With summative usability tests, products are broken down into and tested on key user tasks. Performance is captured on each task like the percentage of users who successfully finish the task (i.e. completion) and how long it took them (i.e. time on task). These performance scores can be compared with competitive products, or previous versions of the same product.
We apply inferential statistics to this data in two ways. First, confidence intervals are calculated, which indicate the the precision of our sample measurements, as they relate to true population values. Second, we use hypothesis testing validate differences in usability performance between two or more products. We test to see if these differences are "statistically significant", that is, usability differences exist at the population-level, and are not just due to noise in the sample.
Ещё видео!