Skip to content
JournalTORIS / FIELD NOTES

← Blog Project notes

Keeping Measurements I Can Explain Again, Not Just Performance Scores

Lessons from image optimization and code splitting on a blog and mobile screens, and how I separate past measurements from current state, and lab results from real-user performance.

2 min read Read the Korean original

Joohwan Yu · Founder, Full Stack Developer

Toris · Published

Contents

While building a blog and mobile screens, I worked on image optimization, code splitting, and cutting unnecessary rendering. I also kept past measurement records in my portfolio. But when you want to reuse those numbers later, you first have to ask: which screen did I measure, and under what conditions?

A number needs a time and conditions attached

A score measured once is a result for the code, content, and environment of that moment. As posts and images grow or shared scripts are added, even the same site can change. If you present past results as the performance of every page today, the meaning of the record changes.

In this retrospective, I don’t carry old scores forward as if they were current guarantees. Instead, in the next measurement I want to record the route, the code before and after the change, the device conditions, and the measurement date and time together. A record that someone else can follow under the same conditions matters more than the sentence “it got faster.”

Separate the bottleneck before optimizing

What to do depends on whether a slow first screen is caused by a large image, JavaScript execution, or an external request. Applying React.memo doesn’t reduce every loading problem. An image pushing content around and a component re-rendering too often are also different axes.

Today I would pick one representative user flow and find the biggest delay within it. If you make several changes at once, it’s hard to separate what actually worked. Narrowing down the cause, making only the needed change, and then comparing again under the same conditions gives you a basis for the next task.

Between lab scores and user experience

Checks in a fixed environment are useful for catching regressions. But real users arrive with different networks, devices, and saved state. In the project retrospective at the time, I also noted a regret that real-user performance monitoring was lacking. I don’t want to cover that over with a single tool’s score.

The standard that stuck with me from performance work isn’t applying many optimization techniques. It’s explaining the problem, recording why I changed things, and not claiming more than the scope of the result. In my next portfolio, rather than listing only good numbers, I’ll also record the conditions that produced them and the parts I haven’t verified yet.

Related post: Toris Blog development experience

Advertisement