web analytics 2.0
For many years eye tracking has been limited to a controlled, virtual environment, making precise data collection and analysis relatively simple:
- the computer screen doesn’t move,
- participant movement is limited,
- and stimulus presentations on screen are generally consistent across participants.
The biggest challenge researchers face is correcting for head movement (turning sideways or leaning forward) and managing point of gaze data on sometimes unexpected dynamic stimuli (pop-ups, animated ads, video, scrolling, etc.).
Nonetheless, it was relatively simple to track samples of 100 or more and evaluate their viewing pattern as they looked at on-screen presentations. Dynamic backgrounds posed a challenge, but as long as all participants looked at the same stimulus, it wasn’t too big of a deal to analyze this. It was time consuming to identify the areas of interest (AOI) for a background that is constantly changing, but once they are identified, everyone’s data can be run against the same AOIs (as identified, for example, in a television commercial). The output is then the same as for static backgrounds: precise dwell time and fixation information for areas of interest that can be aggregated across all participants. And this can easily be plotted on a static image of the background for visualization.
This type of testing was great for websites, TV commercials and software usability testing to name a few, but was less realistic when evaluating stimuli such as shelf displays, package designs, magazines or products in hand. This posed a problem; a golden rule in user research is to test real users in real environments. Eye tracking participants looking at virtual shelf displays on a computer screen just isn’t the same as eye tracking them as they look at actual packages on a shelf (though this certainly is up for debate).
Thus came the shift to mobile eye tracking - recording a person’s point of gaze as he or she is moving about in a real, 3-D environment. This could be selecting magazines off of a shelf and reading them, moving about a store selecting products off of a shelf, or interacting with signage at a baseball game. The same technology could apply and has been modified for this type of testing, but there have been a number of obstacles along the way involving either the hardware or the software:
- Hardware: While hardware has been reduced in scope and size, a fast processor with a big enough hard drive is still needed, as are at least two cameras (one for the eye, one for the scene). The connection between computer and cameras needs to be wireless or all be so lightweight that it’s easily portable. The cameras have to be secured to the head in a way to limit any shift between camera and eye. How to manage these limitations effectively?
- Software: The calibration process and recording the data is much the same. The challenge is in the follow-up analysis. How do you identify regions and analyze point of gaze data when the background is constantly changing and is unique for each and every participant? How can you identify AOI rapidly and accurately with such variability? How can we aggregate data across participants when the stimuli varies so significantly? Can it be done to match the stable scene analysis that we are accustomed to, or is it necessary to make a significant paradigm shift in how we approach mobile eye data?
Different manufacturers and engineers have approached and managed these obstacles in unique ways. With the next few blogs I plan to explore that a bit further. I can only provide information based on personal research and experience, and am eager to hear more about experiences others have had. It’s an exciting new technology; although it has been around for years, there have been and continue to be substantial developments that bring this research approach more to the forefront.
