- Data quality
When it comes to our eye tracking systems, we hold ourselves to the highest standards of data quality. We also share our performance test procedures so our customers can learn how we test and report our quality metrics.
As a researcher, it is incredibly important that you can trust the quality of your eye tracking measurements and make informed equipment decisions for your projects.
We use extensive test methodologies to measure and report eye tracker performance under optimal and variant conditions, including typical data collection scenarios.
The test reports detail our accuracy and precision metrics and specify experimental procedures, so you can understand and interpret eye tracking specifications.
This methodology provides information on the eye tracker performance under ideal conditions, as well as the eye tracker’s robustness to different changing conditions.
The test specification includes ideal measurement conditions, head positions in the track box, large gaze angles, and different illumination conditions. Subjects are tested in a controlled lab room setup and use a chinrest to control for distance from the screen and visual angle relative to the eye tracker.
Accuracy and precision are calculated from recordings with a high proportion of valid data. Both metrics are presented based on monocular (dominant eye) and binocular data.
Precision is also measured on artificial eyes to differentiate between system noise and precision of human eyes.
This new methodology is used to understand eye tracker performance in less constrained conditions, as well as how participant characteristics influence eye tracking data.
Data is collected over a larger population of subjects and typically in an indoor office or university environment. Subjects are tested without a chin rest – allowing them to move naturally and freely rotate head in front of the screen.
We are proud to share below the first published report, showing results for Tobii Pro Spectrum with over 400 subjects, and Tobii Pro Fusion with over 500 participants.