How do you handle cold start in a recommender — and how does the recommender change the data it will be trained on?
Answer it out loud before you open anything. The value of the flags below is in comparing them to what you actually said — including whether you asked about the data before naming a model.
A news app has a homepage recommender trained on clicks. New articles get few impressions and few clicks, so they rarely surface; new users see the globally popular feed. A product review found the homepage increasingly narrow: the same dozen topics, and a click-through rate that rose for two months and then fell.
What it is really testing
Whether the candidate can address cold start for both items and users with mechanisms rather than "use content features", and whether they see the feedback loop: the model chose what was shown, clicks happened only on what was shown, and the next model learns from those clicks. The declining CTR after an initial rise is the loop closing.