- Cloudflare reports it is the fastest provider in 74% of the 1,000 largest networks, up from 60% in April, based on its own measurements.
- Cloudflare says the gain coincides with a new measurement method, and it does not say how much of the rise comes from each cause.
- Before using any vendor speed ranking in a purchase, ask what was measured, where, and how large the gaps are.
A speed ranking is a measurement decision before it is a result. Change how you measure, and the leaderboard can move even if the network has not. That is the useful lesson in Cloudflare's latest performance update, and it applies to any vendor that ranks itself.
What Cloudflare reported
In August, Cloudflare says, it had the lowest connection times of any provider in 74% of the 1,000 largest networks. Its April update put the figure at 60%. The company says it gained 150 networks and 38 countries over that period.
These are Cloudflare's own figures, from its own measurements. The post does not describe an independent audit.
How the test works
Cloudflare starts with the 1,000 largest networks, ranked by estimated user population using APNIC data. For each one, it measures connection time. That is how long a user's device takes to complete a TCP handshake, the first exchange that opens a connection to a server.
It then ranks providers by the trimean of those times. The trimean blends the 25th, 50th and 75th percentiles into one weighted figure. This reduces the pull of unusual outliers while still reflecting the range of user experience.
The data comes from users' own browsers. A small background test fetches lightweight files from Cloudflare, Amazon CloudFront, Google, Fastly and Akamai. It records whether each request finished and how long it took.
Since 2021, those tests have run on Cloudflare-branded error pages. The new step is to run them on Cloudflare Challenge Pages as well. These are screens, delivered by Cloudflare Turnstile, that check visitors before they reach a site. Cloudflare says the rollout is cautious. Only a small share of eligible free Challenge Pages carry the test today, and the company says it will raise that share only if data quality improves and users notice no slowdown.
The ruler changed at the same time
Cloudflare says the improvement "coincides with" the new method. It also explains why that matters. In many networks, the leading providers sit within a millisecond or two of each other. Cloudflare gives one case: 50 ms for itself against 51 ms for Fastly. A gap that small can flip with ordinary daily variation.
More measurements make rankings steadier, Cloudflare argues. It says it may previously have shown as second in some networks while trailing by only 1 or 2 ms. More data narrows the error range around each provider's figure.
That is a reasonable argument. It also means the 14-point rise mixes two things: possible real speed gains and a steadier, larger sample. The post does not say how much comes from each. Cloudflare calls the results early.
What this means for buyers
The ranking says a provider wins on a specific test. That test is connection time, not full page load or application speed. It runs through pages that Cloudflare serves. A leader by a millisecond may not differ from a rival in any way your customers can feel.
None of this makes the result wrong. It makes it a vendor's evidence, to be weighed as such. Questions to put to your network or platform team:
First, where do our own users sit? A win in 74% of networks does not tell you about the networks your customers use. Check your own traffic against the networks that matter to you.
Second, how large is the gap? A one or two millisecond lead sits inside the noise Cloudflare itself describes. Ask for gaps, not ranks.
Third, can we test it ourselves? Real-user measurement on your own sites is better evidence than any vendor's leaderboard, this one included.
Fourth, does the vendor disclose method changes? Cloudflare did, and that is what makes this analysis possible. Ask every provider the same.
A leaderboard is only as steady as its ruler. Before you let a ranking drive a contract, ask who holds the ruler and when it last changed.
Produced by the WebPulse Newsroom with AI assistance from the original reporting credited below, and checked against that source by our editorial review. How we use AI.
Original reporting: Cloudflare.
