Design a Feature to Improve Ad Relevance for Users

To improve ad relevance at Meta, we must bridge the gap between implicit behavioral signals and explicit user intent. My proposed feature is an enhanced 'Ad Feedback Loop' integrated directly into the feed interface. First, when a user hovers over an ad, a subtle tooltip offers two primary actions: 'Hide Ad' for immediate suppression and 'Tell Us Why' for granular feedback. The latter opens a minimal modal asking the user to select from specific reasons: 'Irrelevant Product,' 'Too Frequent,' 'Offensive Content,' or 'Already Purchased.' Second, this explicit signal feeds directly into our real-time bidding engine. If a user selects 'Already Purchased,' the system immediately suppresses similar creatives across the platform, preventing wasted impressions and reducing annoyance. Unlike passive signals, this provides high-confidence training data. Third, we must mitigate survey fatigue. We will implement a frequency cap, allowing only one explicit feedback interaction per ad category every seven days. Success will be measured by an increase in Ad Relevance Diagnostics scores and a reduction in negative feedback rates, balanced against any potential drop in CTR due to higher filtering standards. This approach aligns with Meta's mission to build community trust while optimizing advertiser ROI through precise targeting.