80 On-Site Survey Questions for Online Shops

Choose a goal first

Every question should serve one decision you're about to make. Some common goals for shops:

Two kinds of answer types:

• Open questions ("What's stopping you…?"): slower to analyze, but they show you problems you didn't think of, in your customers' words.
• Closed questions (multiple choice, scales, yes/no): fast to answer and count, but only find what you already listed.

Start open. After 100–200 answers, turn the top themes into a multiple-choice list with an "Other" option.

Exit-intent questions

Exit intent means the survey appears when a desktop visitor moves the mouse toward the tabs or address bar, as if to leave. Phones have no mouse, so on mobile these surveys usually appear after a set time or scroll depth instead.

Cart and checkout questions

Ask these when visitors are about to leave the cart, or site-wide on exit intent if you don't run surveys on the cart page. Never put surveys inside the payment step itself.

Product page questions

Category and search questions

Post-purchase questions (thank-you page)

Buyers have just said yes, so they answer more willingly. This is the place for attribution and "what almost stopped you".

A note on "How did you hear about us?"

This question (often called post-purchase attribution) shows channels that analytics miss: podcasts, word of mouth, a friend's Instagram story, offline ads. That's why many shops use it.

Its limits:

• People remember the last or most memorable touchpoint, not the full path.
• Answer order matters. The first options get picked more. Shuffle answers (keep "Other" last).
• "Google" means many things: an ad, an organic result, or just typing your name after seeing you elsewhere.
• Not everyone answers, and the people who answer may differ from those who don't.

Use it next to your analytics, not instead of it. When both point to the same channel, you can be more confident. When they disagree, ask why.

Follow-up questions that get better answers

A follow-up appears only after a specific answer (also called logic or branching). It turns a vague first answer into something you can act on.

Loyalty questions for returning visitors

Triggers and targeting

The same question gets very different answers depending on when and to whom you show it.

Targeting tips:

• One survey per page at a time. Don't stack pop-ups.
• Don't show it again to people who answered or closed it, at least for a while.
• Show to a share of visitors (say 20–50%) on busy pages, so most people are never interrupted.
• Split by device if mobile and desktop pages differ a lot.
• Use the visitor's language. Answers in the visitor's own language are richer.
• Respect consent. If the survey links answers to a tracked visit, it relies on the same consent as your other tracking in the EU (GDPR and CRO).

How many answers you need

For themes in open answers: 100–200 answers usually show the main themes. A practical check: if the last 30 answers you read added no new theme, you have enough.

For percentages in multiple choice: the margin of error shrinks slowly. For an answer picked by 30% of respondents, the 95% margin is about ±1.96 × √(0.3 × 0.7 / n):

So with 100 answers, "30% say shipping is too expensive" really means "somewhere between about a fifth and two-fifths". That's fine for spotting the biggest problem, not for small differences between options.

How long will it take? Divide the answers you need by your expected answers per day. If a survey reaches 2,000 visitors a week and 3% answer, that's 60 answers a week: about two weeks for 100 answers. Response rates vary a lot by question, trigger and design, so measure your own after the first week.

Analyzing open answers (manually and with AI)

Manually (up to a few hundred answers):

1. Read 50 answers and write down themes as they appear.
2. Tag every answer with one or two themes (a spreadsheet column).
3. Count per theme; pick 2–3 quotes each.
4. Split by device and by buyers vs non-buyers if your tool links answers to orders.

With AI (any number): paste the answers into an AI assistant with a clear prompt: group into at most 6 themes, count them, quote word for word, flag uncertainty. Then check by reading 10 raw answers per theme. Our AI for CRO guide has a prompt you can copy.

Turn themes into action:

• Missing information → add it to the page (often no test needed).
• Bugs or confusion → fix.
• Bigger changes (layout, offer, messaging) → write a hypothesis and test if your traffic allows it (A/B testing with low traffic).

Avoid these mistakes

1. Asking too much. Five questions on exit intent get few complete answers. One to three is plenty.
2. Leading questions. "How much did you love our new design?" Ask "How would you rate…?" instead.
3. Two questions in one. "Was delivery fast and affordable?" Split them.
4. Only multiple choice. You'll only learn about options you already thought of.
5. Interrupting at the wrong moment. Never during payment; not the second someone lands.
6. Asking for an email to answer. It cuts responses. Make it optional, at the end.
7. Collecting and never reading. Schedule a monthly review; turn off surveys you don't analyze.
8. Changing a question mid-survey. Older answers then appear under the new wording. Add a new question instead.
9. Treating one answer as a trend. Count before you act.
10. Forgetting privacy. Open answers can contain names, emails and order numbers. Store and share them carefully.

How Varly fits in

Varly, a CRO platform, includes on-site surveys with templates for several of the goals above ("Why are you leaving?", "Anything stopping you from checking out?", "Is anything unclear about this product?", NPS, "How did you hear about us?" on the thank-you page, and more), follow-up logic, the triggers listed above, and an AI summary of open answers with themes and quotes. Each answer is linked to the visit, so you can compare buyers and non-buyers or watch the recording behind an answer (create a survey, read your results).

FAQ

What's a good response rate for an on-site survey? It varies too much by question, trigger and design for one benchmark. Measure your own in the first week and compare versions of your own survey.

Should I offer a discount for answering? Usually not on exit-intent surveys: it attracts discount hunters and changes the answers. A thank-you message is enough.

Open or multiple choice first? Open first, to learn the words and options. Multiple choice later, to count.

How often should I change my surveys? Keep one evergreen survey (for example, "What almost stopped you?" on the thank-you page) and rotate one research survey per month.

Can I ask for an email address? Yes, optionally, at the end, with a clear reason ("so we can reply"). Mention it in your privacy policy.

---

Varly links every survey answer to the visit behind it, with AI themes and quotes. Request access.

Related: How to analyze heatmaps · AI for conversion optimization · GDPR and CRO · Ecommerce CRO guide · A/B testing with low traffic