The platform that built the pipeline is the one watching it.
Monitoring attaches the moment a pipeline deploys — nothing to install, nothing to wire. It knows the shape of what it built, so it can tell you which stage failed and which numbers downstream are now stale. It reads what the pipeline reports. It writes nothing to your data.
Last run · 02:00
Incident
fct_settlements failed at build. A column changed in raw_payments.
Root cause
Why did this fail, what does it affect, and what can I do about it?
What it catches.
Deterministic, and honest about being deterministic. No guesses, no scoring, no alert storm — the events below are facts about your pipelines.
The run that did not happen
A job that fails tells you. A job that never starts tells you nothing. The missed schedule is caught on its own and raised as an incident.
Output going stale
Tables that should have been refreshed and were not are marked stale, and so is every KPI that depends on them — before anyone opens the dashboard.
A column that changed
When the source schema no longer matches what the pipeline was built against, the drift is recorded and reported with the run.
The stage that failed
A live view of the pipeline's stages, with the failed one marked and the path downstream from it shown as affected.
It will tell you why. It still will not act.
When a run fails, an agent reads the brief the pipeline was built from, the pipeline itself, and the run history, then writes a plain-language account: what broke, the most likely reason, which KPIs and dashboards are affected, and the steps you can take. A hypothesis is labelled as one. Every action stays yours to take.
- What broke, in the words of your own tables and models
- The most likely cause, with a guess named as a guess
- Which KPIs and dashboards downstream are affected
- The steps to resolve it — for a person to run, not an agent
The limits, up front.
- Pipelines built here
- Monitoring covers pipelines this platform built and deployed. Pipelines built elsewhere are not watched today.
- It does not write
- Monitoring reads what your pipelines report and writes nothing back to your data. That holds in the next release too.
- It does not own your on-call
- There is no SLA transfer and no paging rota. Your team stays on call; this makes the first five minutes shorter.
Watch a pipeline you built this week.
In a pilot, the pipeline we build together shows up under watch on the day it deploys.