The world is being quietly rearranged by people who write very long documents.


The title they went with Simpson's Paradox in Behavioral Curves: How Aggregation Distorts Parametric Models of User Dynamics Noisy translates that to

Apps are misreading how much you use them. By 3x to 5x.


It turns out that when companies measure how much people use their apps, the numbers are systematically wrong. The way they combine data from many users makes it look like people engage three to five times more than they actually do, distorting everything from ad targeting to drug trials.
Companies have been building their entire understanding of user behavior on a flawed assumption. The numbers they use to decide what to recommend, how to target ads, or even how to dose drugs are systematically overestimating engagement by three to five times. This means they are likely misallocating resources and making incorrect predictions.
Watch for companies that build recommendation systems or advertising platforms to acknowledge this distortion and adjust their models.

If you insist
Read the original →