Eye tracking makes it possible to investigate website user experience issues in detail. It’s discrete and informs other important research, giving additional insights when used in our clients’ conversion optimisation projects.
For example, it helps with understanding what the analytics platform and other data sources indicate as friction points. Eye tracking enables us to drill down to what the problem could actually be by monitoring and measuring real-time activity.
It doesn’t solve the problem (although it might confirm something that we identified during heuristic analysis as a quick fix); but, it does provide an additional layer of important information.
We combine this with other data and findings to help develop and define a hypothesis and design the treatments, and we can also use it to establish whether we’re on the right course.
Previously we’ve shown how eye tracking can uncover what people are and are not looking at on a website (and for how long) and how we can use that information to identify friction points and inform our design updates.
But what else can the technology tell us that goes beyond what a visitor is looking at?
Well, one of the powers of the best eye tracking technology is in how they build their model of the eye. The most common technique is pupil centre corneal reflection (PCCR), which essentially uses cameras to capture the reflection of light on the cornea and in the pupil of the eye.
This data undergoes numerous calculations to model eye and gaze position but it also enables a calculation of pupil size (for a more in-depth understanding of modelling eye position and pupil size, see references 1-9).
Knowing the size of the pupil and, importantly, how it changes over time can be used to study emotional responses to stimuli. In conversion optimisation, we can apply this to a website visitor’s experience to identify actions and behaviours that are initiating some kind of response (positive or negative).
As pupil dilation is an unconscious response, it can provide indicators of a state of mind that you would not necessarily know from where they are looking, or even what they might be telling you in a testing session.
By monitoring how pupil size is changing over the course of a lab session, we can begin to understand what elements and tasks on the website are prompting both positive and negative responses in the visitors’ experience.
So what has an impact on pupil size changes? Research has shown that it can be an indication of the following:
In optimisation, how do these translate to onsite behaviour—and what might be indicated in this data?
Let’s take a look at some examples to illustrate the changes that you might see in pupil size when on a website.
Example 1 – Winning versus losing engagement
The video above shows an edit of a £10 winner game on the lottery and while my reactions may not seem overly excited, what does my pupil size tell us?
First, you need to look at the below comparison of the change in pupil size during that game (orange) compared to a losing play of the same game (blue).
We can see that over the game—as the possibility of winning (and winning more) becomes apparent, pupil size increases and peaks around the winning moments—my engagement increases when being rewarded (three peaks associated with three winning moments).
Compare this to the losing game when pupil size—and my engagement and interest—decreases relatively as the game progresses and the prospects of winning are fading.
The above example is an indicator of a potentially big positive effect—a monetary win is likely to generate a pretty positive response but your website can’t be giving away money all of the time (though as we’ll mention there are ways to reward your customers in other ways).
Example 2 – Shopping frustration
See the below graph taken from a purchase test on a clothing website:
While the task itself was actually relatively straightforward—there were no extreme points of website functionality impacting the experience, for example—when analysing pupil size over the task, we noticed a concentrated period of increased pupil dilation.
What did that tie into? When reviewing the session what we found was that this period correlated with the time spent on a product page reviewing the product images.
And why would that cause an increased pupil size? Well by this stage in the test, the visitor had found a product they liked and were looking to potentially purchase but were struggling to make a decision as to whether it was right for them.
So, this was a period of increased cognitive load and frustration that was highlighted by pupil size.
Whilst this may indicate there was a problem, you still have to follow up to find the reason why it was happening.
In this particular case, the subject had selected a specific product but they did not have sufficient information to make and justify a purchase – there were no “model” images of the item being worn to provide the key pieces of information the subject wanted (how it fit & hung, length etc).
The visitor had already invested time into finding a suitable product and then had to put in further effort trying to find and process the key decision critical information and when they couldn’t it caused frustration.
This particular issue may have been overlooked if not highlighted by the pupil size data.
Combined with your other research, pupil size helps identify the blockages in your own website conversion and can help you to assess whether your test variations are tackling them.
But, it also can help identify what is providing a positive response to your website and you can use this to help employ techniques such as using “micro-rewards” and moments of “surprise and delight” that keep your customers engaged and enjoying the journey to conversion and beyond.
This is an important consideration as monitoring pupil size can help identify both positive and negative responses to website tasks—the best optimisation is not only focused on removing the negative barriers to conversion but should also seek to maximise the positive and enjoyable engagements your customers have on your website.
By eliminating the causes of negative responses and increasing the positive, you can ultimately maximise your onsite conversion by improving the complete customer experience.
 Tobii, Eyetracking whitepaper, https://acuityets.files.wordpress.com/2010/07/systemrecommendations_txseries_tobiistudio.pdf
 JT Wang, “Pupil dilation and eye-tracking”, http://homepage.ntu.edu.tw/~josephw/Pupil%20Dilation%20and%20Eyetracking_rev.pdf
 JT Wang, et al., “Pinocchio’s pupil: Using eyetracking and pupil dilation to understand truth-telling and deception in sender-receiver games”, American Economic Review, https://people.hss.caltech.edu/~camerer/pinocchio2.pdf
 M Pomplun, S Sunkara, “Pupil dilation as an indicator of cognitive workload in human-computer interaction”, http://www.cs.umb.edu/~marc/pubs/pomplun_hci2003.pdf
 M Pomplun, et al., “Using pupil size as a measure of cognitive workload in video-based eye-tracking studies”, http://www.cs.umb.edu/~marc/pubs/pomplun_sunkara_fairley_xiao_draft.pdf
 T Partalaa, V Surakkaa, “Pupil size variation as an indication of affective processing”, Int. J. Human-Computer Studies, 59, 185–198 (2003), https://www.researchgate.net/profile/Veikko_Surakka/publication/222659277_Pupil_size_variation_as_an_indication_of_affective_processing/links/0fcfd50cadbc323e87000000.pdf
 TP Flavio, et al., “Discriminating the relevance of web search results with measures of pupil size”, CHI 2009 ~ New Input Modalities, https://www.researchgate.net/publication/221516350_Discriminating_the_relevance_of_web_search_results_with_measures_of_pupil_size
 J Klingner, et al., “Measuring the task-evoked pupillary response with a remote eye tracker”, http://graphics.stanford.edu/~klingner/publications/MeasuringPupillaryResponse.pdf