Quick answer: Lighthouse flags "Avoid enormous network payloads" when the total transfer size of a page load is too large. A good score ends at about 2,667 KiB, and 4,000 KiB scores at the median. Fix it from the largest rows down: resize and convert images, replace autoplay video and GIFs, split JavaScript, and defer or remove third-party scripts.
Lighthouse prints Avoid enormous network payloads when the page downloads too many bytes in total during load. The report line reads like "Total size was 4,318 KiB", followed by the ten largest requests. Every byte has to cross the network before the page is done, and on the throttled mobile profile Lighthouse uses, a large payload means slow LCP and a slow Speed Index. In our study of 269 Shopify storefronts, 217 of them (80.7%) failed this audit. This page covers the threshold, how to read the table, and the fixes that cut the most bytes for the least work.
TL;DR
- What: The total transfer size of every network request during page load, and the ten largest requests.
- Threshold: Lighthouse scores the total on a curve. About 2,667 KiB (2.6 MB) is where a good score ends, and 4,000 KiB (3.9 MB) is the median score.
- Fix: Start with the largest rows. That is usually images, then JavaScript (often third-party), then fonts and video. Compress, resize, remove, or load later.
What does "Avoid enormous network payloads" mean?
Lighthouse adds up the transfer size of every request the page made during the run: HTML, CSS, JavaScript, images, fonts, video, and third-party tags. Transfer size is the compressed size on the wire, so a file served with gzip or Brotli counts at its compressed size. Requests that never touched the network, such as data URIs, do not count.
The score uses a log-normal curve, per the audit source:
| Total transfer size | What Lighthouse does |
|---|---|
| Under about 2,667 KiB | Good score, the audit passes |
| About 4,000 KiB | Median score, the audit fails |
| Well above 4,000 KiB | Score falls toward zero |
The table under the audit lists the ten largest requests by transfer size, with URL and size. It does not group them by type, so read the URLs.
The audit has no weight in the Performance score by itself in Lighthouse 13. It explains the score through the metrics: bytes on the critical path delay Largest Contentful Paint and Speed Index, and large scripts add to Total Blocking Time when they run. Large payloads also cost visitors on metered mobile data real money, which is why the audit description mentions it.
What usually makes the payload enormous?
Sort the table by size and read the top five. On most sites the pattern is one of these:
| Largest rows | Typical cause | Fix first |
|---|---|---|
.jpg, .png, .gif | Full-size photos served to phones, no modern format | Resize and convert (section 1) |
.mp4, .webm, .gif | Autoplay video or animated GIF in the first screen | Replace or defer (section 2) |
.js from your domain | One large bundle, unused code, legacy polyfills | Split and trim (section 3) |
.js from vendor domains | Chat, reviews, ads, A/B testing, tag managers | Defer or remove (section 4) |
.woff, .woff2, .ttf | Many font weights, full character sets | Subset and limit weights (section 5) |
On Shopify and WordPress storefronts, images and third-party scripts are the top rows on most failing pages.
How do I reduce the network payload?
1. Resize and compress images
An image wider than its largest display size wastes bytes on every phone. Serve each image at the widths it actually displays with srcset, and serve WebP or AVIF instead of JPEG or PNG.
<img
src="/product-800.webp"
srcset="/product-400.webp 400w, /product-800.webp 800w, /product-1600.webp 1600w"
sizes="(max-width: 600px) 100vw, 50vw"
width="1600" height="1600" alt="Linen shirt, white">
An image CDN or your platform's resize parameter does this without a build step. See Serve images in next-gen formats for the format half. Lazy-load images below the first screen with loading="lazy": Lighthouse only counts requests made during the run, so images well below the first screen, which Chrome does not fetch until the visitor scrolls near them, stop counting.
2. Replace autoplay video and GIFs
A hero video or a large animated GIF can be bigger than the rest of the page. Convert GIFs to MP4 or WebM, which are often a tenth of the size. For a hero video, show a poster image and start the video after load or on interaction, with preload="none".
<video autoplay muted loop playsinline preload="none" poster="/hero-poster.webp">
<source src="/hero.webm" type="video/webm">
</video>
3. Ship less of your own JavaScript
Split the bundle by route so each page loads only its own code, remove unused libraries, and stop shipping polyfills to modern browsers. The Reduce unused JavaScript and Legacy JavaScript audits list the exact files. Make sure every text file is compressed: see Enable text compression.
4. Defer or remove third-party scripts
Vendor tags are often the largest JavaScript rows, and you cannot shrink their files. Your options: load chat, reviews and recommendation widgets after window.load or on first interaction, remove tags nobody uses (old pixels, abandoned A/B tests, duplicate analytics), and replace heavy embeds such as YouTube iframes with a lightweight facade that loads the player on click.
5. Subset and limit fonts
Each weight and style is a separate file. Keep two or three weights, serve WOFF2 only, and subset to the character sets you use. A Latin-only subset is often a fraction of the full file.
How do I fix it in Shopify or WordPress?
- Shopify: use the
image_urlfilter with awidthand theimage_tagfilter, which writessrcsetfor you. Most of the remaining weight is app scripts. In the Lighthouse table, rows from app domains are the apps to review; remove apps you no longer use, because many leave their scripts behind in the theme after uninstall. - WordPress: turn on WebP or AVIF conversion in your image plugin or host, check that the theme outputs
srcset, and use the asset manager in your performance plugin to stop plugin CSS and JavaScript loading on pages that do not use them. Page builders often load their full library on every page.
How do I verify the fix?
- Re-run Lighthouse. The total in Avoid enormous network payloads should be lower, and the top rows should change.
- In DevTools, open the Network panel, reload with Disable cache on, and read the transferred total in the status bar. Throttle to a mobile profile to see the effect on load time.
- Watch LCP and Speed Index. Those are the metrics a smaller payload moves.
What mistakes should I avoid?
- Lazy-loading the first-screen image. It lowers the payload number and delays LCP. Lazy-load only below the first screen.
- Chasing the total instead of the critical path. 3 MB that loads after the first paint hurts less than 1 MB of render-blocking CSS and JavaScript. Fix the bytes that sit in front of the first screen first.
- Compressing already-compressed files. JPEG, WebP, MP4 and WOFF2 do not shrink with gzip. Resize or re-encode them instead.
- Counting on the cache. Lighthouse measures a cold load. A cached repeat visit does not change this audit, though an efficient cache policy still helps real visitors.
Related audits
- Serve images in next-gen formats: usually the largest byte savings
- Reduce unused JavaScript: script bytes the page never runs
- Enable text compression: gzip and Brotli for HTML, CSS and JavaScript
- Largest Contentful Paint: the metric a heavy page delays most
Audit your URL at https://lighthouse-md.com.
Audit your page now
Paste your URL, get scores plus a CLAUDE.md plan for Claude Code.