Quiz Exit Intent
Quiz exit intent is a technique that detects when a respondent is about to abandon a quiz and triggers a message or offer to keep them engaged.
Key takeaways
- Desktop exit intent listens for the cursor crossing the viewport's top edge.
- Mobile has no equivalent signal, so back gestures and inactivity stand in.
- Recovery rate rises with answers already invested and falls with prompt frequency.
- Offering to save progress converts better than repeating the original offer.
- A question causing heavy drop-off deserves rewriting, not a recovery prompt.
In depth
Exit intent is a listener rather than a prediction model. On desktop the browser fires a mouse-leave event when the cursor crosses the top edge of the viewport toward the tab bar or address bar, and the script treats that crossing as the signal. Touch devices have no equivalent, so mobile implementations substitute proxies: an intercepted back gesture, a spike in upward scroll velocity, or inactivity beyond a set number of seconds. Each proxy also fires on ordinary behaviour, so a threshold and a frequency cap sit in front of the handler.
Recovery rate rises with how much the respondent has already invested and falls with how often the prompt appears. Someone six answers into a diagnostic has sunk effort and responds to a reminder of what they are about to lose, while someone on the first question has nothing at stake. The trade-off is attention, because every trigger spends goodwill and a prompt shown twice in one session teaches dismissal. Session caps, a minimum step threshold and a suppression cookie keep the mechanism scarce enough to keep working.
In practice teams arm the trigger only across a middle band of questions and offer something that removes the reason for leaving instead of restating the original pitch. Common variants save progress and email a resume link, shorten the remaining path, or deliver the partial result immediately. In a scorecard funnel the strongest version names the score already in progress, since a respondent who has answered most questions is close enough to a tier that the promise of a number feels concrete rather than abstract.
The signal is weak on its own. A cursor leaving the viewport often means someone is switching tabs to look something up and fully intends to return, so the prompt interrupts a person who was never abandoning. Mobile proxies are weaker still, and an intercepted back gesture can read as a broken browser. Exit intent also treats a symptom: if one specific question drives most abandonment, rewriting that question returns far more volume than any recovery layer placed after it.
Example in practice
How to measure it
The number that matters is recovered respondents divided by triggers fired, not divided by all abandoners, because many abandoners never trip the listener at all. Track triggers, acceptances, and how many of those acceptances eventually complete the quiz or return through the emailed link. A high acceptance rate paired with low eventual completion means the offer is attractive but the resume path is broken.
Watch a guardrail alongside it. Compare completion rate and time on step for sessions that saw the prompt against sessions that did not, matched on the step they reached. If exposed sessions complete less often, the prompt is interrupting people who would have finished anyway. Count repeat dismissals per visitor too, since rising dismissals mean the frequency cap is set too loosely.
Common mistakes
The first mistake is firing on every step of the funnel with identical copy, so the respondent meets the same modal at question two and at question eight. The message that works late is a reminder of a nearly finished result, which is meaningless early on. Set a minimum step before the listener arms, cap it at one appearance per session, and write copy that references how far along the person actually is.
The second is treating exit intent as a substitute for fixing the funnel. A team adds the recovery layer, sees a small share of abandoners return, and stops investigating why abandonment was high in the first place. The layer only recovers a fraction of a leak that could have been closed. Read the per-question drop table first, repair the worst question, then add exit intent for the residue rewriting cannot remove.
Frequently asked questions
How does quiz exit intent detect a leaving visitor?
On desktop it watches cursor movement toward the browser bar or rapid upward scrolls, and on mobile it uses back-button presses and inactivity. When those signals fire, the funnel shows a recovery prompt.
Does exit intent hurt the user experience?
Only when it is overused or irrelevant. Triggering rarely and offering something genuinely helpful, like saving progress, keeps it from feeling like a nuisance.
Where in the quiz should exit intent appear?
It works best in the middle of the funnel where respondents have already invested several answers. Those warm, near-complete visitors are the most cost-effective to recover.
How does exit intent work on mobile devices?
There is no cursor to track, so implementations substitute proxies: intercepting the back gesture by pushing a history state, detecting a fast upward scroll, or waiting out a period of inactivity. Each is noisier than the desktop signal and fires during ordinary browsing, so mobile prompts need a higher threshold and a tighter frequency cap than desktop ones.
At which question should an exit-intent prompt appear?
After the respondent has invested enough answers to feel the loss, typically past the midpoint but before the final step. Early prompts have nothing to reference, and prompts on the last question compete with the result the person is already about to see. Arm the listener at a fixed step number and allow one appearance per session.
What should an exit-intent prompt offer in a quiz?
Something that removes the reason for leaving. Saving progress with an emailed resume link works when the barrier is time, showing the partial result works when the barrier is doubt about the payoff, and shortening the remaining questions works when the quiz feels long. Repeating the original offer rarely helps, because that offer already failed to hold the visitor.
Does exit intent hurt SEO or page performance?
Not directly for search, since these prompts fire on user interaction rather than page load and are not the intrusive interstitials that ranking guidance targets. The performance cost of the listener script itself is small. The real risk is behavioural: prompts that make a page feel hostile increase bounces, and that pattern does affect how the page performs.
Why is my exit-intent conversion rate so low?
Most often because the trigger fires too early, when the respondent has invested nothing, or because the offer restates what they already declined. Check what share of triggers come from the first two questions, then check whether the prompt appears more than once per session. Both problems suppress acceptance quietly, without producing any visible error to investigate.
Can exit intent capture an email from someone who never finished?
Yes, and that is its most common use in a lead funnel. Offering to send the in-progress result gives a reason to hand over an address that the unfinished quiz alone does not provide. Treat those contacts as a separate segment, because they carry a partial answer set and a partial score and need a different follow-up sequence than finishers.