Mobile conversion problems that desktop testing never reveals

This is not a design philosophy argument. It is a measurement problem. You approve the site on a 27 inch monitor, your designer builds on one, and your customer is holding a five inch screen in one hand with a toddler on the other arm.

Resizing your browser window does not close that gap either. A narrow desktop window has a mouse, a keyboard, fast internet, and no onscreen keyboard eating half the display.

Why does testing on a desktop miss so much?

Because the four things that break on mobile do not exist on a laptop.

There is no thumb. A mouse pointer hits a ten pixel link perfectly. A thumb does not, and the miss rate on small links is where a lot of quiet loss lives.

There is no onscreen keyboard. When it opens on a phone it covers most of the screen, and whatever was under it, often the submit button, is gone until the visitor knows to dismiss it. Plenty of people do not.

There is no slow connection. Your office wifi is not the parking lot where a customer is standing. Google's analysis of 11 million mobile web domains, loaded over a globally representative 4G connection in January 2018, found it took more than five seconds for above the fold content to display and more than seven seconds for a mobile landing page to fully load. That is the median experience of the internet your customer is actually on.

And there is no vertical stacking. Desktop layouts sit side by side. On a phone those columns stack, so a three column layout becomes a very tall page and the form that looked like it was next to the text is now four screens below it.

Everything on a phone is longer, narrower, slower, and touched by a thumb. Every one of those changes something you approved.

Why can't I just use the published mobile benchmarks?

Because most of the numbers you will be quoted about web performance and layout were not measured on phones, and the people who produced them say so.

The strongest experimental evidence in this field is the 2014 KDD paper by Ron Kohavi, Alex Deng, Roger Longbotham and Ya Xu, written out of Microsoft and LinkedIn, which reports isolated slowdown experiments at Bing covering millions of users. It is the source of almost every credible speed figure in circulation. It also carries a sentence that vendors quoting it tend to drop, and the sentence is theirs: the above experiments are all on desktop, and there is a lot to learn for mobile.

So when a proposal tells you that every hundred milliseconds costs you a fixed share of mobile conversions, ask where that was measured. The honest answer, for most of the figures in circulation, is a desktop search engine in the early 2010s.

The same caution applies to form research. The Baymard Institute has tracked checkout flows since 2012 and concluded that the number of form fields matters far more than the number of steps. That is a sound directional principle and it is worth acting on. It is also ecommerce checkout research from a company that sells UX research and benchmarking, conducted on buyers who already have a cart and a reason to persist. It is not a study of a plumber's quote form on a phone, and no such study with a disclosed sample and method exists.

Use the direction. Refuse the number. Then measure your own.

What breaks specifically on phones?

The same list turns up on site after site.

Sticky headers and cookie banners eating the screen. A fixed bar at the top, a promo bar under it, and a cookie notice at the bottom can leave a thin strip of usable page. On desktop all three look modest.

Tap targets too small or too close together. Menu items stacked tightly, two links in adjacent lines, a tiny close button on a popup. If a visitor misses twice they leave.

Forms that fight the phone. A phone field that opens a text keyboard instead of a number pad, an email field with autocapitalize on, dropdowns with thirty options in a scrolling wheel. Each one adds friction at the worst moment, on top of the fact that every field already costs you submissions.

Popups with no visible exit. On desktop the close button is obvious. On mobile it is often offscreen or hidden behind the fold, and a visitor who cannot close a popup does not browse around it. They leave.

Horizontal scrolling from a wide table or an oversized image. It makes a page feel broken even when everything else works.

Phone numbers that are text or images rather than links, which is the most common and most expensive one on this list.

What is the mobile gap costing me?

How should I test my own site on mobile?

On a real phone, on cellular data, standing somewhere other than your office. Ten minutes, once a quarter.

Do the full path a customer takes, not just the homepage. Arrive from a Google search result rather than typing the URL, because that is how they get there and it sometimes lands them on a different page than you assume.

Then complete an inquiry yourself, thumb only. Fill in the form. Watch what the keyboard covers. Count the taps. Submit it and see what you land on afterward, because the page after the form is usually wasted and you will only notice while you are standing in it.

Try it one handed. Try it in sunlight. Try it with your phone's text size set larger, which many older customers use and which breaks more layouts than anything else on this list.

Write down every moment you hesitated. You know this business and you still hesitated, so a stranger stopped there.

What about tablets and older devices?

Tablets are usually fine, because they get the desktop layout on a big enough screen. They are rarely the problem.

Older and cheaper Android phones are a different story, and they are well represented among local customers. Slower processors, smaller screens, older browsers. If you can borrow one for five minutes, that test is worth more than any emulator.

The accessibility settings are the underrated case. Larger system text, higher contrast modes, and reduced motion all change how your page renders, and a layout built with fixed heights tends to break under them. The people using those settings are often the ones with the most money and the least patience.

If you are about to start a rebuild, put device testing in writing at the beginning. It belongs in what to prepare before a web project starts, because "works on mobile" means nothing as an acceptance criterion and "completes an inquiry one handed on a small Android phone over cellular" means everything.

Where does the mobile problem sit outside your website?

Two places, and both are worse in 2026 than they were two years ago.

Sterling Sky's 2026 field research, using Jepto and Places Scout data, documents AI powered local packs that are mobile only and US only, appearing on roughly 7% of tracked keywords. They show one or two businesses instead of three, and they carry no call buttons at all. Across 322 markets, 88% surfaced fewer unique businesses in the AI pack than in the regular three pack.

The same research tracks the shift to paid on mobile specifically. Local pack ads went from about 1% of mobile reports in early 2025 to about 22% by December 2025, and Local Services Ads went from about 11% to 31% of tracked queries.

Neither of those is a layout problem and no amount of thumb testing fixes them. They are the reason a mobile conversion rate can be flat while your mobile inquiries fall, and they belong in the same conversation as your tap targets.

What to do this week

Do the ten minute field test today. Real phone, cellular, offsite, arriving from a search result, completing an inquiry with your thumb.

Fix the two worst things you find before you touch anything else. In most cases that is an unlinked phone number and a form field that opens the wrong keyboard, and both are quick changes.

Run the six step device split with your own analytics and write the dollar figure down.

If mobile is much lower and mobile is most of your traffic, you have found where your inquiries went. That comparison only exists if your events are set up properly, which is why conversion tracking that survives a website change is worth doing before the next redesign rather than after.

Then repeat the test after every change to the site. Mobile breaks quietly and usually during someone else's update.

If your mobile traffic converts poorly and is also low quality, look at what brought them. Ad copy that filters out the wrong customers fixes a problem no layout change will.

Questions about how your site behaves on a phone?

Email me at eric@seod.com with your website address and the one action you most want a mobile visitor to take. I will run through it on an actual phone, complete that action myself, and send you the list of every place I hesitated or had to try twice.

It takes me about ten minutes and there is nothing attached to it. If it all works cleanly, the reply will say so and name the next thing I would look at.

There is more on conversion in the library.