← All questions
Retrieval & EmbeddingsHard· Indexing Strategy· System Latency· Data Churn

How often do you re-index, and how did you pick that nu...

Determining the re-indexing frequency requires a quantitative approach rather than an arbitrary schedule. You must derive the optimal interval (T*) by balancing the build cost of the index against the incoming query rate and the rate of data churn.

Additionally, you must account for operational constraints. Specifically, you should identify the point (ϕ) at which beam inflation causes query latency to exceed your p95 budget. Finally, you must ensure that your re-indexing strategy respects the mandatory erasure deadline. The most restrictive of these factors dictates the final re-indexing frequency.

Code playground (Python)