A common intent shaping problem
Your setup questionnaire loses users
The short answer: ask less now, more later. Here is how to recognise this problem in your funnel, why it happens, and how the best consumer apps design it away.
How to recognise it
- Drop-off inside setup, before the product is ever seen
- Questions whose answers nothing uses for weeks
- A questionnaire shaped by stakeholder wishlist
Why do setup questionnaires cause drop-off?
A long setup questionnaire charges the user upfront for a product they have not tasted: ten questions of effort, paid on faith.
Each additional question raises the bill, and the user's balance, patience backed by zero delivered value, was small to begin with. Drop-off is not impatience; it is a rational refusal of a bad trade.
The questionnaire is usually long for internal reasons: teams pooling their data wishes into the one flow every user must pass. The user pays for that org chart.
How do you collect data without losing users? Progressively, each question in its moment
Split the questionnaire by urgency: what genuinely changes the first experience gets asked now, one or two questions at most. Everything else moves into the product, asked at the moment each answer becomes useful.
Progressive collection converts better twice over: fewer walls upfront, and each later question arrives with its payoff attached, so answer rates rise too.
Let behaviour replace questions where it can: what users do answers more honestly than what they say. Watch before you ask.
And if a first-run question survives the audit, spend design on it: a beautiful, obviously consequential question feels like the product starting, not paperwork in front of it.
The building blocks that solve it
- Progressive Disclosure: Questions arrive one layer at a time, each in its moment.
- Setup Defaults: Defaults answer what questions used to ask.
- Investment: Later asks, each attached to a visible payoff.
Who does this well
Pinterest: Three taps of taste, then straight in: everything else is learned from behaviour later.
Duolingo: The placement quiz is the setup: the questions are the product demonstrating itself.
Questions founders ask
How many setup questions are acceptable?
As many as visibly shape the first experience, which is usually one to three.
The honest test per question: if the user could see what their answer changes, would they call it worth answering before trying the product? Most questionnaires fail that test by question two.
Won't we lose the data we need for personalisation?
You collect it later, better: behaviour is more honest than stated preference, and questions asked at their moment of payoff get truthful answers.
What you lose is junk data from users grinding through a wall. That loss is a gain.