WPT vs. Navigation API

Support
1

Hi Pat,

First of all, let me say thank you for this great piece of software. We are using it as private instance for nearly a year and it really helps working on performance-issues.

My question:
Besides WPT we use RealUserMonitoring to get loading-times. For that, we measure the “loadeventStart” of the W3C navigation timing API.
If I compare these values with the “document complete” values from WPT, the RUM-values always are clearly worse: around 1-2 seconds which is nearly 50%. We compare same pages and same browsers over a longer period (2 weeks).
TTFB values do not differ significantly.
Before investigating further, it would be helpful to know, if the W3C “load event star” event is the same as the WPT “Document complete”.

Should the events used by us be the same (from a technical point of view)? Or did we miss something? Is it actually possible to compare WPT and navigationtiming-API timings in a meaningfull manner?

I’m happy to hear from you soon,

Regards

2

Yes, loadEventStart is the same as WPT’s Document Complete. If you run tests on WebPagetest using Chrome or Firefox you can actually see the W3C navigation timing times directly on the waterfalls. There is a pink area that spans the dom content loaded end-start and a light blue area for the load event (end and start). The W3C resource timing data is also in the raw data of the WPT results if you pull those.

They are usually really close but they have been off by a little bit historically as Chrome was not starting the navigation event early enough for the W3C navigation timing (few hundred ms off at most).

Take a look at the distribution of your RUM values to see what the spread looks like. Usually they are all over the place and averages can be influenced like crazy from outliers.

3

Thanks Pat,

I’ll check that.