Weijia Ren

I led the design and deployment of a real-time, on-platform, inline survey measurement system to evaluate search quality on Facebook. I designed cognitive interviews and controlled experiments to validate survey wording, UI placement, and cooldown intervals, proving that query-level "usefulness" at the first pagination point provided the most sensitive signal without degrading engagement. By integrating survey responses with session-level search behavioral log data, I established "usefulness" as a primary guardrail metric for product launch. This live diagnostic system enabled engineering teams to quickly uncover and fix misleading, clickbait query suggestions that inflated clicks but severely harmed user satisfaction.

I led end-to-end quantitative research to evaluate and optimize the integration of Meta AI into Facebook Search following its high-profile launch. Initial launch metrics revealed ambiguity, with high interaction volume but mixed user sentiment. I designed a multi-wave survey to diagnose user behavior, segmenting interactions into intentional AI queries and accidental triggers driven by existing search habits.

My statistical modeling proved that most of the Meta AI usage in Search was "unintentional", which drove low perceived search usefulness and satisfaction among users. These findings directly guided leadership in changing the product direction from the Meta AI threadview to the blended view. I adopted intent-aligned trigger mechanisms, establishing foundational UX quality guardrails for generative AI across Meta.