Exit Intent: When Leaving Visitors Are Worth Asking
Exit intent can turn a lost visit into a useful answer, if you ask the right question on the right page. When it works, when it does not, and how to do it on mobile.
Most visits end without a conversion. That is normal, and for most pages it is not a problem you can fully fix. What you can change is whether you learn anything from the visitors who are about to leave. Exit intent is one way to do that: you notice the moment someone seems to be leaving and ask a single, useful question.
Done badly, it is the “Wait, don’t go!” overlay that everyone has learned to close without reading. Done well, it is a short question on the right page that tells you why people did not buy, sign up or book a demo. This article covers what exit intent actually detects, where it makes sense, where it does not, and how to run it on mobile.
What exit intent actually detects
On a desktop browser, exit intent usually means the mouse pointer moves quickly toward the top edge of the viewport, where the tabs, the address bar and the back button sit. A script watches for that movement and fires a trigger. It is a guess about intent, not a fact. The visitor may be reaching for another tab, a bookmark or the browser menu, and they may leave without ever moving the pointer upward, for example by using a keyboard shortcut.
That uncertainty is the first thing to accept. Exit intent catches a share of departures, not all of them, and some of the people it catches were not leaving. This is a good reason to keep the prompt small and easy to dismiss. If the guess is wrong, the cost to the visitor should be close to zero.
When asking a leaving visitor makes sense
The strongest cases are pages where a visitor has shown real interest and then stalled. A few examples:
- Pricing pages. Someone who reads the plans and leaves has a concrete objection: price, limits, a missing feature or a missing trial. One question about what was missing often surfaces the answer within a few dozen responses.
- Product and feature pages. A visitor who looked at a feature in detail and then left probably did not find what they were looking for. Asking what they expected to see is more useful than any heatmap.
- Long articles and guides. Readers who scroll deep and then leave can tell you whether the article answered their question and what they wanted next.
- Cart or signup steps before the final commitment. Baymard’s research on checkout abandonment shows that many reasons are visible on the page itself, such as unexpected extra costs or forced account creation. A short question can confirm which one applies to your visitors.
What these cases have in common is that the visitor has already invested some attention. You are not interrupting a stranger, you are following up with someone who almost converted.
When it does not make sense
There are moments where exit intent is more annoying than helpful, and the Nielsen Norman Group lists several of them in its overview of problematic popup patterns. Prompts that appear before the visitor has seen any value, in the middle of a task, or stacked on top of each other all damage the experience.
For exit intent specifically, avoid these situations:
- Payment and other high-focus steps. A visitor typing card details should not be interrupted by anything.
- Confirmation pages. Someone who just completed an order is not leaving a failed funnel. If you want feedback here, ask about the experience instead.
- Inside the product, during a task. Logged-in users who move toward the tab bar are usually switching tasks, not abandoning your product.
- Repeat visitors who already answered. Cap the frequency. If a person has seen or answered the prompt once, do not show it again for a long time.
- Right after the page loads. A visitor who has been on the page for four seconds has not formed an intent yet. Require a minimum dwell time or scroll depth before the trigger can fire.
Ask a question, do not pitch a discount
The typical exit overlay tries to talk the visitor out of leaving with an offer. That can work for a coupon-driven shop, but it teaches visitors to wait for the discount, and it tells you nothing about why they hesitated.
A question does the opposite. It costs the visitor a few seconds and gives you information you can act on. Some formats that tend to work well:
- A single closed question with three or four answers, such as “What stopped you today?” with options like price, missing feature, not ready yet and just browsing.
- A single open question with a short text field, such as “What were you hoping to find on this page?”
- A yes or no question with an optional follow-up, such as “Did this page answer your question?”
Keep the prompt visible but small. A corner card or a slim bar is easier to ignore than a centered modal, and that is the point: people who want to answer will, and people who do not can carry on. Make the close button obvious and do not make the visitor confirm that they want to close it.
If you offer an incentive, say so plainly and do not pretend it is anything else. A small, honest thank-you is fine. A countdown timer or a guilt-trip button label is a dark pattern, and it damages the trust you are trying to protect.
Exit intent on mobile
On a phone there is no pointer, so the desktop trigger does not exist. Tools that support mobile approximate intent with other signals: a quick upward scroll toward the top of the page, a long period of inactivity, or a return to the previous page. None of these is as reliable as the desktop movement, so treat mobile results as rougher data.
Two practical points follow. First, screen space is limited, so a prompt that is fine on desktop can cover most of a phone screen. Use a compact bottom sheet or a slim bar and keep the text short. Second, be cautious about tricks that interfere with the back button. Intercepting navigation to show a prompt frustrates people and is widely treated as a deceptive pattern.
Keep search and speed in mind
Google’s guidance on interstitials says that dialogs that obscure the whole page, especially promotional ones, can make it harder for search engines to understand your content, and it points to small banners as the acceptable alternative. Exit prompts appear late in a visit, but the principle applies: do not cover the content, and do not show a full-screen overlay on arrival.
Speed matters too. A prompt script that loads late and shifts the layout hurts the page it is supposed to help. Load the script asynchronously, reserve space for it or render it as an overlay that does not push content, and check your Core Web Vitals before and after you add it.
How to tell whether it works
The obvious metric, how many visitors the prompt “converts”, is the least interesting one. Look at these instead:
- Response rate. What share of the people who saw the prompt answered? If it is very low, the question or the timing is off.
- Answer quality. Do the answers repeat? Patterns across twenty or thirty responses are more useful than any single comment.
- What you changed because of it. The real return is a fix on the page: a clearer pricing note, a missing comparison, a shorter form.
- Side effects. Watch bounce rate, complaints and page speed to make sure the prompt is not hurting the experience.
Run it on one page first, for a few weeks, and review the answers on a fixed day each week. If nothing you learn changes the page, the prompt is not earning its place.
A simple setup to try this week
- Pick the page with the most visitors who show interest and then leave, usually the pricing page.
- Write one question. Make the answers short and add an “other” option.
- Require at least 15 seconds on the page or 50 percent scroll depth before the trigger can fire.
- Show it once per visitor and cap it at once per month.
- Make it dismissible with one click and place it so it does not cover the main content.
- Review the answers weekly, group them into themes and fix the most common one first.
In conversionloop, exit intent is one of the signals you can combine with dwell time, scroll depth and inactivity when you define who sees a campaign, so the conditions above can be set without code. The same approach works in any tool, though. What matters is that you ask a real question, at a moment that makes sense, and then act on what you hear.
Frequently asked questions
What is exit intent?
Exit intent is a trigger that fires when a visitor seems about to leave a page, usually detected on desktop by the pointer moving quickly toward the top of the browser window.
Does exit intent work on mobile?
Not in the same way, because there is no pointer. Tools approximate it with signals such as a quick upward scroll, inactivity or navigating back, and these are less reliable than the desktop trigger.
Should an exit prompt offer a discount or ask a question?
A question usually teaches you more. A discount can lift short-term conversions in some shops, but it tells you nothing about why people hesitated and can train visitors to wait for offers.
How often should I show an exit prompt?
Once per visitor, with a long pause before it can appear again, and never to someone who has already answered.
Sources and further reading
- Popups: 10 Problematic Trends and Alternatives , Nielsen Norman Group
- Interstitials and dialogs , Google Search Central
- User-Feedback Requests: 5 Guidelines , Nielsen Norman Group
- Cart Abandonment Rate Statistics , Baymard Institute
- Web Vitals , web.dev (Google)