Techniques for evaluating software systems are provided, including measurement of performance ratings of features of an app-under-test are described. The evaluation may include analysis of a video recording of the user interface of the app-under-test, where the video analysis may include comparing the changes over time of co-located pixel in the video to produce a performance rating, for example a speed index. The results of multiple such tests can be compared by comparing the performance rating produced during execution of different tests.