Feng-GUI: "Visual Attention" Heatmaps
Feng-GUI is an interesting heatmap creation service. Unlike click-based heatmaps from Crazy Egg, FuseStats, and others, Feng-GUI creates heatmaps based on where it thinks the human eye would most likely be attracted. Eye tracking is something that designers have long used to measure the effectiveness of advertising, or design more usable web sites (among other commercial applications). But Feng-GUI doesn't use real eye tracking, which would require that humans look at each object being measured and would hardly scale very well. Instead, the site uses an algorithm that attempts to guess what a real human would be most likely to look at.
According to Feng-GUI, their ViewFinder algorithm is based on research from the field of robotics into how humans see and what sort of physiologic and neurological processes go on in our brains when we look at things. The algorithm than creates a saliency map. Saliency is a neuroscience term that describes how much an object stands out relative to its neighbors.
Feng-GUI has been around for awhile -- it was pitched on the TechCrunch forums to little reaction over a year ago -- but just recently the company released the results of a test that it says proves the accuracy of its algorithm. According to the company, Feng-GUI attention maps capture 70% of what traditional eye and mouse tracking report -- or, in other words, they are 70% as accurate. Because the work is being done by a computer rather than humans, one has to assume that it is cheaper.
The visual results of the test are impressive -- Feng-GUI does indeed seem to stack up well to traditional eye tracking. But because the methodology of the study, who did it, and how it was conducted wasn't released, it is hard to trust it.
There is, however, one major problem with Feng-GUI: I couldn't get it to work. The heat map generator on the Feng-GUI page timed out on every attempt to generate a map for ReadWriteWeb (it supposedly times out after 30 seconds, but RWW has loaded quickly all day), or any other web site I tried. Nor could I get it generate a map for a screenshot of ReadWriteWeb that I uploaded.
Google's Portuguese site as seen by Feng-GUI.
Regardless, I do like the idea of Feng-GUI. If the algorithm can be trusted, it has a lot of useful applications for designers and web site owners. The service could theoretically be used for testing early stage prototypes of new designs without having to pay people to perform usability studies or exposing half-baked new ideas to beta testers. Or it could be used by advertisers to figure out which display ad would attract the most attention on a specific web site before making an ad buy. Theoretically, sites could even use data from the service to charge more for a specific ad spot based on visual attention rather than higher click-thru rates (though selling advertisers on a robot's opinion might be a bit of a tall order).
What do you think of the idea of eye tracking vs. click tracking? What about algorithmic pseudo-eye tracking? Could you get Feng-GUI to work for you? Leave your thoughts in the comments below.
ShareRelated Entries
1 TrackBacks
TrackBack URL for this entry: http://www.readwriteweb.com/cgi-bin/mt/mt-tb.cgi/3285
Comments
Subscribe to comments for this post OR Subscribe to comments for all ReadWriteWeb posts
Would be good to see a comparison of the heatmaps produced by Feng GUI vs. heatmaps produced by real (i.e. non-algorithmic) eye tracking. Hard to tell how good the service is without this info (A video clip of real eye tracking is available here btw.)
Definitely need a comparison between real tracking and Feng in order to properly assess the worth of one over the other.
Interesting concept, how does it differ with other tracking sites that knows where we click etc. It may have its usefulness for specific functions but not overall I think.
We know alot about how the eyes and brain process vision these days and there are lots of ways to take these learnings and make them more available to website design teams.
Mouse tracking, alas, is truly a poor substitute for eye tracking:
http://alwaysbetesting.com/abtest/index.cfm/2007/4/29/Eye-Tracking-vs-Mouse-Tracking
But there is some correlation. There are also alternate ways to reveal the underlying drivers like our eye gaze simulator in a browser, Scrutinizer: http://about.stompernet.com/scrutinizer
Very interesting concept.
I think this has a very limited use since it only is an (70%) aproximation and not the real thing.
As most sites follow one of a few best practise patterns for navigation and content it really is not a big challenge to map those areas on a heat map.
What you really need eye tracking for is to check if secondary content is visible at all to the users. Does 1.2 em suffice for an h2 in the right column? Is the link to "support" really as prominent as we think?
We did an eye tracking study for our corporate site at my work a few years ago and it was really interesting but served more to confirm what we had done well than to give us surprises.
Also: It is important to follow up on the test with an interview of each test person immediately after the eye tracking session. Then you can understand where there were problems with the graphic design or the navigation design and where information design was the main obstacle.
I wrote an article about a similar concept two years ago, after seeing someone doing a presentation of "Feng-Shui and Graphical User Interfaces". It gives a little bit more info on how it works and how to do it yourself: http://www.ivy.fr/blog/index.php/2005/11/17/9-feng-shui-et-interfaces-graphiques (in French)
If they can perfect their algorithm more, it would be interesting to see if this becomes a tool in the web designers and usability designers toolbox. I know I would certainly run it against my website to see what the results were, even if it is only 70% accurate. I might not make major changes based on that, but it at least might open my eye to something otherwise not seen.
But that's just dreaming. I couldn't get it to load, either :-p
I am skeptical... would have to see it scientifically tested by impartial 3rd parties.
I've tried Feng Gui and the output compared to actual eye tracking had limited correspondence, but I am still hopeful. The images are here - http://usableworld.terapad.com/index.cfm?fa=contentNews.newsDetails&newsID=48379&from=list
The algorithms were developed off real eye data I understand and if Feng Gui continue to tune it with real data that would be good.
I invited the Feng Gui guys to a new facebook Group about eyetracking and some useful discussions are beginning. Please join us!
http://www.facebook.com/group.php?gid=8187313273
Interesting, but VERY slow... :(
For me this service has proven very usefull. You get instant feedback on design changes that you want to try out. Real user research is a whole different ballgame of course.
Tried this service but it didn't produce a heatmap which I would trust, as it just stuck a pink blob in the top left hand corner even though our design is more centralised. Still, if they can refine the system to be a bit more sophisticated, then it can defintely become a much more useful tool for web designing.