Exit-intent survey questions that tell you why shoppers leave

The short answer

An exit-intent survey is a short question shown to visitors at the moment they seem about to leave. The best one asks one question, on the page where you lose the most people. Make it about the decision: “What’s stopping you from buying today?” Read every answer, group them into reasons, and act on the biggest one.

The rest of this article covers each of those steps: the trigger, the question, the number of answers and what to do with them.

What exit intent means, on a computer and on a phone

On a computer, “exit intent” usually means the cursor moving quickly towards the top of the window, where the tabs and the close button are. It’s a decent signal that someone is about to leave.

Phones have no cursor, so the usual triggers don’t exist. Workable alternatives are:

• Coming back to the tab. The browser fires a visibilitychange event when someone switches tabs or, on a phone, switches to another app. That’s often to compare prices. Asking when they come back catches them at exactly the right moment.
• A quick scroll back to the top, which often comes just before tapping back.
• Time on the page without progress, such as a minute on the cart without starting checkout.

Whatever the trigger, keep it small on phones. Google’s guidance on pop-ups warns that intrusive ones “may frustrate them and erode their trust” and can hurt how a page performs in search, and suggests banners that take up only a small part of the screen. A one-line question that slides up from the bottom does the job.

Ask about the decision, not the experience

Many exit surveys ask questions whose answers you can’t act on:

• “Did our site meet your expectations?” A “no” tells you nothing about what to change.
• “How would you rate your visit from 1 to 10?” A 6 has no cause attached.
• “Would you recommend us to a friend?” Someone who hasn’t bought yet has nothing to recommend.

The useful questions ask what stood between the shopper and the purchase, in words that invite a specific answer:

• “What’s stopping you from buying today?”
• “Is anything missing from this page that would help you decide?”
• “Was the total what you expected?”

Start open, then switch to multiple choice

Whether a question is open or multiple choice is, in the words of Pew Research Center, “one of the most significant decisions that can affect how people answer questions”. In one of Pew’s experiments, 58% of people picked “the economy” from a list of issues, but only 35% named it when asked the same question without a list. And 43% of those answering the open version gave an answer the list didn’t include at all.

A list puts words in shoppers’ mouths, and it can only contain the reasons you already thought of. So do what Pew does: start with an open question to find out which answers are common, then build the multiple-choice list from those.

The switch is worth making. Multiple-choice questions are much easier to answer, so more people reply, and the answers add up by themselves. When you do switch:

• Use the shoppers’ own words for each option, not your internal jargon.
• Keep “Just looking”: plenty of visitors were never going to buy today, and you want to know how many.
• Keep “Something else” with a text box, so new reasons can still surface.
• Shuffle the order of the options, so the first one doesn’t win just by being first, but keep “Just looking” and “Something else” at the end.

Questions to ask on each page

Different pages lose people for different reasons, so each deserves its own question. These are good starting points.

The thank-you page question is the one most stores forget. An answer like “I almost didn’t buy because I couldn’t find the returns policy” tells you exactly what non-buyers ran into.

For the cart and checkout, it helps to know the reasons that come up across the industry. Baymard Institute’s survey of shoppers who abandoned a checkout (updated September 2025) found that extra costs (40%), slow delivery (20%) and not trusting the site with a card (19%) top the list. Baymard leaves out shoppers who were just browsing, so set your “Just looking” answers aside before you compare. Your own answers tell you which of these applies to your store, and this guide shows how to fix each cause of cart abandonment.

How many answers you need before acting

Fewer than you might think, but more than five. A rough guide, using the standard margin of error for a share:

• With 100 answers, a reason given by half of them would be accurate to about ±10 percentage points, if those answers were a random sample.
• With 400 answers, about ±5 points.

Yours won’t be random (more on that below), so treat these margins as the least uncertainty you have, not the most.

For comparison, Pew notes that a properly random sample of 1,067 people gives a margin of about ±3 points. For ranking the reasons shoppers leave, you don’t need that precision. Around 100 answers per page is enough to see which two or three reasons dominate. Split by device when you can: phone and desktop shoppers often leave for different reasons.

Some answers need only one occurrence. If a shopper writes “the discount code field gives an error”, check it today.

Who answers, and why that matters

An exit survey is not a poll. The people who answer chose to. Survey researchers call this self-selection bias: it matters whenever the choice to answer is related to the topic. Annoyed shoppers may be more eager to tell you what went wrong; rushed ones skip the question.

So read the answers as a list of real problems and a rough ranking, not as exact percentages of all your visitors. And don’t bribe people to answer. A discount for replying attracts the shoppers who wanted a discount, and teaches everyone that heading for the exit earns a coupon.

Turn answers into fixes and tests

A pile of answers is only useful once it’s sorted.

1. Read the first 100 or so answers and write down the reasons as they come up. Most will fit into five to eight groups: price, delivery cost, delivery time, missing information, trust, payment, comparing, just browsing.
2. Tag every answer with one group, and count.
3. For each of the top reasons, decide: are the cause and the cure obvious? A missing size chart, a hidden delivery cost or a broken field gets fixed straight away. If it isn’t obvious (“too expensive” could mean the price, the delivery cost or the value not coming across), it becomes an A/B test.
4. Check the result in your numbers, not in the survey. If delivery costs were the top reason in the cart, orders per visitor should improve once they’re shown earlier; the cart-to-checkout rate shows where the change acted. How long a test needs to run depends on your traffic.

The test idea library has a ready-made test for most of the common reasons, and session recordings of real visits show where in the page shoppers got stuck.

Privacy and consent in Europe

This isn’t legal advice, but three points are worth checking with whoever advises you on privacy:

• Free-text answers can contain personal data. Don’t ask for names or emails in an exit survey, and treat what people type with care.
• Remembering who has already seen the survey (so you don’t ask twice) usually means storing something in the browser. Under the ePrivacy rules, that can need consent. France’s regulator, the CNIL, states that cookies can’t be set or read until the person has consented, apart from a short list of exemptions such as the shopping cart.
• Say what the answers are for, in one line next to the question.

Mistakes to avoid

• Asking several questions. Each extra question loses people. One question, perhaps with one follow-up, is plenty.
• Showing it on every page, every visit. Pick the pages that leak most, and show it once per visitor.
• Interrupting the checkout. Ask when someone is clearly leaving, never while they’re typing their address.
• Collecting answers nobody reads. Put 30 minutes a week in the calendar to go through them.

Surveys of this kind are what Wavelynks surveys are built for: one question, in the shopper’s language, with answers grouped into reasons automatically.