CTR Up, Catalogue Shrinking
Decide what you would do from the brief alone, including whether you would change anything at all. Everything below it is available, but the exercise stops working if you open it first.
The homepage recommender has been retrained weekly for six months. CTR is up steadily (illustrative). Merchandising complains that new products never appear, sellers in the long tail are leaving the platform, and users in a survey describe the homepage as "the same ten things". Leadership sees the CTR chart and does not see the problem.
Retraining daily instead of weekly "so new products get picked up faster", and adding a recency feature. Daily retraining closes the loop seven times faster; the recency feature gives new products a brief bump until their lack of clicks — on impressions they never received — ranks them down again. CTR rises slightly, which looks like the fix working, and the catalogue narrows faster than before. The approach is intuitive because it treats "new products do not appear" as a freshness problem, when it is an exposure problem.
Read this even if you are confident. It is here rather than behind a button because it is the answer most teams actually ship, it passes review, and its cost arrives weeks later when the labels do.