This one took me a second to get traction on.
Start by defining offline ranking metrics (e.g., NDCG, MAP, precision@k, recall@k) and online business KPIs (CTR, conversion rate, revenue per session). Explain how precision and recall trade off in ranking, then describe how to validate offline improvements with online A/B tests and reconcile discrepancies through deeper analysis.
Pro tip: Emphasize that offline metrics are proxies, not goals—always tie them to online business impact and be ready to explain why an offline win might not translate online (e.g., position bias, feedback loops).
List metrics like precision@k, recall@k, NDCG, MAP, and MRR to evaluate ranking quality against labeled relevance data. Explain how precision and recall apply: precision measures how many retrieved items are relevant, recall measures how many relevant items are retrieved.
Identify online metrics such as click-through rate (CTR), conversion rate, add-to-cart rate, revenue per session, and user engagement. These reflect actual business impact and user behavior.
Discuss how increasing recall (retrieving more relevant items) may lower precision (more irrelevant items shown), and vice versa. In ranking, the trade-off is managed by position and threshold; e.g., top-k precision vs. overall recall.
Describe how to validate offline improvements with online A/B tests. If offline metrics improve but online KPIs don't, investigate factors like position bias, presentation, or metric misalignment. Use online metrics as the ultimate arbiter.
Continuously refine offline metrics to better correlate with online KPIs. Consider using counterfactual or unbiased offline evaluation methods to bridge the gap.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.