Dating Apps Rank Popularity, Then Call It Compatibility

The word arrives like a measurement. Compatibility. Matching features showed up in the early 2000s to narrow the pool, and they let the sites charge higher fees while holding attention longer; as a review in the Harvard Data Science Review writes, these algorithms did not have to work perfectly to be profitable. 1

What a recommendation actually ranks turns out to be closer to popularity. A study of more than 240,000 users of a major dating platform found a user's chance of being recommended rose significantly with their average attractiveness score, and the researchers' model showed that giving popular and unpopular users equal odds lowered the platform's revenue. 2 The same review covers the other half of the effect: in OkCupid's own testing, 30 percent matches were raised to 90 percent and 90 percent matches lowered to 30, and suggesting a match worked about as well as being one. 1

The study usually quoted as proof that app matches do better — marriages that began online were slightly less likely to break up and slightly higher in satisfaction — was commissioned by eHarmony, and its author list includes eHarmony's scientific advisor and the former director of eHarmony Laboratories. 3

Compatibility is what happens in the room, between two people. That part stays unmeasured.