There are a few companies in the market that offer a participant-free, predictive eye tracking method. Most of them claim a 75-90% correlation with a real eye tracking data. This data seems quite promising, but we weren’t able to find any convincing studies that would support this claim.
We wanted to learn how the attention distribution looks like on a Dell homepage. To conduct the tests we used EyeQuant (predictive eyetracking method), Attensee (behavioral attention mapping) and a real eye tracker - Tobii T60 with 5 seconds exposure to homepage screenshot, tested on 15 participants (conducted in 2 days time). Attensee study was set to 9 seconds exposure, conducted on 60 participants and took an hour to conduct.
Attensee and Tobii heatmaps show high correlation, while the predictive eye tracking results seem to be far from the truth in this particular example. Eyequant simulation predicts that people will pay a lot of attention directly to the Dell logo, whereas real participants were mostly focused on the promotional content: 'Get a $279.99 laptop...' and button 'Dell preferred account'.
Moreover, the prediction shows some fixation on 'Laptops and Ultrabooks' area, but in fact, more spotlight was put on the first section with the offer for Dell Club members.
Also, EyeQuant’s prediction did not foresee that the boat displayed on one of the laptops, would steal users attention in such a fair amount. Neither did it reflect that users were interested in 'Do more with Dell' section in the footer.
In this analyzed case, we’ve found no support for the claim of high correlation between the real & predictive eye tracking. Without any doubts, Attensee is the winner here and these heatmaps speak for themselves.