11 May 2026

Stop calling the job done when the API is still thinking

Why automation success metrics that ignore upstream response time mislead ops reviews.

Most schedulers mark a run successful when the process exits without an exception. That is a useful process health signal. It is not a measure of API automation performance.

If your bot waits twelve minutes on a partner endpoint, then writes a row and exits cleanly, the dashboard still smiles. The customer who needed that row an hour earlier does not.

We ask teams to record three timestamps: work accepted, first dependency resolved, and business outcome committed. The gaps between them tell you whether to tune the runner, renegotiate the partner SLA, or stop batching so aggressively.

Application analytics for automation only becomes trustworthy when those gaps are visible every week—not only after an incident review.