Geek Question: Cloudflare Statistics Gone Wild!

This is a pure geek question about a problem I’m having with Cloudflare statistics. I use Cloudflare to speed up the website and block “threats.” It also delivers stats, which they claim are usually more accurate than any other log system.

Its regular traffic numbers have always been somewhat higher than other stats loggers’. Through most of 2013, the Cloudflare/Other Loggers ratio was 5.5 (SD=2.8; Pearson correlation of the two stat sets R^2=0.18: lots of variance and not an impressively high correlation).

Since January 1, the ratio has been almost exactly 100 (SD=63, R^2=0.31). The correlation increased but the scatterplot showed this was entirely due to a few outliers; it’s meaningless. Two days ago Cloudflare recorded 420 times as many pageviews as the other stats loggers!

All these services claim to exclude bots, spiders, and other non-human interactions.

Does anybody have any idea what would explain this? And could you please persuade me that Cloudflare is right after all? At this point I can’t begin to believe its high numbers—but I’d like to!

(The “other stats loggers” are StatCounter and Clicky, averaged. They produce different numbers from each other every day, with Clicky coming in 5% higher on average, but they correlate very highly with each other: r=0.99.)

Commenting Restored

The comment function here has been out of service, possibly causing frustration, for which I apologize. You can comment again now, and it will save and post as it should do. First-time commenters' comments will not appear, however, until approved in moderation.

4 thoughts on “Geek Question: Cloudflare Statistics Gone Wild!

  1. This link hypothesizes browser pre-fetching as one reason why CloudFlare overestimates page views. But it doesn’t explain why it suddenly shows 100x more, which seems extremely unlikely to be accurate.

  2. I’d already seen that pre-fetching theory proposed.

    I didn’t mention that while Cloudflare was skyrocketing unreasonably, my other stats were also increasing–but by a factor of two or three, not one or two hundred.

    Very strange.

  3. I ran into a period of time over the weekend when I couldn’t get to the site and was redirected to Cloudfare (I think). It was something about DDoS protection if I understood it correctly.

  4. There was actually a very short period when I had to increase the security level, because the server was very overloaded, looking as if it were under attack or some other unexpected load.

    That lasted an hour or so. It had no effect on the statistics, though, since Cloudflare counts real page views separately–or so they claim–and the stats have been high, both before and after.

Comments are closed.

Subscribe

Subscribe here to receive updates and a free Too Good To Be False preview chapter!

"Engaging… exhilarating.… This might be the most surprising and refreshing book you’ll read this year!" — Lee Strobel

"Too Good To Be False is almost too good to be true!" — Josh McDowell

Purchase Here!

More on the book...

Goodreads

Too Good to be False: How Jesus' Incomparable Character Reveals His Reality

Serving with:

Discussion Policy

By commenting here you agree to abide by this site's discussion policy. Comments support Markdown language for your convenience. Each new commenter's first comment goes into moderation temporarily before appearing on the site. Comments close automatically after 120 days.

Copyright, Permissions, Marketing

Some books reviewed on this blog are attached to my account with Amazon’s affiliate marketing program, and I receive a small percentage of revenue from those sales.

All content copyright © Thomas Gilson as of date of posting except as attributed to other sources. Permissions information here.

Privacy Policy

%d bloggers like this:
Clicky