Drift Watch
Included on every plan
Continuous distribution monitoring. Every column you watch is scored on every run against its own fourteen-day shape, per segment, on twelve signals.
- Population stability index, calibrated so one threshold works across every column
- Segment-level scoring — catches a collapse in one market while the total holds
- Day-of-week seasonality modelled separately, so Mondays don't page anyone
- Before-and-after distributions attached to every alert
Schema Guard
Team and Scale
Contract enforcement at the boundary. Declare what a table promises; Driftline blocks the run that breaks the promise instead of reporting it afterwards.
- Contracts as code, versioned in your repo next to the model that reads them
- Fails the run in CI, not in production, when a producer breaks a consumer
- Type coercion and nullability changes caught before merge
- Every consumer of a column listed on the pull request that changes it
Replay
Team and Scale
Reconstruct any day's baseline and ask the only question that matters in a post-mortem: when did this actually start?
- Rewind any column's distribution to any run in the retention window
- Bisect a drift back to the run and commit range that introduced it
- Compare two arbitrary runs side by side, per segment
- Export the timeline into the incident doc as a figure, not a screenshot
Module availability
What comes with what.
| Module | Starter | Team | Scale |
|---|---|---|---|
| Drift Watch | yes | yes | yes |
| Segment-level scoring | — | yes | yes |
| Schema Guard | — | yes | yes |
| Replay | — | 30 days | 400 days |
| In-VPC scoring engine | — | — | yes |
Start with Drift Watch.
Free on three pipelines, no card. Add the other modules when the failures change shape.