Peter Booth
Jul 21, 2017 · 2 min read

Er no. There are a bunch of great reasons why you should be focussing on RUM data if you are motivated towards improving your website’s performance. If you glance at the history of web performance engineering it starts to make sense …

In the beginning there was Keynote, and there was Gomez. Two companies who offered services to remotely test websites from a bunch of remote locations and report performance data. These services published league tables, or indexes, of businesses so that, for example, the relative performance of the Citibank and Chase websites could be seen over time.

The test agents were stripped down browsers, the URLs they hit were limited. One of these two market leaders deployed it’s test agents in datacenters all over the world, and coincidentally, the market leading CDN deployed it’s edge servers in exactly the same datacenters. This meant that these services did a great job of accurately reporting the performance of a website for a user who lives in a datacenter with a PEER1 connection to the internet.

The consequence of this is that large Fortune 500 companies used a bad metric to measure and report the performance of their website. Their metric was overly optimistic, and it largely ignored the cost of images, css, js, tracker pixels that make up the bulk of website response times for real users.

Bottom line — you could improve your performance on Gomez or Keynote by 33% without making a noticeable difference to what a real user experiences. I can recall in the mid 2000s being greeted by laughter when i suggested that the time to build a web page wasn’t the most important variable in web performance. Developers wanted to focus on server response times because thats what they knew how to tune. We have come a long way since then, and tools like webpagetest.com, redbot, NewRelic, pagespeed, yslow, etc are the reason why.

I once worked at a place where the CTO’s bonus depended on the company’s Keynote ranking. Can you guess what variable we were instructed to optimize for?

Choice of KPI is crucial. If you are tuning a website because there’s a commercial benefit from a faster user experience (greater customer retention) then RUM makes sense. Can you suggest a better metric?

)

    Peter Booth

    Written by