UserProof

Why users leave your website without signing up, and how founders of a small site or SaaS find out why

Updated . By Ryan Doyle, founder of UserProof, a session replay tool for coding agents.

For founders of a small site or SaaS, the method is five steps. Pick the one goal, such as sign-up. Find the funnel step where most visitors drop. Watch real sessions of people who left at that step. Ship one fix. Measure the same step again. Sales.co ran this loop in six days (October 20 to 26, 2026), and the share of its x.com visitors who stayed past the first screen went from 24% to 47%.

It works the same whether you build by hand or with a coding agent such as Claude Code. This page shows the same home page losing 76% of its x.com visitors and 28% of everyone else, then the one fix that moved it. The data is from Sales.co's recorded visits on October 20 to 26, 2026. Next to it sits our check of October 28, 2026: 122 of 378 YC startup homepages have no sign-up button on the first phone screen. To score your own homepage on the same 10 checks, use the free homepage grader.

Jump to: why they leave · were they buyers · which step · watch the step · fix and check · tools · sources · FAQ

Why do users leave a website without converting?

Usually because of one specific thing on one specific screen, which a site-wide average hides. Sales.co is a done-for-you outbound service, and much of its traffic comes from x.com. In the 14 days before its fix, 90 of 118 x.com visits to its home page bounced (76%), against 28% of visits from other sources. 69% of the x.com visitors were on a phone (Sales.co story, October 2026). The phone's first screen held a headline and a seven-line paragraph, and both buttons sat on the second screen. Those visitors left at a median of 7 seconds, 12% of the way down the page, and 136 phone visits produced zero button clicks, per the story's check of October 26, 2026.

No survey would have caught that: nobody writes in to say the button was one swipe too low. The Sales.co story has the full before and after.

Did the visitors who left ever intend to sign up?

Some did not, and you can tell which by splitting your leavers by source: on Sales.co's home page the same page lost 76% of x.com visitors and 28% of visitors from other sources. If most leavers were never buyers, moving the buttons could not have helped. It did: after Sales.co put a one-line subhead and both buttons on the phone's first screen, 47% of x.com visitors stayed past the first screen, up from 24%. Phone visitors went from 0 button clicks in 136 visits to 4 in 155.

That is one site and one week, so treat it as a worked example, not a rate to expect. The bounce drop was 23 points, with a 95% interval of 10 to 36.

How do you find the step where visitors drop?

Count the people who reached each step of one goal, in the same date range from the same tool, and compare the steps by how many finishers you would gain back. Our funnel calculator does the arithmetic in your browser. Its example funnel, which is made-up numbers for illustration and not a measurement, starts with 8,000 home page visitors and ends with 31 who paid (0.39%).

Funnel calculator example (illustrative numbers), extra paying customers if 10% of the people who leave a step stayed
StepPeopleLeft hereExtra at the end
Visited the home page8,000startstart
Clicked Sign up9607,040+22.7
Submitted the sign-up form610350+1.8
Finished setup240370+4.8
Paid31209+20.9

The biggest loss is between the home page and the Sign up click, so that step is the one to fix first. The ranking says where the most finishers are, not how hard each fix is, so weigh the size against the work. It also says where people leave, not why.

How do you see why visitors left at that step?

Watch the sessions of people who left at that step, and look for the same stall in several of them. A recorded session shows the click that did nothing, the button below the first screen or the error that ended the visit. For the home page, a first-time visitor on a phone runs into the same few problems again and again. We checked 378 YC startup homepages on a phone and a desktop on October 28, 2026, and these were the problems found: On a site with too little traffic to watch, run a website usability test with five people who match your buyer instead.

Startups failing each check, 378 YC homepages (Winter 2025 to Fall 2026 batches), UserProof check, October 28, 2026
ProblemShareStartups
Main button too small to tap35%95 of 268
No main button on the first phone screen32%122 of 378
Slow to load25%96 of 378
Dead clicks8%29 of 378
JavaScript errors5%19 of 378
Main button covered4%11 of 290
Broken links3%13 of 378
Main button does nothing1%2 of 298

Each share is out of the startups where the check could run, which is why some rows have fewer than 378. Of the 378, 93 passed all 10 checks and the median score was 95 out of 100. The YC homepage report lists every startup's score and the 10 checks.

A dead click, a click that changes nothing, is the one you can count from events. In our separate scan of 493 YC startup homepages on October 28, 2026, 33 (6.7%) had at least one, and 43 of the 49 dead clicks were on real buttons and links, not plain text. The dead click report has the method. To test your own page now, paste it into the dead click finder.

How do you fix it and check that it worked?

Ship one change aimed at the stall you watched, then measure the same step over the same kind of visit. Sales.co's change was phone-only: a one-line subhead and both buttons (START NOW and BOOK CALL) on the first screen, with desktop left alone. It shipped on October 20, 2026, and UserProof checked it six days later:

x.com visits to the Sales.co home page, UserProof check_fix, read October 26, 2026
MeasureBefore (14 days)After (6 days)
Bounced90 of 118 (76%)45 of 85 (53%)
Took a next step25 of 118 (21%)38 of 85 (45%)

Over the same period the share taking a next step fell on Sales.co's other pages, so the lift was not site-wide drift. Change one thing at a time, or you will not know which change moved the number.

Which tools can show you the sessions?

Several can: Microsoft Clarity, Hotjar and PostHog all record sessions, and UserProof is one route among them. Our session replay guide covers what these tools record, what they mask and what they cost. UserProof records real visitor sessions with a script tag and counts dead clicks, rage clicks, errors on click and slow responses from recorded events, never from an LLM. It then hands your coding agent the one highest-impact fix with the file and line, and checks afterward whether your goal moved. Field values are never recorded. It has no heatmaps and no surveys. Which tool fits depends on whether you want a chart for yourself or a fix for your agent, and our ranking of 12 replay tools compares them on six stated criteria. If you are leaving Hotjar, start with the Hotjar alternative page. If you also want A/B tests, our ranking of conversion rate optimization tools for AI coding agents prices eight of them.

UserProof plans, read from the pricing page on November 7, 2026
PlanPriceRecorded visits a month
Free (3 sites)$05,000
Pro$29 a month50,000
Scale$99 a month250,000

FAQ

How do I find out why users leave my website without converting?

Pick one goal, such as sign-up. Find the funnel step where most visitors drop. Watch real sessions of people who left at that step. Ship one fix. Then measure the same step again. Analytics shows where people leave; only the sessions show why.

Is a high bounce rate always a problem?

No. A bounce rate only says visitors left from the first page. Split it by traffic source before you act: on Sales.co's home page, x.com visitors bounced 76% and visitors from other sources 28%, so the page itself was not the whole problem. The first screen those visitors saw on a phone was.

How many visitors do I need before session replay helps?

Enough to see the same problem more than once. Our dead click rule needs at least 10 sessions and 3 visitors on a page before it reports a finding. Sales.co's report was based on 136 phone visits. On a smaller site, watch the sessions yourself and look for one repeated stall.

Can I find out why visitors leave without a paid tool?

Yes. Microsoft Clarity has no cap and no paid tier, according to our replay ranking, and UserProof's Free plan covers 3 sites and 5,000 recorded visits a month. Our free funnel calculator and dead click finder need no sign-up.

Sources, read November 7, 2026