I can’t believe it’s been a week since I released my last article in this series. That article was on a really quick Hugo shortcode that I wrote to do a better job of optimizing how my images load on pages.
Checking in on my results
As I promised two articles ago, I finally ran another website performance audit. My results were fairly good but I have definitely ran into a point where I have optimized performance as much as is worth it. My scores were quite good, but I’ve reached a point where improving one area is actually hurting others. That’s a good sign to stop - I don’t have any glaring performance problems and my site is performing extremely well. There are a few more things I could do, but I don’t think they will improve performance enough to be worth the time they’ll take.
Here are the average scores from the three audits. There was a slight change between my first and second articles. My average transfer size at initial load dropped significantly between the second and the third (because of the shortcode I wrote in the last article), and while performance improved a little in mobile, average performance dropped (by 0.1 over 26 pages).
Overall, my average performance scores increased from 96.5 and 96.2 (mobile and desktop) to 97.6 and 97.0 (again, mobile and desktop). My site’s core web vitals either improved or stayed exactly the same. My site’s average cumulative layout shift score was 100 in both forms at the beginning of my improvement program and they’re 100 at the end. First input delay scores started as 97.3 amd 98.3. Now, they’re 99.6 and 99.5. My largest contentful paint scores went up from 94.8 and 94.4 to 96.9 and 96.1.
My site’s average total transfer at page load dropped by nearly 20,000 bytes. That savings was created by getting rid of Font Awesome entirely and changing how my pages load images.
Was it worth it?
Objectively, no. My performance scores started off extremely good and I can’t think of many reasons that I would suggest that anyone else optimize performance when their scores are already as high as mine were.
However, I built SiteImp and this was my first chance to use it as a real client would. I learned a lot about the product and have a number of changes to make my next technical sprint. From a product perspective, it was extremely worth it.
As well, I’m happy that I finally wrote a shortcode so I can properly use the uk-img class with my Hugo sites.
Finally and perhaps most importantly, I got to go through a really good case and collect data on when a site is about to over-optimized. The latest round of enhancements improved some things but caused a few regressions on my page metrics reports. I’m going to spend some time analyzing what happened and see if I can write some code to predict when it’s about to start happening.
So all in all, I’m happy that I had the chance to work with SiteImp and am grateful for the learning experience. But again, my actual improvements were fairly small.
- Improving hluska.ca’s performance with a Siteimp audit
- Eliminating render blocking resources part 1 - rel preload
- Thinking through css on hluska.ca
- Checking in on those changes
- Investigating Font Awesome on my contact page
- Minifying my site with Hugo on DigitalOcean App Platform
- Writing a Hugo shortcode to optimize my images with UI Kit - last post
- Final results from my site improvement program - this post