Explain How Ads Get Personalized: Signals, Targeting Models and Privacy Limits

You are interviewing for a product manager role at Conde Nast, and the interviewer asks a technical question: how would you tailor the ads people see so that each person gets ads suited to them? Explain the main approaches to ad personalization, what data each one relies on, how those inputs turn into a decision about which ad a particular person sees on a particular page, and where the approach is limited. ```hint Person versus page Some signals describe the content on screen right now and others describe the reader over time. Keep them apart: they need different data, behave differently for readers you cannot identify, and are treated differently by consent rules. ``` ### Clarifying Questions - Are the ads shown on our own websites and apps, where we control the content and know our signed-in readers, or are we buying placements on other publishers' sites? - Are advertisers mainly buying awareness (reach among the right audience) or direct responses such as clicks, sign-ups and purchases? - Roughly what share of readers are signed in or have agreed to personalized advertising, and which privacy laws apply in the markets we serve? - If more targeted ads raise revenue but readers find them intrusive, which matters more for this business? ### What a Strong Answer Covers - The distinct personalization approaches (contextual, declared and demographic, behavioral and interest-based, advertiser audiences and lookalikes, predictive response models) and the data each one needs. - How a single ad request is served: identifying the reader and the page, filtering eligible campaigns, scoring them by expected value and choosing one within a tight latency budget. - How personalization is evaluated offline and with a controlled holdout, using advertiser, revenue and reader-experience metrics. - Consent, privacy law, browser and mobile tracking limits and sensitive categories, plus a fallback for readers who are anonymous or opted out. - Failure modes such as cold start, feedback loops, excessive frequency and ads that feel invasive. ### Follow-up Questions - Most of your readers never sign in, and their browsers block third-party cookies. How do you still personalize ads for them? - A personalization change lifts click-through rate, but complaints about ads that feel creepy also rise. How do you decide what to do? - How would you show an advertiser that paying more to target a specific audience beat running the same campaign untargeted?