The 60-second answer
Retrieval maximizes candidate recall cheaply over a huge corpus; ranking spends more compute to optimize precision/order on a much smaller candidate set. Two-tower/ANN methods are common for retrieval, while richer cross-feature models such as DLRM, Wide & Deep, GBDTs, or cross-encoders fit ranking.
Build the answer in this order
Retrieval maximizes candidate recall cheaply over a huge corpus; ranking spends more compute to optimize precision/order on a much smaller candidate set.
Two-tower/ANN methods are common for retrieval, while richer cross-feature models such as DLRM, Wide & Deep, GBDTs, or cross-encoders fit ranking.
Evaluate retrieval with recall@K and coverage, ranking with NDCG/MRR/CTR-style outcomes, and always measure end-to-end impact.
Tie the explanation to a concrete production failure mode or decision criterion.
A useful interview mental model
This is the shape of a strong answer—not a script to memorize.
Senior-level signal
- Senior answers discuss where business rules, freshness, exploration, and safety filters sit between stages.
- Call out training-serving mismatch: optimizing a ranker cannot recover candidates the retriever never surfaced.
What the interviewer is really testing
Likely follow-up questions
Common weak-answer patterns
- Reciting a definition without mechanism or assumptions.
- Claiming one technique is always better without a data regime.
- Stopping before failure modes, validation, or deployment implications.