The first version of our local acquisition loop crawled a client site every five minutes. Cloudflare showed tens of thousands of requests. That looked energetic. It wasn’t evidence that more customers were finding the company.

Most of the volume came from our own tooling repeatedly checking pages, sitemaps, redirects, and assets. The crawl was useful as a deployment test, but the chart mixed internal proof work with humans, search crawlers, random bots, redirects, and errors.

Classify before you count

We now separate human-like traffic, verified search crawlers, Valen Systems tooling, known bot noise, unknown automation, and status-code traffic. Cloudflare remains valuable for seeing whether pages are served, which paths are hit, and where errors occur. It’s supporting evidence, not ranking truth.

  • Search Console measures Google impressions, clicks, queries, and average position.
  • A SERP provider measures where a URL appears for a controlled query and location.
  • Analytics and call systems measure human behavior and conversion events.
  • Cloudflare measures requests at the edge, including requests no customer ever made.

The useful question isn’t whether traffic rose

The useful questions are specific. Did the right page gain impressions for the right query? Did its position improve? Did the search result earn clicks? Did the visit produce a call, form, booking, or useful conversation?

A request chart can help explain a failure. It can’t by itself prove demand. Once we stopped asking one system to answer every question, the reporting became less exciting and much more honest.

Delivery evidence still matters

The correction wasn’t to throw Cloudflare analytics away. Edge data can show a spike in 404s, a redirect loop, a cache miss pattern, an unexpected hostname, or a crawler hitting a page family that shouldn’t be public. Those are important operational facts.

We now label the evidence before using it. A successful edge request can prove that a release is reachable. A Googlebot request can prove discovery activity. Neither event becomes an impression, a customer, or a ranking without evidence from the system that measures that outcome.