Vector Search easy
A marketplace runs ANN search for the top 10 listings, then removes listings outside the user's city. Users in small cities often get only 0 to 2 results. What is the best fix?
- A Increase the number of HNSW graph links per node so that each query explores more neighbors.
- B Switch the distance metric from cosine to dot product so that local listings score higher.
- C Apply the city filter during the search (pre-filter or filter-aware traversal), not after it.
- D Re-embed all listings with a larger model so that location is captured in the vector itself.
Answer: C. Apply the city filter during the search (pre-filter or filter-aware traversal), not after it.
Post-filtering with a **selective** filter starves results: most of the top 10 get thrown away. Pre-filtering or filter-aware traversal only considers matching listings, at the cost of more complex index logic. Tuning the graph does not solve this, because the nearest neighbors are still mostly in other cities. Post-filtering works fine when most results pass the filter.
One system design question a day, with spaced repetition so the ones you miss come back.
Get the appRelated questions
- You are building an IVF index on 20 million vectors. A teammate asks why the build needs a 'training' step before any vectors can be added. What is the reason?
- An HNSW index serves a marketplace where about 30% of listings are updated or removed each week. Each update is applied as a delete plus an insert, and deletes mark tombstones. Over a month, recall@10 drifts from 97% to 89% and p99 latency rises at the same ef_search. What is happening, and what should you do?
- Users delete documents in your primary database, but the deleted documents keep appearing in vector search results for hours. The vector index is rebuilt nightly. What is the best fix?
- Your embedding model outputs unnormalized vectors whose magnitudes vary widely. A teammate wants to switch the index from cosine similarity to dot product because it is cheaper to compute. What is the consequence?
- You are setting an SLO for a new ANN-backed recommendation search over 50 million vectors. A product manager asks for 100% recall@10 with single-digit millisecond latency. What is a realistic response?
- A B2B SaaS app serves 500 customers, and every query searches only one customer's documents. One large customer runs heavy batch queries that slow down search for everyone. Which sharding approach fits best?