by Jerry Johnson

You know the old saying: something that sounds too good to be true...is.

Webcam eye tracking has been touted as a sort of redeemer: you don't have to pay those ugly meanies at Tobii, SMI, ASL, etc. those exorbitant prices to track the eye with expensive cameras, extra infrared lighting, fancy monitors, and elaborate software.  Easy as you please, you can just use the camera embedded in your computer and voila! -- eye tracking results at Wal Mart prices. 

Well, Aga Bojko has done us all a great service by conducting practical tests with webcam eye tracking and exposing these fallacies.  Her results are here.

It's not so much a scathing expose as a simple reminder that you get what you pay for.  As Aga points out, webcam eye tracking can indeed detect the gross regions where the respondent is focused, but not much else.  If that's what you need, go for it.

Most of us need more, and more precisely.  She rightly finishes by saying "it's a nice quick-and-dirty option to have handy" (or words to that effect), but note that the list of limitations is quite long!  For example, low frame rate, poor accuracy, head movements harm results, static pages only, short sessions required, and so on.

More time in market will help.  For the moment, webcam eye tracking is a work in progress.

Eye tracking technology has evolved rapidly over the past dozen or so years so that studies employing eye tracking within a true 3D VR environment have become a reality.  Surprisingly, many eyetracking studies still employ first generation technology. 

Here's a brief review of how eyetracking has evolved: 

First generation - simple eye tracking using a static image with no interaction between the respondent and the image being viewed. There is little or no use of virtual reality technology.  Useful output is restricted mainly to findability metrics.  

Second generation - utilizes VR to provide a more realistic shopping experience but requires two separate sessions on the part of respondents, the first session is a VR shopping study followed up by a second session, the eye tracking component, that measures what the respondent is looking at.  The two-part nature of this study make it nearly impossible to obtain a single integrated view of show the respondent is shopping.  Attempts to overlay the two datasets are performed manually and are subject to analyst interpretation.  Output includes gaze trails, dwell times and incidence of views along with separate tables of items purchased or rejected.

Third generation - integrates eye tracking and virtual shopping into a single session that is being driven by the respondent in a dynamic VR environment.  The key distinction that true integration brings to shopper marketing studies is that robust statistical analysis can be applied to the data being collected. Gaze trails have now been replaced with regression models that can be used to project the findings of the studies into nationally reliable sales numbers.

The latest developments in technology are seeing new biometrics being captured to augment the insights gained from eye tracking.  For example, brainwave measurements captured at the same time with eye tracking can pinpoint the portions of the brain that are stimulated during key periods of VR shopping.  Researchers can now measure how emotionally involved the respondent is during the shopping experience which may lead to the ability to manipulate the very nature of the decision-making process itself! 

Eye tracking technology has evolved to the point that even different types of shopping styles have now been identified.

The eye track gaze of respondents in a recent study was submitted to sequential pattern analysis; the same type of computer analysis done to map the human genome or in a fictional sense to populate Jurassic Park.  This analysis revealed at least four (4) different gaze trails that corresponded with different shopping styles:

1. Spontaneous
2. Task Relevant
3. Preoccupied
4. Manipulative

Initial findings have shown that these styles impact which stimuli are more effective leading to the possibility of coordinating multiple stimuli within the same aisle to achieve maximum impact  on sales. The results were made possible because of the capture of continuous eye gaze data within an interactive VR shopping environment that was controlled by the respondent.

Recently released US Consumer Expenditure Survey data confirm what eyetracking studies have been telling marketers for some time.  Our Nimbus eyetracking studies revealed a widening gulf between the shopping patterns observed between younger generations and their older counterparts over the past 5 years.  The eyetracking data revealed differences in dwell times and rejection rates among the different age cohorts.  Not only were there differences in what they purchased but there were differences in price sensitivity and the amount of time spent on considering a specific product before either purchasing or rejecting.  Recent releases of the CES data for 2009 reveal the impact of those shifts in shopping behavior on the distribution of common household purchases.  Some categories of household goods look to be on a declining trendline.  Meanwhile, other newly emerging products are siphoning off the discretionary spending of Millennials.  The problem is that the decision-making process of Millenials has changed and new ways of reaching the younger generations with product information are needed.

Clients often ask whether the results from a virtual study reflect actual shopping behavior in a real store.  During a recent oral care study, we measured a 97% correlation between dollar value of purchases on a brand-by-brand basis made by respondents buying in a virtual store versus reported industry sales.  The Pearson Correlation Coefficient was significant at better than 95% confidence level for results summed up across more than 13 brands represented in the study.  This high degree of correlation meant that all sorts of different stimuli could be tested for effectiveness in a virtual reality study with confidence that the results could be applied to real marketing strategies.

Mobile device usage in China has surpassed the US and indications are that penetration rates will continue to climb faster in China than either the US or Europe.  Mobile device users are far more likely to access the Internet and do so more often than other internet users.  Understanding these mobile device users is going to pose a major challenge to marketers since they tend to ignore traditional surveys.  It stands to reason that any research needs to make use of the same technology platforms. 

Is mobile eye tracking better than tracking from a computer monitor?  There's debate.

Most mobile eye tracking solutions provide analog output in real time.  In other words, researchers and observers can "see what the respondent is seeing" in the moment.  That's great for qualitative insight, but what kind of precision or quantitative reliability does it provide?

Not much.

On the other hand, eye tracking from a computer monitor in virtual reality provides real coordinate data, precise time coding, and in some cases fixations on special areas of interest, like a brand name on a label.  So far so good.  Is there any downside to eye tracking from a computer monitor?

Some people claim the virtualizations are clumsy and unreal.  But with each new year that critique fades, as the virtual reality simulations are approaching the real world much better.  The odds seem to be in favor of tracking the eye from a monitor.

Maybe there's somebody out there who sees it differently?