Netflix Product Sense: AI-Generated Previews Matched to Each Member's Taste

You are a product manager at Netflix. How would you use generative AI to give each member a trailer made for them, with the aim of raising engagement? Explain what engagement should mean here, which members you would focus on, how the tailored trailers would be produced and where they would appear, and how you would measure whether they work. ```hint Start from the moment of choice A trailer exists to help a member decide whether to press play. Find the members for whom that decision goes worst today, then ask what a cut made for them could show that a single trailer made for everyone cannot. ``` ### Clarifying Questions - Should engagement mean more viewing hours, more titles started, fewer visits that end without anything played, or better subscriber retention? - Does this cover the whole catalog (licensed and original films, series, games) or start with titles Netflix produces, where the rights are simpler? - Where would the trailers play: the preview that starts automatically on the home screen, the title's detail page, or channels outside the app such as email? - May the AI create new footage, voices or narration, or only select, order, trim and caption footage that already exists in the title? ### What a Strong Answer Covers - A precise definition of engagement that rewards satisfying viewing, not just clicks on a preview - A chosen member segment with concrete decision pain, ranked against the alternatives - A concrete personalization design: which signals drive it, which parts of the trailer vary, where it appears, and where people stay in the loop - Success metrics measured against a control, with guardrails that catch misleading cuts and lost viewing - Risks specific to this idea: misrepresenting a title, spoilers, licensing and talent rights, and the cost of generating video at scale ### Follow-up Questions - Tailored trailers raise play starts, but more members also stop watching within the first few minutes. What do you conclude, and what do you change? - A studio that licenses a popular series to Netflix refuses to let its footage be re-edited. How does your plan adapt? - How would you keep generation and serving costs below the value created, given how many members and titles Netflix has?