Screenshot statistics that don't hold up

Some numbers about visuals and screenshots get repeated on blog after blog, and almost nobody checks where they came from. We traced four of them. None holds up the way it is usually quoted. The real case for good screenshots doesn't need them.

"We process visuals 60,000 times faster than text"

Where it comes from. A two-page 1997 brochure from 3M, Polishing Your Presentation.

What the source says. Jon Schwabish traced the claim to that brochure in 2015 and found no references in it at all. In 2012, Alan Levine started looking for the study behind the number and offered a reward to anyone who found it. His 2014 post says the $60 prize still stands unclaimed.

What to cite instead. If you want to say pictures help people learn, cite Richard Mayer's multimedia research: across 13 experimental comparisons, people who learned from words and pictures did better on a transfer test than those who had words alone, with a median effect size of d = 1.35. It is real and measured, and it is about short lessons, not software docs.

"80% of knowledge bases are out of date"

Where it comes from. A 2025 blog post that cites a survey of 224 contact centre professionals by CallCentreHelper.com.

What the source says. In that survey, 19.1% rated their knowledge base "very accurate" and 43.1% "reasonably accurate". The 80% is everyone who did not pick "very accurate", which includes the people who called theirs reasonably accurate. It measured how accurate staff felt their knowledge base was. It didn't measure whether anything was out of date. The same post also cites one source for a 23% figure twice: once as the rise in tickets from a poorly kept knowledge base, and once as the drop from a well-kept one.

What to cite instead. For how common outdated docs are, the GitHub Open Source Survey 2017: 93% of respondents had seen incomplete or outdated documentation. Say that it is a perception survey and that it doesn't separate incomplete from outdated. For screenshots specifically, cite the VS Code study: 210 issue reports about outdated screenshots. More on that study.

"The first three app-store screenshots drive most installs"

Where it comes from. App store optimization agencies. One 2026 article says "the first three screenshots are the most influential for conversion performance".

What the source says. That article cites no data for the claim. Its "research consistently shows" line about first impressions cites nothing either.

What to cite instead. Apple's own guidance, which says less and says it plainly. You can have up to 10 screenshots, and depending on orientation, up to three can appear in search results. Apple's advice is to lead with your best features and show what people can expect. That is a good reason to make the first three count. It is not a measurement of installs.

"People spend 7 seconds on a store listing" and "only 13% scroll screenshots"

Where it comes from. A 2017 guest post by an app marketing agency on Phiture's ASO Stack blog.

What the source says. It says "the average time people spend on a store listing is 7 seconds" but never says how that average was calculated. Its other figures come from a handful of apps, and it reports iOS and Android separately: 0.5% of iOS users watched a video, against 2–4% on Android. Those are observations from the apps it tested, not a benchmark for store listings in general. The "13% scroll screenshots" figure comes from the same post and carries the same limits.

What to cite instead. Nothing precise. We found no public, described measurement of how long people look at store listings. Apple's guidance above is the honest version: put your best screens first.

What to cite instead

The real evidence is less dramatic and more useful. Buyers want to see the product: 35% of SaaS sales sites don't show the UI clearly enough, according to Baymard. Buyers notice mismatches: 69% of B2B buyers in a Gartner survey reported inconsistencies between a sales organization's website and what its sellers told them. And outdated screenshots are common enough that VS Code had 210 issue reports about them. None of that needs a made-up number.

What the research says about product screenshots collects all of it, each finding labelled by how strong it is.

Sources

Every page below was opened on 13 November 2026, and each number on this page was checked against it.