How to find where visitors leave
For owners and marketers who know conversions are low but not why. After reading, you can map your funnel, find the step that loses the most people and gather evidence on what to fix.
Map the steps between arriving and converting, count how many visitors reach each step, and fix the step with the biggest drop first. Use a funnel report in GA4 for the numbers, then watch session recordings and heatmaps in a tool such as Microsoft Clarity to see why. Say an apparel store has 10,000 product page views, 900 add-to-carts, 600 checkouts started and 150 orders. Checkout loses three in four of the people who start it, so start there, not on the homepage.
Write your funnel down as four to six steps
A funnel is the path a visitor takes from landing on your site to converting. Writing it down turns a vague worry into numbers you can compare. Keep it short; each step should be a page view or event you already track.
- Online store: product page, add to cart, start checkout, add shipping, add payment, order
- Med spa or clinic: treatment page, booking page, choose time, enter details, confirmed
- Café: menu page, add item, cart, checkout, order placed
- B2B or SaaS: landing page, pricing page, form started, form submitted
Numbers show where people leave, recordings show why
Use analytics to find the leaky step, then watch real sessions at that step to understand it. GA4's funnel exploration report shows how many people complete each step and where they drop. Look for the largest drop in absolute numbers, not just percentages.
Then open a behavior tool. Microsoft Clarity is free and offers session recordings, heatmaps and scroll maps; Hotjar is a common paid alternative. Filter recordings to visitors who reached the leaky step but did not finish. Watch for rage clicks, which are repeated taps on something that does not respond, dead clicks on things that look clickable but are not, and people scrolling up and down hunting for information.
Split every report by device and traffic source
A blended average hides most problems, so always compare mobile with desktop and one channel with another. A checkout might convert well on desktop and poorly on phones because a payment button sits below the keyboard. Paid social visitors often behave differently from people who searched for you by name.
Look at new versus returning visitors too. If returning visitors convert well but new ones do not, the problem is often trust or clarity, not the checkout itself.
Ask visitors directly
Short surveys give you reasons that no report can. A one-question survey on the leaky page, or right after purchase, is enough. Good questions are open and specific, such as 'What, if anything, nearly stopped you from ordering today?' or 'What information is missing from this page?'
Your team already holds answers too. Read recent support emails, chat transcripts, reviews and the questions staff hear on the phone. The same three or four worries usually come up again and again: price, shipping, sizing, timing or trust.
Check speed and errors at the leaky step
Slow pages and silent errors cause drop-offs that look like disinterest. Run the page through Google's PageSpeed Insights and check the Core Web Vitals report in Search Console. These measure loading speed (Largest Contentful Paint), how quickly the page responds to taps (Interaction to Next Paint) and how much the layout jumps around (Cumulative Layout Shift).
Then try to break the step yourself on a real phone. Use a discount code, pick an out-of-stock size, enter a phone number with spaces, choose express shipping. Errors that only appear in edge cases are common and easy to fix once found.
Do this next
Questions people ask
How many session recordings should I watch?
Twenty to thirty recordings filtered to the problem step usually reveal the main patterns. Stop when you keep seeing the same behavior and start fixing.
Are session recordings allowed under privacy law?
Generally yes, if you mask personal data, mention the tool in your privacy policy and respect consent rules where you operate. Most tools mask form fields by default; check that yours does.
What is a normal drop-off rate?
Some drop-off at every step is normal, and the rate varies widely by business. Compare your own steps against each other and over time, and focus on the step that loses the most people.



















