Eyetracking Study - How Query Intention Changes the Use of Search Results

[tedster]

I just read an interesting eye-tracking study: User behavior in SERPs. Eye tracking study July 2010 [dynamical.biz]. This article was translated from the original Spanish version, so it is information that many English speakers may not have seen.

The big surprise for me was not that different intentions meant different eye tracking behavior - although there is some meat in there too. The big surprise was that the snippet attracted a higher number of fixations than the title element did - and this was for EVERY intention category.

The study looked at these four query types: informational, navigational, transactional and multimedia. One big takeaway for me - there's a reason Google Webmaster Tools gives us quality reports on our meta descriptions, and it's a stronger data-driven reason that I would have guessed!

[jimbeetle]

And check out the heat map and gaze plot for the transactional query. Assuming it was similar to the [barcelona restaurant] example, it's striking that the local results were mostly overlooked.

[tedster]

Good catch, Jim. Maybe that's partly the reason behind this test [webmasterworld.com]

[aristotle]

Here are the results for informational transactions:

Title: 37% of fixations, 34% of fixations time

Snippet: 51% of fixations, 53% of fixations time

URL: 12% of fixations, 13% of fixations time


I'm a little confused about the meaning of "fixation". Suppose a surfer takes a quick glance at the title, just long enough to read it. Does this count as a fixation? If not, then I think the above data may be misleading, and may understate the importance of the title.

[Robert Charlton]

Interesting study, though I'd like to see more explanation about methodology, assumptions, and interpretation of data.

What the article doesn't make clear (and the graphics are shown too small to interpret) is the order of viewing, and how fixation time and number of fixations are parsed from the data.

Fixations are shown in milliseconds and are reported in fractions of milliseconds to two decimal places in the charts, way too precise IMO for this kind of experiment. There probably should be some rounding off for experimental error, with a +/- error range indicated.

It's also hard to know whether to attribute the greater number of fixations and fixation time for snippets, eg, simply to the greater number of words. If you considered, say, fixation time per word to be important, it's possible that snippets might actually be underperforming titles, since on Google, at any rate, snippets are generally at least twice as long as titles. That said, I feel that descriptions are very important for conversions. I'd like to see a stronger test to back that up.

[jimbeetle]

I think the above data may be misleading, and may understate the importance of the title.

Maybe as in a well-constructed title encourages searchers to read the snippet?

Then, as in Robert's observation, it simply takes longer to read the snippet.

It kind of just reinforces what we've been saying for years: title and snippet/description are very important and work hand-in-hand to get the click.

[CainIV]

This signals the inevitable fact that Google will eventually develop further nuances to what it refers to as its "SERPs" pages as they look to use this data to increase conversions on areas that generate revenue. :)

[mercedesP]

While I haven't done a detailed reading (yet), I can't see how they can make comparisons between searcher intentions and their eye-tracking, when the "distribution" of search results elements are soooo diferent according to the query. i.e. for transactional query the first "organic" results are from Google maps. This "block" of results is clearly defined by a blank space element, indented and reinforced by an image .... but this "informational block" is in direct eye-catching competition with top and side adwords results, while the results for the other two query intentions barely show any adwords competing blocks.

So, if on any given intentional search results page, the diverse blocks of elements are not very, very similar how can the results be trusted (or to what degree)?