Screens, Research and Hypertext

Powered by 🌱Roam Garden

No One Reads PDFs

This has been true since at least 2014. Seriously.

If your content is inside a PDF, it's probably not going to get read.

In 2014, the World Bank conducted a study of its website traffic to determine how people were using its reports. What they found is that 1/3 of its reports had never been downloaded. Indeed, only 13% of all World Bank reports were downloaded more than 250 times.

Graphic showing downloads of World Bank PDFs.

These numbers are pretty typical

Of course we shouldn’t expect a World Bank report to have millions of downloads. They’re niche products, so maybe 250 downloads is pretty good.

Except it really isn’t.

According to the Bureau of Labor Statistics, there are around 19,000 economists working in the United States. So 250 downloads represents just over 1% of the World Bank’s core American audience. That’s a pretty low bar.

Part of the reason PDFs are unloved—they don't really accommodate the way most people read on screens. Or, more to the point, a PDF is hard to scan.

Internet users love to scan text.

That’s been true pretty much since the web really took off. In 2006, web usability pioneer Jacob Nielsen measured the eye movements of 232 users as they looked at thousands of web pages, a process known appropriately enough as “eye tracking.” Nielsen discovered that users read text in patterns that roughly resemble the letter F.

Image from a Nielsen study showing eye tracking patterns for three different page types. In each case, the tracking pattern roughly resembles the letter F.

Since then, study after study has confirmed that users scan text online. Another Nielsen study, for example, found that users read only about 20% of the text content on a given page. Usability expert Gerry McGovern discovered that only 1 in 15 users could locate an item they were specifically looking for when it was placed in the middle of a page. And Chartbeat—a web analytics company that is widely used by media companies—says that most users will scroll through only about 50% of an article before leaving.

For more context

Think tanks are built for producing the wrong sorts of things.

What to read next

PDFs just don't work with the way people find content.

Other items of interest

It's not just ordinary readers who have difficulty with reading things online.

What even is a standard think tank output?

What might a format optimized for research outputs look like?