Web performance is worth investing in because it directly affects how long users wait to see content and whether a page responds promptly to their actions. web.dev notes that improving these two aspects can enhance the user experience and help a website achieve its goals. However, not every optimization is worth its cost. To evaluate results, consider loading and interaction performance, user behavior, and the maintenance costs of technical solutions. Do not judge success by a single score or figures from an individual case.
Performance is more than how quickly a page opens
How quickly content loads and how promptly a page responds to user actions are two important aspects of performance. Content appearing sooner does not necessarily mean subsequent interactions will be smooth. web.dev cautions that sending too much JavaScript can make a page slow to respond both during loading and in later interactions, frustrating users.
- Loading performance: Observe when content starts appearing, not just whether resources have finished downloading.
- Interaction performance: Focus on whether the page responds promptly to user input.
If an improvement only makes content appear sooner without addressing sluggish interactions, it improves loading performance alone. That is not enough to claim that the experience has improved across the board.
Critical CSS illustrates why optimization involves trade-offs
Critical CSS consists of the styles needed to display the initially visible part of a page. According to web.dev, correctly inlining these styles in the document eliminates the network request needed to fetch that portion of the CSS. If the browser does not already have the relevant resources cached, this may shorten initial load time. The remaining styles can be loaded asynchronously.
But this approach also has costs. Teams need to decide which browser viewport sizes to support and how to extract and maintain the styles. Users who scroll down before the remaining CSS has loaded may also see unstyled content. Inlining large amounts of CSS increases the size of the initial HTML. Unlike an external stylesheet, these inline styles cannot be reused directly from the cache on subsequent pages.
This example shows that evaluating performance gains also means considering when a technique is suitable. If a website’s architecture makes critical styles difficult to maintain, the gains may not justify the investment. Use testing to assess the trade-offs rather than treating any technique as a standard solution for every website.
Behavioral data can supplement evaluation, but does not guarantee results
Alongside technical performance, examine user behavior relevant to the website’s goals. A 2025 article on tenten.co lists measures such as static generation, loading content in stages, and reserving space for ads. It reports that, in the case it describes, the bounce rate fell by 29% and time on page increased by 41 seconds after implementation.
These figures should be treated only as claims about the individual case described in that article. The supplied excerpt does not specify the sample, observation period, comparison baseline, or other changes occurring at the same time. It therefore does not establish that the performance measures alone caused the improvements. Nor should these figures be treated as expected results for other websites. A more cautious approach is to present changes in technical performance and user behavior separately, explain the testing conditions and limitations, and then assess whether the investment helps achieve the website’s goals.
Web performance: perspectives from across the web
Can waiting time be translated directly into business value for a website?
Not directly; the value assigned to time depends on the context and assumptions. In 1998, Jakob Nielsen estimated that waiting for a typical banner ad to download cost about three cents in employee time. Assuming leisure time was worth one-third as much as work time, he inferred that people browsing for leisure might be willing to pay one cent to avoid the ad. These were estimates and inferences made at the time, not evidence of users' actual willingness to pay, and they should not be treated as a universal price for websites today.
References
- Welcome to Learn Performance!web.dev — Source used in the article above.
- Optimize resource loadingweb.dev — Source used in the article above.
- The Case For Micropaymentswww.nngroup.com · 1998-01-25T06:00:00+0000 — Discusses: Can waiting time be translated directly into business value for a website?
- 別再瞎猜 SEO!Google 演算法洩密:E-E-A-T 才是王道!tenten.co · 2025-03-23T15:09:34.000Z — Source used in the article above.


