Hacker News new | ask | show | jobs
by babs474 4548 days ago
A single data point is too small to reason about. But lets say there was an ad campaign that was blocked at a high rate, or even just placed poorly where nobody was seeing it or something. Our scheme would assign those impressions a low value, ie it would come out that cranking impressions or lowering impressions isn't effecting conversion rates anywhere else, so why not turn the ad off.