Real world network delays can be caused by many different factors which can make it difficult to analyze and evaluate how quickly resources load on different network and device conditions. This is why use use benchmarking and testing to enable analysis of potential web performance issues.
However benchmarking tools can use utilize different techniques to manage the network connection; commonly known as network throttling. Therefore, it is important to know the impact the different throttling techniques may have on your analysis, otherwise you may be missing opportunities to improve performance.
This is of paramount importance if you use Lighthouse, WebPageTest or Chrome DevTools as your testing tool as each has a different approach to network throttling. This, of course, leads to different results. Catchpoint recently released their own article explaining the differences between three of the most common methods.
- Simulated throttling as used by Lighthouse
- Request level throttling as used in Chrome DevTools
- Packet level throttling as used by default in WebPageTest
Packet level is considered the "Gold Standard" as it is applied at the network packet level and therefore more closely emulates a real network connection at the throttling level bandwidth selected.
As the other approaches are not able to emulate network connection as precisely, they may not report connection data as accurately. Consequently, if network delays are inaccurate, delays in the network that impact on the overall page load may be hidden and resolution opportunities may be missed.
Catchpoint's article covers this well and in more depth so is worth a read.
