Eyetrack Study May Surprise You

By Thomas Young, MBA

The Poynter Institute's decade-old "Eyetrack" study showed how consumers read a printed newspaper page. [In May 2000], results were released for the online equivalent. The results are remarkably different, confirming what we all know, but don't always practice - that the Web is a different medium where rules about user behavior have little to do with what works in the print medium.

But this Eyetrack study may shock some of you. One of its principal findings is that news Website users tend to look first at and look most intently at text, glossing over photos and images in search of meaningful textual information. This is 180 degrees away from the print Eyetrack study, which found that newspaper readers typically get drawn first to photos on a page, before letting their gaze goes off in search of text.

Camera-heads

This new Eyetrack study, which has its roots in the work of Stanford University communications professor Marion Lewenstein started some four years ago. It is a joint project between Stanford and Poynter, the Florida-based journalism education and research center. Poynter joined in about two years ago, and the research moved from videotaping test subjects as they read news Websites, to outfitting them with tiny head-mounted cameras that could track where on a computer screen the subjects were looking.

The study team worked in three U.S. cities: Chicago; St. Petersburg, Fla.; and San Francisco-San Jose. A total of 67 test subjects were recruited to spend time reading online news sites as they normally would, while their eye movements were tracked and recorded into a database. All subjects described themselves as online news consumers, and they were free to look at whatever sites they desired for as long as they liked. They brought along their own list of browser bookmarks, so it was possible for them to have a near-normal online news surfing session.

In all, the test subjects surfed 211 unique news sites, viewing nearly 6,000 pages, over an elapsed time period of 40 hours. Some significant numbers are that the average news surfing session was 34 minutes, and the average number of news sites visited during that time was six. (On one extreme, one test subject visited 19 different sites.) So, if you have any illusions that Web users are spending lots of time poring over your site, think again.

A key measurement in the study is an eye "fixation," which is defined as when a subject's eye focuses on a single point for at least 1/10th of a second - enough for the brain to comprehend a bit of information. In all, the researchers recorded 608,063 eye fixations. That's a lot of data, and the study team will have plenty of work ahead of it still to pull useful conclusions after further analysis.
Initial analysis

What you'll find on the Poynter Website now is an analysis of the study team's initial findings, which are substantial, yet still only scratch the surface in understanding Web news consumers' behavior.

Lewenstein says that perhaps the most significant finding is the importance of good, solid, useful text to the online news reader. Photos and graphics aren't looked at anywhere near as much as text, "so have plenty of text. Don't obfuscate it with lots of fancy graphics," she says.

Poynter fellow Andrew DeVigal, one of the leaders of the study, explains that the test subjects exhibited behavior that surprised the researchers. Analyzing subjects' fixations on news sites' home pages, it became apparent that they immediately looked for headlines, news briefs, and captions. They only looked at photos afterward - and sometimes not until they had gone to another page and then returned to the home page. Then they would fixate on a photo on the page.

(The subjects surfed the Web on a high-bandwidth connection, which brought photos on screen quickly. On a typical slow modem connection, text would appear on a news site before photos, of course.)

Graphics fared worse. While 64% of photos on a typical page were looked at, only 22% of graphics were. Compare that to news briefs, which were looked at 82% of the time, and article text, which was looked at (not necessarily read intently) 92% of the time. Interestingly, banner ads fared quite well. While DeVigal had expected to see dismal fixation rates on banner ads, they actually were seen 45% of the time - and the average fixation period was 1 second, which he says is enough for the Web user to comprehend the brand message.

(Here's a useful tip. DeVigal says that given the brief typical fixation on any banner ad, if a banner is animated, every animation frame must have the brand name included. Not to do so can mean the consumer will look at the ad but not comprehend who placed the ad. Tell that to your advertisers.)

Serious text is best

It makes sense that text is what online news users focus on most. Unlike a medium like TV or radio, the Web is used primarily by people who are seeking something specific. Pleasure browsing is less common than specific information seeking and gathering. Eyetrack test subjects seemed to search text elements of a page looking for news, and apparently regarded photos and graphics as of secondary importance.

During post-recording sessions with subjects, they were asked questions about what they saw during their Web news surfing time. Lewenstein says one thing that came out was that the online news users preferred straightforward headlines to funny or cute ones. A common comment was that online site headlines were better than many print newspaper ones, because print headlines writers often try to be cute and witty - the result being that the headlines don't do as good a job of quickly explaining what a story is about. The lesson: Be straightforward and efficient with headlines. Recognize that getting too cute may actually turn off online users, who just want to quickly discern if a story is worth reading.

Likewise, in the interviews, subjects commented on how important "good writing" was to them, and that they look for it. An interesting aside occurred in Chicago, where the Website of the Sun-Times was visited more often by test subjects than the site of the Tribune. Lewenstein recalls that some subjects commented that they thought the writing in the Tribune was better, but disliked the Tribune site's interface (which has since been redesigned) and so they went to the Sun-Times site instead - because it was easier to navigate.

Forcing them to scroll is OK ... really!

Another interesting finding is that online news users "do know how to scroll," says DeVigal. For articles on a news site, if a Web user has clicked to get to the page, chances are high that the article will be read - because the headline or blurb that led the user to click to the page gives enough information to the user that he/she knows in advance it's something worth reading. The Eyetrack study showed that about 75% of article text was read. By comparison, print studies typically show that 20%-25% of any article's text gets read, on average. Print readers have less vested in any given story, because they haven't done anything pro-active to get the article in view (other than turn the page, not knowing what to expect next).

In light of his findings, DeVigal says he thinks it's perfectly appropriate to have home pages that require scrolling to see all the content. CNN.com's home page, for example, is very long, but test subjects who visited the page had no problems scrolling to find what they were interested in. What works about what on the surface may seem like an awkward home page is that content is effectively grouped into categories and subcategories. It's easy for a site user to focus in on desired content.

Photographers won't like this study

What do we do with our news Web pages if users largely ignore photos and graphics? The study authors do not suggest that pages need to be devoid of art, but they do suggest that photos be used appropriately. Part of the problem, of course, is that on-screen photos generally are small (so as not to take up too much screen real estate), and the typical screen resolution is only 72 dpi. Those two factors make for photos that are not particularly compelling, and thus users ignore them.

DeVigal suggests that better photo editing is required on the Web. If you just take a photo as delivered by The Associated Press and post it on your Website, you've fallen into the trap of thinking like a print publisher. Photos that go on your news site must be high-impact shots; routine or bland photos are certain to be ignored. Some photos that work in a newspaper or magazine need to be cropped tightly, down to their essence. A shot of a witness testifying before a Senate panel probably won't work if it's too wide angle, and perhaps should be cropped to nothing more than a mug shot of the witness for the Web, for example.

Running bigger photos to get them noticed is not the answer, says DeVigal. Running appropriately sized photos for a Web page that are dramatic and compact is. One problem with many news sites is that they don't have trained photo editors working for them; those staffers who do select photos aren't as skilled. The Eyetrack study indicates the importance of getting good photo editing skills into a news site staff.

And here's another photo-related tip. Be wary of running headlines within photos (as is a common design trick on news Websites). This study indicates that they might get overlooked.

As for editorial graphics, DeVigal thinks that there is still a place for them, but the most effective will be interactive graphics (such as those produced with Flash). The trick to getting Web users to look at them may be to integrate them into the content package, and not let them "stand out as an island" where they're more likely to be ignored.

Reason for optimism

Lewenstein says that this Eyetrack study has some findings that should please publishers of general news Websites. Researchers found that these online news users favored general news sites, especially local ones, over specialty sites; started their surfing sessions with general news sites; and returning to them often after visiting other news sites.

The study also confirmed that online news users still value serendipity. They seemed to use news sites in the same way they read newspapers - clicking to different new sites in search of "stuff they should know about" in the same way as turning the pages of a single newspaper. This would seem to indicate that past fears that online users would burrow into their own little world of personalized news and ignore the "larger picture" are unfounded, Lewenstein says.

The key difference is that online users have a wide array of news sources available to them and don't stay on one news organization's site for more than a few minutes, typically. That's something to which online news sites must learn to adapt.

(For more detail on the Poynter-Stanford online news Eyetrack study, see the Poynter Website at: http://www.poynter.org/.)