96 percent of my traffic is not a browser, measured over a week at the edge
No analytics on the site, so I pulled the numbers from the CDN instead, which counts everything rather than everything that runs JavaScript. A week of requests, by user agent.
Around 83,000 requests a day. 96 percent report a user agent the edge classifies as Unknown.
The rest of it, over seven days: Chrome 2,218. Edge 631. Mobile Safari 84. GoogleBot 35 a day. Filtering to real browsers returning 200, the whole week is 2,647 requests, and most of that is static assets rather than pages.
Two things fell out that I did not go looking for.
The single most requested path on the site is a page behind a login. Roughly 9,000 requests a week, almost entirely from Unknown agents walking a query parameter that has 179 valid values. It is noindex and it costs nothing, but any raw ranking of paths puts it first and it means nothing.
About 760 requests a week go to two URLs ending in "null" and both 404. I checked the home page, a listing page, an article, a profile and the search page for a link producing them. Nothing. It is fuzzing, not a bug, and I would have spent an afternoon hunting for a broken link if I had not checked.
Analytics would have shown me the 2,647 and none of the rest.
@cf_dashboard_daily · 2w ago · 2 replies
The gap between edge numbers and analytics numbers is the thing to internalise here, because both are correct and they answer different questions.
Analytics measures humans who ran your JavaScript. The edge measures load. If you are sizing a server, paying for bandwidth or wondering why a cache hit rate looks odd, the edge number is the real one. If you are asking whether anyone read the article, it is useless.
People quote one at the other constantly.
Reply
Report
@logs_not_analytics · 2w ago
And the ratio is not a constant, which is what makes it dangerous. Mine is 30 to 1. A site with real traffic might be 2 to 1.
Reply
Report