TL;DR

Most global teams measure their website with tools that report from their own region, so the number they trust says nothing about what a visitor in Mainland China actually sees. Measuring performance for China means measuring from inside China, across the right metrics (time to first byte, render start, visually complete, and load time), from more than one city and network, and more than once. It also means accepting that your usual analytics often under-report China traffic, because trackers like Google Analytics may load inconsistently from inside the country. This guide covers what to measure, how to measure it, and how to read the results without drawing the wrong conclusion. Chinafy measures how a site performs from inside Mainland China and optimizes both the code-level and infrastructure-level issues behind the numbers.

Why teams find out about their China problem last

Most teams find out their website has a China problem the same way. Someone inside the country tells them. A colleague in Shanghai, a distributor in Shenzhen, or a customer who writes in to say the page never loaded. By the time that message lands, the site has often been underperforming for months.

This isn't uncommon. A site can score well in Lighthouse on a laptop in London, return a clean uptime report from a US monitoring service, and still take twenty seconds to render for a visitor in Guangzhou. That gap is the whole reason to measure from inside the country. Performance in Mainland China is a separate question from performance anywhere else, and it can only be answered by testing from inside China.

The reason sits in how a page loads. When a browser in China requests a page, it still has to reach whatever the page depends on, and some of those dependencies live on services that respond slowly or not at all from inside the country. The browser doesn't skip them. It waits, retries, and moves on only once the connection gives up. A test run from Virginia never sees that wait, because from Virginia the resource answers instantly. Same site, same code, different vantage point, completely different result.

So the first step in measuring website performance for Chinese markets isn't choosing a metric. It's choosing where you measure from.

What "performance" actually means when you measure it

"Slow" is not a measurement. Before comparing tools, it helps to agree on what the numbers describe, because a single page-load figure hides most of what matters.

The metrics worth watching:

Time to First Byte (TTFB): how long the browser waits before the server sends the first piece of data. High TTFB from inside China usually points to cross-border routing and DNS resolution rather than anything on the page itself.

Render Start (First Contentful Paint): the moment the visitor sees the first pixel of real content instead of a blank screen. This is the number that maps most closely to the feeling of "it's loading."

Speed Index: how quickly the visible part of the page fills in. Two pages can finish at the same time while one feels far slower, and Speed Index captures that difference.

Visually Complete: when the page stops changing above the fold. A useful proxy for when the visitor can actually read and use the page.

Load Time and fully loaded: when the page has finished pulling everything, including scripts and third-party calls that fire after the visible content appears.

For a China audience, the spread between these numbers tells you more than any single figure. A page can reach First Contentful Paint quickly and then stall for another fifteen seconds while it waits on an unreachable script, an embedded map, or a reCAPTCHA widget. Looking only at "load time" hides that stall. Looking at the full sequence shows you where it happens.

Chinese users are not patient with the gap. Industry benchmarks put the expectation at under two to three seconds before a visitor starts to abandon a page, and Chinafy's 2026 testing found a median visually-complete time of 17.1 seconds across 614 global sites loaded from Beijing. The distance between the expectation and the reality is the thing you are trying to measure.

How to measure website performance from inside Mainland China

The practical challenge is getting a real measurement from inside the country. A handful of tools can do it, each with trade-offs.

Synthetic testing runs a scripted page load from a fixed location and records the result. It's the most repeatable way to compare a site against itself over time. WebPageTest by Catchpoint offers test nodes in Beijing, full waterfall charts, and a filmstrip view that shows the page rendering frame by frame. The waterfall is where the useful detail lives, because it names the individual resources that stall and shows how long each one held things up.

Other synthetic tools test from Chinese nodes too, including Dotcom-Monitor, Site24x7, and WebSite Pulse, each covering a different set of cities such as Beijing, Chengdu, Chongqing, and Guangzhou. Chinafy's own speed test tools like the China Readiness Score compare load times and resource behaviour across Beijing, Shanghai, and Guangzhou against a global baseline, and flag the specific third-party resources causing incompatibilities.

Two caveats are worth building into how you read any of these results.

First, Hong Kong is not a proxy for Mainland China. Several popular free tools only offer a Hong Kong node, and Hong Kong sits outside the mainland's network conditions. A clean result from Hong Kong tells you almost nothing about Beijing or Chengdu. If a tool only tests from Hong Kong, treat it as a control region, not a China measurement.

Second, synthetic tests describe the machine, not the visitor. They tell you how a page loads from a data centre in a given city. They don't tell you what a real person on a mobile network in that city experiences, which is why synthetic data is a starting point rather than the whole answer.

The analytics blind spot: why your own data under-reports China

Once a site is live, most teams reach for their own analytics to understand how it performs. In China, that data is often incomplete, and the gaps run in a direction that hides the problem.

Analytics platforms rely on a tracking script that has to load and fire before it can record anything. If that script sits on a service that loads slowly or fails from inside Mainland China, the visits that suffer most are the least likely to be counted. A visitor whose page times out before the script runs never appears in your data at all. The sessions you lose are precisely the ones you most need to see.

Google Analytics is the clearest example. The tracking snippet can collect data when a page fully loads, but connectivity issues with Google's servers can cause delays, discrepancies, and lost sessions, and the reporting dashboard at analytics.google.com is blocked from inside the country. So a China-based team typically can't view the data, and the data itself under-counts the visitors who had the worst experience.

This is why many teams operating in China run a China-friendly analytics tool alongside or instead of their global one. Baidu Analytics (Tongji) is the most common, with Alibaba-owned Umeng and the lightweight 51LA also in use. Each hosts its measurement inside the region, so the script is more likely to load for the visitors you're trying to measure. Setup carries its own requirements, including a Chinese mobile number to register for some platforms, so it's worth scoping before committing.

The takeaway for measurement is simple. If your only view of China performance comes from a global analytics dashboard, assume it is flattering the picture. Pair it with synthetic testing from inside the country and, where possible, a locally hosted analytics tool, so the visitors who never made it onto the page are still part of the story.

Reading the numbers: variance by region, network, and time

A single test from a single city on a single day is a snapshot, not a measurement. Conditions inside Mainland China vary in ways that a one-off test can't capture, and reading too much into one number is a common way to reach the wrong conclusion.

Three sources of variance to account for:

Region. A resource that loads cleanly from a Beijing node can time out from Shenzhen, or the reverse. Testing from at least two or three cities gives a truer range than any single location.

Network. China's main carriers (China Telecom, China Unicom, and China Mobile) route traffic differently, and cross-border performance can differ meaningfully between them. A result on one carrier isn't a result on all three.

Time. Performance can shift by day and by time of day as routing and congestion change. A test at 3am local time may not reflect the experience at peak.

The practical response is to measure more than once, from more than one place, and to treat the spread between your best and worst results as the real finding. If a page loads in eight seconds from Shanghai and times out from Chengdu, the honest summary is not "eight seconds." It's that a meaningful share of visitors may be getting the timeout. For the same reason, a measurement taken once can drift as conditions and the site itself change, so a single clean test is best treated as true for the day it was taken rather than settled for good.

Load time is one measure of performance, not the whole picture

Everything above measures whether the page loads. That's the layer Chinafy works in, and it's the right place to start, because a page that doesn't load can't perform on any other measure. But a site that loads fast in Mainland China can still underperform commercially, and teams serious about the market eventually need a second set of numbers: whether the site actually works for the people it loads for.

This is a different measurement question. It's less about seconds and resources and more about whether the content fits how Chinese users search, browse, and decide. A page can render in two seconds and still lose the visitor if the messaging was written for a different audience, if the payment and contact options don't match local habits, or if the traffic is arriving through channels the brand hasn't measured. Load time tells you the door opened. It doesn't tell you whether anyone wanted to walk through it.

Measuring that layer sits inside brand and market strategy. For an expert view on how to measure whether a website is actually performing for a Chinese audience, we asked the team at NBH, a China brand team that helps global companies build and grow their digital presence in the market.

A fast-loading page is the starting point, not the finish line. We regularly see European brands ship a site that hits every technical benchmark in Mainland China and still underperforms commercially, because the content reads as a translated global site rather than one built for a Chinese buyer. Speed gets someone through the door. What happens next depends on whether the messaging, the trust signals, and the contact experience feel native to the market, the difference between a page that loads and a page that converts is usually China website localization, not additional load speed optimization.

"We have seen technically flawless websites fail commercially in China because they never adapted the substance, not just the speed," says Jonathan Kullman, CEO of NBH.

We see this pattern often: a European manufacturer invests in getting their site to load reliably from inside Mainland China, sees the technical metrics improve, and then plateaus on inquiries. The visitors are arriving. They're just not converting, because the page that loads fast is still speaking to a European buyer, not a Chinese one. Fixing the load time removed one barrier. It didn't remove the other. Both layers have to hold for the investment to pay off, which is why we would frame technical readiness and market readiness as sequential, not competing, priorities.

A few signals worth tracking once load speed is solved: inquiry rates from visitors who land after reading locally reviewed content versus a directly translated page, time spent on pages that carry visible local trust markers, and whether contact form submissions actually convert to qualified conversations rather than one-off curiosity clicks.

NBH is a China marketing agency helping European companies build and grow their digital presence in the market, from website localization through content strategy and campaign execution.

How Chinafy helps

Chinafy measures how a website performs from inside Mainland China and optimizes the issues behind the numbers. On the measurement side, testing runs from nodes inside the country rather than from offshore control points, so the data reflects what visitors actually experience. On the optimization side, Chinafy combines automated rule-based optimizations with human engineering to address both the code-level dependencies (the external services a page calls that respond poorly from inside mainland) and the infrastructure layer (a managed CDN stack with near-China or onshore nodes built for the region).

The result is that an existing offshore site can reach near-native onshore performance in Mainland China without requiring a rebuild or rehost, while the global version of the site stays exactly as it is for everyone else. Because conditions inside China shift over time, the optimization rules and monitoring are maintained on an ongoing basis rather than set once and left.

For questions that sit outside web performance, including localization, China go-to-market, and presence on Chinese platforms, Chinafy can point you to partners like NBH who specialize in that work.

To see how your specific site performs from inside Mainland China, get in touch with Chinafy for a free evaluation.

Disclaimer: This guide is informational and does not constitute legal or compliance advice. Chinafy is not a legal or corporate advisory entity, and, given that every business is different, we suggest consulting your internal legal counsel on any regulatory questions, or we can connect you with one of our partners.

FAQs

Why can't I measure my website's China performance with my usual tools?

Most speed and monitoring tools test from the region they're set up in, so they measure how your site loads from the US or Europe, not from inside Mainland China. Because performance in China depends heavily on whether the page's external resources are reachable from inside the country, a test run from elsewhere can look clean while real visitors in China wait far longer or see the page fail. To measure China performance, the test has to originate inside Mainland China.

Which metrics matter most for China website performance?

Watch the full sequence rather than a single figure: Time to First Byte, Render Start (First Contentful Paint), Speed Index, Visually Complete, and Load Time. For a China audience, the gap between when content first appears and when the page is usable is often the most revealing number, because that gap is where unreachable third-party resources cause stalls.

What tools can test a website from inside China?

Synthetic tools with Chinese test nodes include WebPageTest by Catchpoint (Beijing, Shanghai, Ningxia) and Chinafy's speed test tools. Be careful with free tools that only offer a Hong Kong node, since Hong Kong sits outside mainland network conditions and isn't a reliable stand-in.

Does Google Analytics work in China?

Partly. The tracking script can record data when a page fully loads, but it may load inconsistently from inside Mainland China, which under-counts the visitors who had the worst experience, and the reporting dashboard is hard to reach from inside the country. Many teams pair or replace it with a China-hosted analytics tool such as Baidu Analytics (Tongji) so measurement is more complete.

How often should I test China website performance?

More than once, and from more than one place. Conditions vary by region, by carrier, and by time of day, and both the network and your own site change over time. Testing from several cities on a repeated basis gives a truer range than any single result, and the spread between your best and worst outcomes is usually the finding that matters.