Live Engine
Select Topic
easyAb Testing And Experimentation
A team launches an A/B test for a new recommendation model. They run the test for 3 days and observe a 4.2% CTR lift in the treatment group (p=0.04). A data scientist declares the experiment a success and recommends shipping the new model. A senior engineer asks: "Did you check for the novelty effect?" What is the novelty effect, and why does it threaten the validity of a 3-day experiment?