Field notes / Thresholds customers notice, not graphs
Thresholds customers notice, not graphs
Performance dashboards are comforting, yet teams still miss regressions that show up in human language first. This article ties k6 scenarios to checkout and account flows we saw in anonymized ticket tags.
We compare two threshold styles: strict wall times versus budgeted errors per thousand virtual users. Each includes a short YAML snippet and a note about when Grafana annotations help.
Finally, we document how to roll thresholds forward after a legitimate slowdown so alerts stay trustworthy.