A controlled experiment that randomly splits users between two versions of a feature so you can measure which one moves the metric you care about, instead of guessing.
Around 2009 Google ran an experiment across 41 gradations of blue for its search and ads link color, with revenue as the deciding metric. The winning shade was credited with about $200M in annual ads revenue, and the experiment became the most-cited cautionary tale of design-by-data, especially after lead designer Doug Bowman quit over it.