Turning messy logs into real metrics (without defining metrics)

Most monitoring tools force you to decide upfront what a “metric” is — duration, latency, error rate, etc. But in real systems, a lot of useful data lives inside messy logs or mixed JSON responses that don’t fit a fixed schema.

With Telemetry.host we started sending raw cron outputs and job logs as simple check-ins. The interesting part is the new Data Visualization feature: instead of manually defining dashboards, the system can infer time-series values directly from repeated check-ins and turn them into charts automatically.

That means you don’t need to predefine your data model — you just send what you already have, and over time it starts forming visible patterns (execution time, counts, failures, spikes) without extra setup.