The conversion leak index

Most sites do not have one conversion problem. They have about forty, and the only question worth answering is which three are costing real money.

You already know something is broken. That is not the hard part. The hard part is that when you go looking, you find twenty things at once. A slow hero image. A form asking for a phone number. A pricing page nobody can compare. All of them are real. None arrive with a price attached, so you fix whichever annoys you most, and the number does not move.

What are the most common conversion problems?

They cluster into five groups: unclear value on the first screen, friction in forms and checkout, missing proof at the moment of doubt, mobile layouts that break under 400 pixels, and slow loading. Across the 134 sites Revslip has audited, almost every one carries problems in at least three of those five.

That is the uncomfortable finding. Not that sites have a problem, but that they have several at once, and the owner only sees the ones that irritate them.

Each group is a bundle of named checks with thresholds, which is what a conversion signal actually is.

Pricing is its own category. For B2B it often comes down to whether you show a price at all.

Is this happening on your site?

Open your analytics and compare mobile conversion rate to desktop over the last 90 days. If mobile is running below 70 percent of desktop, you have a leak worth pricing. That comparison takes two minutes and surfaces something more often than any other check here.

Then the harder one. Open your homepage on a phone, cover the logo with your thumb, and read the headline. Could a stranger say what you sell and who it is for?

Most people cannot run this test on themselves. You know what the page means, so you read the meaning instead of the words.

See every leak on your site, ranked and priced, free.

Where the money actually goes

A problem costs you in proportion to three things: how often people hit it, how badly it blocks them, and whether it keeps hurting after they learn the workaround. That is the severity model Nielsen Norman Group has used for decades, and it is why a raw error count tells you nothing.

A dead link in your footer and a dead payment button are both one error. One of them is free. The other one is your month.

So price them. The formula is traffic × CVR gap × AOV × mobile weight, published openly because you should be able to check it by hand. Here is how to calculate conversion loss, step by step.

Picture this: 18,000 sessions a month, 68 percent on mobile. Desktop converts at 2.1 percent, mobile at 0.9 percent. Average order value €74.

That is 12,240 mobile sessions and a gap of 1.2 percentage points. Which comes to 147 orders a month, or €10,878, sitting behind whatever breaks on small screens.

Nobody closes the whole gap. Mobile intent is lower. Halve it and you are still looking at €5,439 a month, more than most people spend on the tools that failed to find it.

An error count tells you how messy your site is. It does not tell you which mess is expensive.

What the data says

Scale first. The 2026 WebAIM Million, a non-commercial analysis of one million home pages, detected 56.1 accessibility errors on the average page, with 95.9 percent failing at least one WCAG check.

That checkout figure is Baymard Institute, which benchmarked 60 large ecommerce checkouts and averaged 39 improvement areas each. Worth saying: Baymard sells research subscriptions, so read it as a vendor's count. WebAIM sells no CRO anything, which is why it carries the headline claim here.

Speed is the one people underestimate. The 2025 Web Almanac, built on real Chrome user data rather than lab tests, put 48 percent of mobile sites through all three Core Web Vitals.

Loading is the one that fails. The other two are mostly handled.

What to change, in order

Order matters more than the list. Work down by money, not by irritation. Four steps, and the first is the one everybody skips.

1. Price it before you fix it. Run the formula on every candidate leak. Anything under a few hundred a month goes to the bottom however ugly it is. Check: you can state a euro figure for your top three.
2. Fix the first screen. The headline should say what your product does, not what it is. Check: a stranger reads it and describes your business back to you correctly.
3. Close the mobile gap. Test at 360 pixels, not on your own phone. Tap targets, sticky bars covering the button, forms opening the wrong keyboard. Check: mobile climbs toward 70 percent of desktop.
4. Strip the form back. Remove every field you do not act on within a week. Check: completion rate up, not merely more starts.

If your instinct now is to rebuild the whole thing, read redesign or fix before signing anything.

Paste your URL and get every leak priced in minutes.

How to measure the change

Give it four weeks and at least 300 conversions before you believe anything. Below that you are reading noise. Most small sites never reach statistical significance on a single change, which is not a reason to stop testing, it is a reason to stop pretending the test will settle the argument.

Compare matched periods instead. Same weeks, same channels, same promotions. Write down what you expect to see before you look.

When this does not apply

This index assumes your product and your traffic are roughly right. If you are buying clicks from people who were never going to buy, no amount of page work rescues it, and the traffic but no sales diagnostic rules that out faster than anything here.

A caveat on our own numbers. Those 134 audits are not a random sample of the web. They are sites whose owners already suspected something was wrong, which is precisely the group most likely to have three problems running at once. Our frequency claims describe that population, not yours by default.

Questions people ask about a leaky site

Three come up more than any others. The short version: the number of problems on your site is not the useful figure, the order you fix them in is, and whether hiring help pays back is set by your traffic rather than by how bad the site looks.

Should I fix everything, or only the big ones?

Only the big ones, and be strict. Shipping eleven small fixes at once means you learn nothing about which one worked. Take the two with the largest euro figures, release them separately, and measure each on its own.

Is it cheaper to find these myself or pay someone?

It depends on your traffic. Below a certain volume no consultant pays back, whatever they find, because their percentage lift lands on too small a base. We worked the thresholds out in what a CRO audit actually costs.

What about form fields, which everyone argues about?

Field count matters less than field type and rework. Four large datasets disagree on ideal length, covered in how many form fields is too many. The phone field specifically is not the villain it is treated as.

What if the problem turns out to be the traffic, not the pages?

It happens, and it is worth ruling out before you fix anything here. Conversion split by source separates the two in about twenty minutes, using data you already have. The method is in is it your traffic or your website.

Can a heatmap find these leaks for me?

It shows clicks and scroll depth, but it cannot rank a leak or price one, which is the limit built into every Hotjar alternative.

How bad does my conversion rate have to be before I act?

There is no universal threshold, because the same month produced 0.55% in one sector and 5.23% in another. Price the leak instead of judging the ratio, and start with is my conversion rate bad to set an honest baseline.

Does this list apply to a Shopify store?

Most of it does, but Shopify removes some entries and adds others. You cannot rebuild the checkout on a standard plan, and shipping cost appears late by default, which changes where the money goes. The platform-specific version is Shopify traffic but no sales.

Which of these leaks sits in my checkout?

Checkout has its own ranked list, because the reasons people abandon there are measured separately from the rest of the site. Each one is priced and ordered in why is my checkout abandoning.

Where does the very first leak happen?

Before the scroll. The opening screen is judged fastest and, on the evidence, serious business pages are judged harder than entertaining ones. Why people leave a landing page covers that first screen in detail.

Do the trust badges on my checkout do anything?

Only if your buyers recognise them. The one peer-reviewed test of purchase probability found no average effect, and a seal nobody can name is decoration. What the published studies actually found.

Is my hero headline one of these leaks?

Often the first one. If a competitor could use your headline unchanged, it is naming your category rather than what changes for the reader. Run the logo-swap test.