Freshness
Also called Freshness
How up-to-date the information a system can retrieve actually is.
Think of it like
Milk in the fridge — the question isn’t whether it’s there, it’s whether it’s still good.
Example
A pricing assistant that indexes the catalog nightly is a day fresh; if a price changed this morning, its answer is confidently wrong until the next refresh.
How it actually works
Retrieval systems are only as current as their last update. Freshness measures the lag between reality and what the index knows. High freshness matters for prices, inventory, news, and status; it matters less for stable reference material. The tension is cost — near-real-time indexing is expensive, so teams pick a refresh cadence that matches how fast the underlying data actually moves.
For product teams
Decide per data source how stale is acceptable, then set the refresh cadence to match — not everything needs to be live.
For engineers
Track index lag against source-of-truth update time; choose refresh frequency by each dataset’s volatility.
Related
- Staleness — The problem freshness prevents.
- Index Refresh — The process that restores it.
- Knowledge cutoff — Why a model alone can’t stay current.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome