It’s election season again y’all! That means a ton of engineers at news orgs are spending sleepless nights building different reader experiences with various tech stacks. In years past I’ve informally compared all the different approaches because I find it educational. One big reason, the faster your page is, the easier it is for folks to enjoy it regardless of their internet connection speeds.
Links were gathered via Twitter, primarily using a thread by Tiff Fehr. Basic criteria are the page has to feature updating vote and/or delegate counts. Scores were tallied after running PageSpeed Insights at least three times to establish an average. Additionally, AMP pages were used over non-AMP page versions with the main assumption being that readers are coming to the page on mobile either via social media or after searching for election results. I was also really curious to see how AMP performed when it came to interactive data graphics.
Note: This is not comprehensive of all election results pages; there will be more publications publishing results pages later in the election season.
Lighthouse scores have some variance. If you see any numbers that are +/- 5 or more points, please let me know. News organizations are arranged from highest to lowest score. To explore more, here's the same table as a Google Spreadsheet.
In no particular order
- AMP is not exactly a silver bullet for performance gains, but it does help. AMP pages were mostly in the “yellow” of lighthouse scores range (50 to 80 max) compared to non-AMP pages (20s & 30s). Vox & WaPo had some of the fastest AMP pages in the high 70s.
- BUT you don’t need AMP. LAT clocks in as the fastest with a speed index of ~1.5s and a score of ~80, with Reuters not far behind. No AMP versions for either page.
- Page construction matters a lot, in other words more SVG elements means more DOM nodes and slower rendering speeds. Live blogs or canvas elements, along with shorter/smaller pages, generally did better.
Another insight:
The number of requests, one of the oldest performance measurements, is still pretty relevant when measuring performance, and plagues news sites. It's no coincidence that the top three sites have the fewest number of requests and the bottom three have the most: