page load results showing slower than local


I have been running some page test results for a client and I have used a lot of different tools.

The results I get back from WPT are excellent but I am seeing a much slower response from WPT than I am from in Chrome and from a server running

I know that Page Load is generally not a RUM measurement, however it is something that the client cares about.

In Chrome and PageSpeedIO the load time is around 7 seconds with a TTFB of 209ms. However in WPT I am seeing generally 12-25 seconds Load with a first byte of 353ms.

I have changed the location from Dulles to London and Manchester and played around with the bandwidth settings to match that of Chrome & (30000 down 15000 up) but I still get similar results even with max repeat’s using cache?.

Is this standard for WPT as I’m having trouble selling WPT as a reliable tool particularly when the web results are not even comparable. I understand you could argue that they are more indicative of an average user, but I wouldn’t expect to see such a large discrepancy between the load times.

Has anyone else seen this?

Chrome’s traffic-shaping isn’t as accurate as what WPT uses and will tend to skew fast. Since it doesn’t work at the packet-level it doesn’t include things like TCP slow start and also misses out on SSL negotiation round trips, cert checks, etc. If you are testing locally, I highly recommend using OS-level traffic shaping. Depending on the OS you are on there are a lot of good options: