# Data Freshness SLAs and Lateness Windows
Status: public
Confidence: medium (0.725) (verified)
Last verified: 2026-06-02
Generation: ai_structured


## TL;DR

Freshness SLAs and lateness windows tell agents whether data is current enough to answer a question or trigger downstream work.

## Core Explanation

A dataset can have valid schema and still be unusable if it is stale. Agents should inspect source freshness checks, last successful run, expected schedule, watermark, deadline alerts, and allowed lateness before using the data as evidence.

Freshness should be judged against the purpose of the data. A daily finance report, a near-real-time fraud stream, and a weekly product analytics table need different freshness thresholds.

## Source-Mapped Facts

- Great Expectations documentation describes data freshness as validating that data is recent enough for its intended use. ([source](https://docs.greatexpectations.io/docs/reference/learn/data_quality_use_cases/freshness))
- Dagster documentation describes freshness policies as allowing users to define how up to date assets should be. ([source](https://docs.dagster.io/guides/observe/asset-freshness-policies))
- dbt documentation describes source freshness as a way to check whether source data is up to date. ([source](https://docs.getdbt.com/docs/build/sources#source-data-freshness))

## Further Reading

- [Great Expectations Freshness](https://docs.greatexpectations.io/docs/reference/learn/data_quality_use_cases/freshness)
- [Dagster Asset Freshness Policies](https://docs.dagster.io/guides/observe/asset-freshness-policies)
- [dbt Source Freshness](https://docs.getdbt.com/docs/build/sources#source-data-freshness)