The 'I optimized it once' problem

A site that passes Core Web Vitals today can fail them in three months if someone adds a heavy image, a new third-party script, or a plugin with render-blocking CSS. Without continuous monitoring, regressions go unnoticed until Google penalizes them or users complain.

Two kinds of monitoring: lab and field

Lab: tests run under controlled conditions, like Lighthouse. Useful for catching problems during development, reproducible, and easy to automate.

Field: real data from real users on their own devices and connections. It's what Google measures for ranking. You get it through RUM (Real User Monitoring) tools.

Continuous monitoring tools

Google Search Console: the free starting point. It shows real Core Web Vitals grouped by URL. Checking it monthly reveals trends and emerging problems.

Lighthouse CI: integrates Lighthouse into your CI/CD pipeline. Every deploy automatically runs an audit and can be blocked if the score drops below a threshold.

WebPageTest: detailed tests from multiple geographic locations. Ideal for in-depth diagnosis of specific regressions.

Alerts: when to get notified

Set up alerts so you hear about it when something changes. Recommended thresholds: alert if LCP exceeds 3 seconds for more than 20% of users, if CLS exceeds 0.15, or if average TTFB exceeds 1 second.

Performance budget: set limits before they're exceeded

A performance budget is a set of limits the team agrees not to exceed: maximum page weight, maximum number of requests, minimum Lighthouse score. Defining it before the project starts and automating its enforcement prevents performance debt from piling up.