Web Hosting Comparison by Real-User TTFB
Updated
Hosting plans list storage and price, but rarely how quickly the server actually responds. This comparison maps 61 websites measured by PerfDelta to their hosting provider and compares the Time to First Byte (TTFB) of their real Chrome visitors from the Chrome UX Report, separately for each market. Data as of 2026-10-24.
Web hosts for sites in the US
Across all sites in this market with field data, including those behind a CDN, the median is 825 ms.
Sites targeting United States. For each site we take the 75th percentile TTFB of the whole site on phones from the Chrome UX Report; the table shows the median per host with its 95% interval and the share of sites with a good TTFB up to 800 ms, sorted by number of sites. Sites behind a CDN are left out.
| Host | Sites | Median TTFB | 95% interval | Good TTFB | Pass Core Web Vitals |
|---|---|---|---|---|---|
| Amazon Web ServicesWordPress 9%, Nuxt 6% | 30 | 846 ms | 749 ms – 942 ms | 47% | 70% |
| All sites in this market | 2,148 | 825 ms | 806 ms – 844 ms | 48% | 54% |
Only hosts with at least 30 sites with field data in this market. The sample comes from PerfDelta's measurements and does not represent every customer of a host. Core Web Vitals: whole site, all devices.
Web hosts for sites in Germany, Austria and Switzerland
Across all sites in this market with field data, including those behind a CDN, the median is 616 ms.
Sites targeting Germany, Austria or Switzerland. For each site we take the 75th percentile TTFB of the whole site on phones from the Chrome UX Report; the table shows the median per host with its 95% interval and the share of sites with a good TTFB up to 800 ms, sorted by number of sites. Sites behind a CDN are left out.
| Host | Sites | Median TTFB | 95% interval | Good TTFB | Pass Core Web Vitals |
|---|---|---|---|---|---|
| Amazon Web ServicesAdobe Experience Manager 12%, SAP Commerce Cloud 9% | 31 | 656 ms | 452 ms – 977 ms | 61% | 58% |
| All sites in this market | 749 | 616 ms | 586 ms – 651 ms | 66% | 68% |
Only hosts with at least 30 sites with field data in this market. The sample comes from PerfDelta's measurements and does not represent every customer of a host. Core Web Vitals: whole site, all devices.
Method: metric, host mapping, CDN and location
The metric is the Time to First Byte of the whole site on phones, the 75th percentile from Google's Chrome UX Report over 28 days. Every site's value comes from the same query and reflects page loads by real Chrome visitors, not a lab test.
Response headers set by the host come first, for example on Kinsta or WP Engine. Without them, the reverse DNS of the server IP names the operator: a hostname under kasserver.com points to All-Inkl, one under your-server.de to Hetzner. Name servers are not used, because they often only show where the domain was registered.
When a CDN such as Cloudflare or Akamai sits in front of a site, a server at the edge of the network answers the request, and TTFB no longer measures the host. Those sites are left out of the comparison even when the host behind them is known. Platforms whose network is itself a CDN have their own table. Sites on Shopify, Wix or Squarespace are left out as well; their numbers are on the platform pages.
TTFB includes the network distance between visitor and server, so each market gets its own table, based on the site's target country. A host appears once it has 30 sites with data in that market.
What the hosting comparison tells you, and what it does not
TTFB depends on the host, but just as much on the plan, the page cache and the software behind the site. WordPress without caching on a budget shared plan answers more slowly than the same site with a cache at the same provider. The platform mix under each host shows how different the measured sites are.
The comparison therefore describes how quickly the measured sites of a host respond to real visitors; it is not a test under identical conditions. For your own site, your own number counts: the TTFB test shows it for any address, and the guide to Time to First Byte explains the causes.
More figures by industry, platform and country are in the website speed statistics.
Questions about the web hosting comparison
Why does the comparison not name a fastest host?
TTFB depends as much on the plan, the page cache and the software of a site as on the host. The tables therefore show the measured values per host next to the value of all sites in the market, sorted by number of sites. With a page cache, sites reach a good TTFB with many hosts.
Why are sites using Cloudflare left out?
With a CDN, a server of the CDN answers the request, often from a data centre close to the visitor. TTFB then reflects the CDN rather than the host. For a fair comparison of hosts, only sites whose own server answers directly are counted.
What is a good TTFB?
According to web.dev, a TTFB up to 800 milliseconds at the 75th percentile is good and up to 1,800 milliseconds needs improvement. TTFB is not a Core Web Vital, but it delays FCP and LCP.
How often is the comparison updated?
The figures are recalculated daily. Each site's TTFB is fetched from the Chrome UX Report every four weeks, and its host is re-checked every three months.
A free account saves your tests and shows how your values change from one test to the next. Daily monitoring with field data and alerts starts with the Solo plan.