Does your page keep visitors waiting?
Enter your site's address. We load it on a simulated phone, on slow 4G, three times, and tell you what weighs: images, requests, cache, scripts. The measurement runs from our server in Asia (Malaysia): a site that is fast from here is even faster from France or Europe, if it’s hosted there.
The same page, two phones, the same network. The only difference: on the right, three images were recompressed and one word added to a tag. Watch who gets to read first.
The same page, loaded twice side by side. On the right, three recompressed images and one word in a tag. An hour's work.
1.3 seconds. On the right, the text is there, the visitor is reading.
On the left, still blank. Half the visitors are already gone.
5.4 seconds. The left page finally arrives, in front of nobody.
Four times faster, with identical content.
Nothing changed in the site's code. Three images and one word.
What divides the time by four takes three images and one word in a tag.
On a desktop, at the office, on fiber, you'll never see this wait. Your visitors will.
Weighttransferred
Imagesoversized
Requestsin total
Compressionbrotli, gzip
Cacheone day minimum
HTTP/2or HTTP/3
Scriptsrender-blocking
Fontslightweight
Third partiesshare of weight
Serverfirst byteWeighttransferredImagesoversizedRequestsin totalCompressionbrotli, gzipCacheone day minimumHTTP/2or HTTP/3Scriptsrender-blockingFontslightweightThird partiesshare of weightServerfirst byteStructural checks, the same from Paris to Tokyo. Load times are shown as references, with their method: they don't enter the score.
Why not a simple stopwatch?
Because a stopwatch depends on everything: distance to the server, network, device, time of day. Two runs on the same site differ by 15%. What the page weighs, the request count, oversized images, caching and compression are the same from Paris to Tokyo, and they're what causes slowness. The score rests on those.
What are the displayed times worth?
They’re reference points, not the score. They’re measured in a simulated mobile browser, on a slow 4G network with a throttled CPU, from our server in Malaysia, and the median of three runs is shown. The only time that enters the score is the server’s, the first byte, because it depends on neither the device nor the rendering.
Why does the place of measurement matter?
Because the first byte includes the trip between your server and ours, in Malaysia. If your site is hosted in France or Europe, your visitors there get a faster answer than the one measured here. Compare this time mainly from one measurement to the next.
What does the score measure?
Checks grouped into categories, out of 100. Weight (total, images, requests), delivery (compression, cache, protocol), rendering (blocking scripts, fonts, third-party share) and the server. A check with nothing to examine, a page without web fonts for instance, leaves the calculation.
Who fixes this, and how fast?
Images are recompressed in an hour with the named list. Cache and compression are a few lines at your host. Blocking scripts need a “defer”. Nothing here requires rebuilding the site.
Audit of example.com
0,0 sYour score is building.
Each category adds the points it earns, and shows the ones you're missing.