Four weeks of reports showing which changes moved the score and which did not. Performance work goes wrong when it starts with a guess. Every change in this post began with a measurement on real hardware and ended with a second measurement to prove it helped.
Reading the reports
Field data from real users tells you where the problem is; lab tools tell you why. We start with the Core Web Vitals report, pick the worst page template, and only then open the profiler.
As always, the examples are simplified to fit on a page. The production version has more error handling and logging, but the shape of the code is the same.
Questions, corrections, and war stories are welcome in the comments. Several of our best follow-up posts started as a reader pointing out something we had missed.
We will revisit this in six months with fresh numbers. If the results change, we will update the post and note what changed at the end.
If you try this on your own project, start with one page or one service rather than the whole codebase. A small experiment gives you real numbers to argue with, and it is easy to roll back if the idea does not fit.
What we would change next time
If you try this on your own project, start with one page or one service rather than the whole codebase. A small experiment gives you real numbers to argue with, and it is easy to roll back if the idea does not fit.
- Start from field data, not a lab score.
- Fix the largest resource first.
- Compare percentiles, not single runs.
- Re-measure a week later.
This finally clicked for me after reading the docs three times.
Came for the snippet, stayed for the explanation.
Good question. We measured on a mid-range Android phone and took the median of five runs.
Good question. We measured on a mid-range Android phone and took the median of five runs.
Came for the snippet, stayed for the explanation.
Counterpoint: for small teams the simpler approach in the first section is usually enough.
The quote in the middle is doing a lot of work here, and it earns it.
We tried something similar in 2023 and rolled it back. Curious whether the tooling has caught up since.
The part about naming things made me laugh, then made me go rename three functions.