Two Case Studies: Production ML infrastructure and Recommendation Engines - ML 072
Adventures in Machine Learning
00:00
How Would You Evaluate Without an AD Test?
I wouldn't do an AB test in this case. There's really no point. It's only a potential positive upside of launching this. You can also do something where you forecast what the engagement will look like in, let's say, three key metrics and see if the actual metrics are above or below the forecast. Another alternative is often business data exhibits weekly seasonality. So in this case, there's probably more sessions during the week than the weekend. And I would build that solution in the rapidly disabled this without having to do a production code push.
Transcript
Play full episode
