A single speed test is enough more often than monitoring vendors admit. The mistake is using that one test as a standing substitute for a portfolio promise.
Honest jobs for a one-off run
You are launching a marketing page tomorrow and need a lab sanity check. A developer claims a tag manager change fixed LCP and you want a same-day confirmation. A client forwarded a competitor’s score and you need a comparable screenshot for the call. Geography is part of the ticket and you need a real-browser pass from a specific region. In those moments GTmetrix, PageSpeed Insights, or WebPageTest-class tools are the right weight of process.
Where spot checks quietly fail
Spot checks fail when the SOW implies continuous care. They fail when five people each keep a private ritual (“I run mobile on Fridays”). They fail when the only URL anyone ever tests is the homepage while paid traffic lands on templates that never enter the ritual. They fail when CrUX is amber for a month and nobody noticed because the last lab card in Slack was green.
None of that means you should monitor every URL in existence. It means the decision is about which URLs need dates, not whether diagnostics are obsolete.
A simple rule of thumb
If missing a regression for two weeks would embarrass you in a client meeting, those URLs belong on a schedule. If the work is a discrete investigation with a clear end, run the diagnostic and close the ticket. Many agencies need both: a short monitored set for money pages, and GTmetrix when the monitored set says something is wrong.