Lighthouse CI vs Managed Monitoring: Build vs Buy for Agencies
An agency tech lead inherited twelve client repositories, each with a Lighthouse CI workflow that passed on Friday and missed a checkout regression by Monday. Reporting meant exporting JSON, pasting scores into slides, and hoping the account manager could explain why /cart turned red while the homepage stayed green.
Lighthouse CI is the right starting point: merge protection, reproducible lab numbers, no SaaS invoice while one engineer owns the scripts. The friction shows up around client five or fifteen. Per-repository YAML drifts. Preview URL patterns break after a host move. Flaky runners train teams to re-run until green. Field CrUX drift after deploy never enters the pipeline.
Where DIY usually breaks for agency portfolios:
Repository-centric configs vs portfolio-centric clients
Muted flaky jobs that pass without trust
No client-ready narrative from CI artefacts alone
No scheduled production coverage beside merge gates
Managed PageSpeed monitoring does a different job: schedules across many URLs, retains history, fires budget alerts, and gives sponsors a portfolio view. Keep Lighthouse CI as a gate. Layer managed monitoring for production evidence. Watcher is one option for multi-tenant dashboards, discovery, budgets, and lab plus CrUX without babysitting YAML for every client.
