Mobile Site Speed Benchmark: How Fast are the UK Top 5,000?


#1

Found this case study based on HTTP Archive data on Twitter:


Some interesting observations:

  • grouping by category provides a nice way to compare sites
    • categories provided by majestic.com (potentially useful for HTTP Archive)
  • the most popular category is retail
  • averages for performance metrics are used
    • we must continue to get the word out to use medians instead
    • would be great to see how these sites perform in the Chrome UX Report, not sure how relevant synthetic performance data is in this case
  • the news and media category appears to be the slowest
  • distributions are visualized as a scatterplot with lines to show 25/50/75 percentiles

Curious to see if anyone else had any interesting takeaways.


#2

Hi Rick,

Thanks for your comments. I am the MD and founder of Erudite and the main author on that piece.

I just wanted to clarify a couple of points…

Categorisation - we created our own categories, but to help automate them in as far as possible we used Majestic.com “topic” data. That data is far more granular than our level industry categories, so we assigned the Majestic topics a roll-up category. There was still a lot of editorial assignation required and there’s I’m sure a handful of errors in the 5,000+ category assignments.

Totally agree about averages. We manually cleaned a few anomalous outliers and also provided median data; however we’re in the business of marketing - and an average value is much more tangible a benchmark for those we’re trying to communicate with.

One of the most interesting things to come out of the data for me - the News & Media sector has almost the fastest TTFB at 1.27 but the slowest fullyLoaded at 19.6 (14.73 median load time). Suspect this is because of the amount of window load analytics triggers and ad tracking activity that is critical for the survival of these sites but would welcome any other ideas on this.: News & Media All data

Maybe that’s our next project!


#3

@NicholaStott great work on the study! I’ll second Rick’s suggestion on medians: I understand the desire to “keep things simple” but averages are a misleading metric and can easily lead you astray in your analysis and conclusions when working with non-normally distributed data — which is the case here. If you simply swap out all mentions of average to median, you’re already miles ahead.

That aside, interesting results! I’d love to dig a bit deeper into some of these verticals. Curious, would you guys be willing to publish the classified URL sets? :slight_smile:


#4

Hey Ilya - I take your point on average vs. median, particularly as most readers seem to be sticking to the overview data, rather than going into the channel deep dives (where the median data is presented for each measurement).

We did publish the full list of URLs and the categorisations and and it is referenced in the first or second paragraph of the study, though for convenience here’s the link to the G Sheet.

Happy to send over the whole lot if you like? DM me an email address and I’ll share the link.

Nichola


#5

Hi, I recently updraged my website from HTTP to HTTPS. It’s a mobile price website ( https://youmobile.com.pk ) Can you guide me how much has it effected my speed. I want the 301 redirect to be speedy. How can I improve the site speed of my mobile website?