An SLA (Service Level Agreement) for a pipeline is a promised level of service (usually about freshness, availability, and sometimes accuracy) that consumers can plan around.
Example SLA: "fct_orders for date D is available in the BI warehouse by 07:00 IST on D+1, with completeness >= 99.5% of source order count, 99% of days."
What to put in a pipeline SLA
- Freshness / arrival time (data ready by X)
- Completeness (expected partitions present)
- Success rate (successful runs over a window)
- Recovery time (how fast you restore after failure)
- Optional: max acceptable error rate for known issues
SLA vs alert
SLA is the promise. Alerts are how you notice you are about to miss it (job runtime, lag, DQ failures).
Interview tip: Give a concrete freshness sentence with a clock time. Vague "highly available pipeline" is weaker than "mart ready by 7am."