The traffic that GA4 swore didn't exist

The traffic that GA4 swore didn't exist

A mid-size B2B SaaS site had a problem: GA4 showed 40k monthly sessions, but the hosting bill screamed three times the load. The clue lived in the server logs, not the JavaScript.

GA4 only counts visitors who run the tag. Server logs count every request, including the bots GA4 never sees. We parsed a week of raw access logs and found 61% of crawl hits hammering faceted-filter URLs no human ever visited.

The change: a few Disallow rules and a canonical cleanup on the filter pages.

Crawl waste on those paths dropped from 61% to 19% of total requests, and the origin's p95 response time fell from 880ms to 410ms.

Use GA4 to understand humans. Use server logs to understand everything else hitting your box. They answer different questions, and the bandwidth bill is written in the logs.