CrackML by @ml.with.umang
Interview questions / ML Fundamentals
ML Fundamentals interview question

Retrieval vs Ranking

Explain the difference between retrieval and ranking in a large recommendation/search system.

mediumconceptEvidence 41/1001 source reportLinkedIn

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

1
Give the core idea

Retrieval maximizes candidate recall cheaply over a huge corpus; ranking spends more compute to optimize precision/order on a much smaller candidate set.

2
Explain how it works

Two-tower/ANN methods are common for retrieval, while richer cross-feature models such as DLRM, Wide & Deep, GBDTs, or cross-encoders fit ranking.

3
Compare alternatives

Evaluate retrieval with recall@K and coverage, ranking with NDCG/MRR/CTR-style outcomes, and always measure end-to-end impact.

4
State failure modes + validation

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.

01Definition
02Mechanism
03Trade-offs
04Failure modes
05When to use

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

Mechanistic understanding, assumptions, trade-offs, and whether you can turn a definition into a model decision.

Likely follow-up questions

What assumption makes this approach work?
When would you choose the strongest alternative instead?
What production or data failure mode changes your answer?

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.