends. Two sentences reflecting their own answers back is the paragraph that makes quizzes convert. It works because you did read their answers.
• Say when you are not the answer. One honest result page, for the bucket genuinely better served elsewhere, buys credibility for the other four. It costs you a sale you were going to lose in the returns queue.

Failure modes worth naming: an email gate decided by default rather than on purpose; a scoring model that routes every path to the highest-margin item, which is detectable by the third screenshot on a forum; sensitive answers reflected back in a subject line a shoulder-surfer can read; and buckets so blunt that seventy percent of takers land in one, at which point every per-segment number on the board is fiction.

Numbers to hold it against

Quiz numbers look unusually good next to other capture mechanics, which is why they need honest framing rather than celebration.

• A five to eight question quiz commonly completes at 50 to 80 percent of starts. Every extra screen costs completions. If takers are leaving at question six, that is a question problem rather than a traffic problem, and the drop-off table names the culprit.
• Email capture from completers commonly runs 50 to 70 percent when the result is gated and the result feels worth it. A visible skip lowers capture and raises the quality of what you keep, which is a real trade rather than a mistake.
• Start-to-email conversion for a well-built quiz commonly lands somewhere in the 15 to 40 percent range of page visitors, against low single digits for a standard popup. That gap is the entire business case for the format.
• No bucket should hold much more than half the takers. Seventy percent in one means the score mapping is too blunt to segment anything; sharpen the options before trusting a single per-segment number.
• Quiz-sourced subscribers commonly convert better than popup-sourced ones over the first 30 to 90 days, precisely because the first email can be about them. Revenue per subscriber from the segment flows against the generic welcome cohort is the number that justifies this board, and if it is flat after a full cycle, the flows have stopped using the answers.
• Three to five buckets. Two is a coin toss dressed up as a diagnosis; past five you are writing flows nobody has time to keep current.

What moves these: category above all (diagnosis-heavy categories such as skincare, supplements, sizing and equipment fit outperform commodity ones by a wide margin), traffic source, whether an incentive rides alongside the result, and how visual the options are.

What has to be true before a card asks for approval

Question cards. Every option mapped to a bucket score, with the amount. A stated reason the question earns its place, naming the recommendation it changes. The wording in the taker's language, not the brand's. Images on the options wherever the choice is visual, because tapping a picture beats reading four radio buttons.

Segment cards. The score range that lands here, the expected share, and the result page itself: diagnosis in two sentences that quote their answers back, then the recommendation with its why, then the honest alternative where we are not the right answer. The one objection this type has, named, because it is what email two answers.

Emails. One per segment, not one for all. Email one is the receipt for the trade and has to be worth the address on its own: the type, the diagnosis, the saved result, the recommendation restated. Email two answers this type's specific hesitation with proof from customers like them. If two segments' emails read identically, the segmentation is being wasted and it is worth saying so rather than shipping the duplication.

Consent and privacy. The line under the email field, the privacy terms, and the flows that actually send all describe the same reality. Answers are used for the relevance promised at collection and nothing else. Sensitive categories get reflected back gently in body copy and never in a subject line.

The seed cards are the first instance, not the ceiling

Seven cards show the shape. The count on this board is not a matter of ambition: it is arithmetic, and it is the reason quizzes get abandoned half-built.

Design holds the concept card and the pinned rules. Questions holds one card per question, so five to eight, each with its own option-to-score mapping. Segments holds one per bucket, three to five, each carrying a full result page. Flows is where the multiplication bites: at least two emails per segment, so a four-bucket quiz is eight email cards before you have written a single reminder. That number is the board, and a quiz with two emails and four buckets is a quiz that decided to be a popup on the last afternoon. Measure holds one funnel card, plus the per-question drop-off deltas written back onto the question cards themselves.

The layout is one arrangement of many

The columns run by production step: design, questions, segments, flows, launch, measure. That is the right shape while the quiz is being built, because the machine has to exist before the questions do. Once it is live, most teams find a column per segment reads better: each bucket owns its result page, its two emails and its own share of the funnel, and the whole board becomes a set of parallel small programmes. Rearrange it. Spend the answer, and let every question earn its place, both survive any layout.

When this is the wrong playbook

If every customer should buy the same thing, do not build a quiz. It will capture addresses and quietly teach people that your personalisation is decorative. If you cannot build per-segment flows, either because the ESP will not support it or because nobody will maintain them, run a good popup instead and keep the three weeks. If the category is sensitive enough that a diagnosis implies advice you are not qualified to give, the honest result page is mostly a referral, and that is a content problem rather than a funnel one. And if the traffic is too thin to fill five buckets, the per-segment numbers will never be readable: start with three, or start with traffic.