Observability Practical By Samson Tanimawo, PhD Published Jul 7, 2026 4 min read

Defining SLIs for Data Pipelines

Pipeline SLIs differ from request-response SLIs. The three dimensions that matter, the metric definitions, and the alerting that catches drift.

Freshness

How old is the data the pipeline produced? Freshness SLI: 95% of partitions arrive within 30 minutes of source.

Critical for downstream that depends on recency.

Did the pipeline process all expected records? Completeness SLI: 99% of expected daily volume.

Drops indicate upstream issues, transformation bugs, or schema mismatches.

Sample-based: pick a small subset; verify outputs match expected. Correctness SLI: 99% sample agreement.

Hardest to measure. Worth it for high-stakes pipelines.