92% of Downloaded Bytes Go Unused: A Browser Q&A
Archi4 min read
Contents
- What the Browser Throws Away
- How did you find the number?
- Why would a browser download so much data it does not need?
- What is the biggest surprise in the data?
- How does that affect a person on a slow connection?
- What changes after you know that number?
- Is there a downside to focusing on the first payload?
- What should users do about it today?
What the Browser Throws Away#
Most people assume a browser downloads a page and shows it. In practice, the browser spends most of its time fetching data that never reaches the screen. We spoke with a performance engineer who has spent the past year tracing page loads on a large publisher's website. She asked to keep her name out of this piece because her employer does not clear standard interviews. Here is what she told us about the number that changed how she measures speed.
How did you find the number?#
We collected traces from 200 real user sessions on a typical news article. The median page downloaded 14.2 MB of compressed data. Then we filtered the trace to resources that actually contributed to the first paint. That number was 1.1 MB. So 92 percent of the downloaded bytes were unused for that first view. The 1.1 MB includes the HTML document, the main CSS file, and a small amount of JavaScript that runs before the first render.
Why would a browser download so much data it does not need?#
The server tells the browser to fetch everything. Most sites embed large JavaScript frameworks, third-party widgets, and font files. Some of that data is used later, when the user scrolls or clicks. But a large share is never executed at all. In one trace, the page loaded 11 font families and only 2 were actually applied to any visible text. Another trace showed 4.6 MB of JavaScript, and the browser executed 1.3 MB before the page became interactive. The rest sat in memory as parsed code that no function ever called.
What is the biggest surprise in the data?#
The ratio of requested bytes to executed bytes. Many teams measure page weight as a single number. That hides the problem. When we broke down the 14.2 MB, we saw that images made up 7.3 MB, but most images were below the fold and could have waited. The JavaScript was only 4.6 MB, but it caused the longest delay because it blocked parsing of the page. The total page weight was less important than the order in which resources arrived.
How does that affect a person on a slow connection?#
Consider a 4G connection with a real throughput of 20 Mbps, which is about 2.5 MB per second. Downloading 14.2 MB takes roughly 5.7 seconds. But the browser only needs 1.1 MB for the first meaningful paint, which could arrive in less than half a second. The remaining 5.3 seconds of network time is spent on data that the browser may never use. On a metered mobile plan, the user pays for that waste. On a crowded network, the extra requests also delay other data that the browser does need.
What changes after you know that number?#
We changed our performance target from "largest contentful paint" to "first payload time." That is the moment the browser finishes fetching the resources required for the main content. The difference is big. By moving third-party scripts to after the primary HTML and inlining the critical CSS, we cut the first payload from 1.1 MB to 480 KB. That shifted the average first paint on slower phones from 3.4 seconds down to 1.2 seconds. The total page weight barely changed, but the user experience changed completely.
Is there a downside to focusing on the first payload?#
Yes. Some pages need data for the next action, like a search autocomplete or a video player. If you move all of that out of the initial response, you might have a fast first paint and a slower interaction. The fix is to prioritize the resources that support the first user interaction instead of only the first paint. For a news article, that is the text and layout. For a shopping page, it is the product list and the add-to-cart logic. The 92 percent number is useful as a warning, not as a rule.
What should users do about it today?#
The most effective action is to block third-party scripts in the browser. Both major desktop browsers and many mobile browsers now have a built-in option to block cross-site trackers. That typically reduces the total data by 20 to 40 percent, because most of those scripts are pure overhead. Another step is to turn off automatic font downloads, which are usually set by the operating system rather than a site. Users can also enable a data saver mode, if their browser offers one, or use an extension that stops scripts until the page is fully loaded.
The 92 percent figure will not hold forever. New protocols like HTTP/3 and compression schemes will shrink the total. But the underlying problem is a mismatch between what sites order and what users actually consume. The radio on a phone draws power while that 14 MB downloads, so the waste shortens battery life as well. Until that changes, a fast browser will always be the one that throws away the most data.
More articles