How do you run scripted browser monitoring reliably in production?

Start with one genuinely important user journey rather than trying to reproduce the entire frontend. Capture screenshots, logs, and traces whenever a run fails so you can tell whether the page is broken or the script is stale. Choose selectors intended for automation instead of relying on text or fragile CSS generated by the UI framework. Tools such as Playwright can help, but the bigger reliability improvements come from keeping the test focused and updating it as part of the deployment process.

Keep browser checks limited to a small number of critical user journeys and let API checks cover most of the system. Good candidates are login, one core transaction, and perhaps a workflow involving a third-party dependency that API checks would miss. Treat the scripts like production code: use stable selectors, review them alongside frontend changes, and keep them narrow enough to maintain. For alerting, page someone only when failures occur from multiple locations or line up with another signal such as elevated error rates or customer reports. A single failed run may just be a broken selector.

We use browser checks to represent the real customer experience, while the exact setup depends on our uptime and service-level requirements. A practical baseline is running the critical journeys from a few geographic regions every five minutes. For authentication, use a dedicated test account with no meaningful permissions and restrict it to only the workflow being tested. Never use a real customer or employee account, and handle any credentials through the monitoring provider’s secret storage.