Eye Tracking

Eye tracking is a research technique that records where a person looks, in what order, and for how long, using hardware that measures the position and movement of the eyes. In interface research it produces two main artifacts: gaze plots, which show the sequence and duration of fixations for an individual, and aggregated attention heatmaps, which show where a group of participants looked most.

The method's genuine value is that it captures behavior people cannot accurately report. Asked what they looked at, participants describe what they remember noticing, which is a poor account of the rapid, largely unconscious scanning that determines whether an element is seen at all. Eye tracking reveals whether a key message was fixated or skipped, whether people read a heading or scanned past it, and whether attention followed the path the design intended. It is particularly informative for questions about whether something is noticed, which is different from whether it is understood.

Well-known findings from this literature have shaped everyday practice. Research from Nielsen Norman Group described the F-shaped scanning pattern typical of dense text pages, where readers attend to the first lines and the left edge and progressively less of each subsequent line. Banner blindness, the learned avoidance of regions and treatments resembling advertising, was documented the same way. Studies of directional cues showed that people tend to follow the gaze direction of a person in an image, which is why photographs of a model looking toward a headline outperform ones looking away.

The method has significant practical constraints. Laboratory-grade equipment is expensive and requires calibrated, supervised sessions, so sample sizes are small and participants are in an artificial setting where their attention is more deliberate than it would be at home. Webcam-based alternatives have made the technique far more accessible and can be run remotely at larger scale, but their spatial precision is considerably lower, and they support conclusions about broad regions rather than about specific elements. Predictive attention models trained on eye tracking data are cheaper still and useful for early screening, but they estimate visual salience rather than measure real attention.

Interpretation is where most misuse occurs. Looking at something is not the same as understanding it, being persuaded by it, or acting on it, and an attention heatmap does not indicate whether the attention was productive. A region that draws long fixations may be compelling or may be confusing, and only asking the participant distinguishes the two. Eye tracking is therefore most valuable in combination with think-aloud protocol and task performance measurement rather than as a standalone output, and its aggregate visualizations should not be confused with the click and scroll heatmaps produced by analytics tools, which measure entirely different behavior.

A practical alternative that captures much of the value at a fraction of the cost is asking participants to describe what they noticed immediately after a brief exposure, combined with observing where they move the cursor and how they scroll. Neither is a substitute for measured gaze, but for the most common question in commercial work, whether a key element is being seen at all, the cheaper methods usually produce the same conclusion. Reserving equipment-based measurement for cases where the answer genuinely depends on precise fixation data, such as comparing two positions for a critical disclosure or evaluating whether a legally required notice is actually read, keeps the cost proportionate to the decision it informs.

Given cost and complexity, it is normally reserved for high-stakes surfaces where attention allocation genuinely determines the outcome, such as a homepage, a key landing page, or a product detail template. In that context it is one method among several within a user research programme, and its findings are most useful when they feed directly into hierarchy decisions in product design rather than being presented as standalone visual evidence.