Tracking breaks quietly. Monitoring is how you find out.
The failure mode of server-side tagging is not the build. It is month four, when a developer ships a checkout change, an endpoint starts 502-ing, or a CMP update quietly flips a consent default. The container keeps returning 200 and nobody notices until the quarter is reported. This is the part of the category that is usually sold as a status page, if it is sold at all.
SLOs, not a dashboard
Ingest availability, destination success rate, p95 latency, match quality and consent coverage each get a target and an error budget. When a budget starts burning, someone hears about it — not a widget that turns amber on a page nobody has open.
Buffer and replay
Events in flight when a destination goes down are buffered and replayed when it comes back, so an outage at the far end costs you latency rather than conversions. The replay is logged and countable, which is how you tell the difference between a fixed incident and a quiet loss.
- Event volume alerting against a seasonal baseline, not a fixed threshold
- Every destination watched for non-2xx, with the response body captured
- Match quality and identifier coverage tracked on a rolling window
- Consent mix alerting, so a CMP change cannot silently reshape your data
- Container error rate, cold starts, memory pressure and cost burn
- Schema drift detection — a parameter that stopped arriving is an alert, not a surprise
- Teams with nobody to route an alert to. Monitoring that fires into an empty channel is theatre
- Setups with no server-side pipeline yet — there is nothing to watch until there is
Tell us what your stack looks like.
You will get a reply from the engineer who would do the work, usually within one business day, and it will contain an actual opinion rather than a calendar link.