By Christoph Olivier. Published 2026-07-31. Original first-party dataset.

This is an original performance study of small business websites in the United States. Between the crawl dates, CO Consulting requested the homepage of 373 small business websites across six service verticals and eight cities and recorded server response time, transfer protocol, compression, content delivery network use, mobile viewport configuration, HTML document weight, and render blocking resources. Of the 373 sites sampled, 320 returned a successful homepage response and form the analysis set. All figures below come from that first-party crawl unless a source is named.

Executive summary

• The median small business homepage returned its first byte in 1,206 milliseconds. Google documents a good time to first byte as under 800 milliseconds, so the median site in this sample was above that threshold.
• Only 20.9% of reachable sites met Google’s good time to first byte threshold of under 800 milliseconds. 28.8% were in the poor band above 1,800 milliseconds.
• Modern transfer protocol adoption was high: 90.3% of reachable sites served the homepage over HTTP/2 or HTTP/3, and 46.2% advertised HTTP/3 support.
• Text compression was common but not universal: 85.9% of homepages were served compressed, leaving 45 sites (14.1%) sending the homepage uncompressed.
• A mobile viewport meta tag was present on 89.1% of sites. 35 sites (10.9%) had no viewport meta tag, a basic mobile readiness gap.
• Sites behind a content delivery network returned a lower median time to first byte (1,018 milliseconds) than sites with no detected CDN (1,361 milliseconds). This is an association in the sample, not a controlled test.
• Only 17.8% of reachable sites (57 of 320) cleared a basic four point readiness bar at once: HTTP/2 or better, compression on, a mobile viewport tag, and a good time to first byte.
• Accounting and tax sites were the slowest vertical by median time to first byte (1,389 milliseconds); marketing agencies and financial advisor sites were the fastest.

Key findings

• Median time to first byte across 320 reachable homepages: 1,206 milliseconds. 75th percentile: 2,005 milliseconds.
• Time to first byte distribution against Google’s documented thresholds: 20.9% good (under 800 milliseconds), 50.3% needs improvement (800 to 1,800 milliseconds), 28.8% poor (over 1,800 milliseconds).
• HTTP/2 or HTTP/3: 90.3%. HTTP/1.1 only: 9.7%. HTTP/3 advertised via alt-svc: 46.2%.
• Compressed homepage response (gzip, brotli, or similar): 85.9%.
• Content delivery network detected from response headers: 46.2%. Cloudflare was the most common, on 124 sites (38.8% of reachable sites).
• Mobile viewport meta tag present: 89.1%.
• Median HTML document weight: 31 kilobytes. 75th percentile: 53 kilobytes. 10.0% of homepages had an HTML document over 100 kilobytes.
• Render blocking resources in the document head: median of 2 blocking scripts and 5 stylesheet links. 25.9% of sites had five or more render blocking scripts in the head.
• Lazy loading of images: among 249 sites with at least three images on the homepage, 62.7% used the native loading=”lazy” attribute.

Reachability and excluded sites

Of the 373 sites sampled, 53 did not return a usable homepage during the crawl window and were excluded from all performance figures. Of those, 29 returned an HTTP 403 status, which typically indicates a bot filter or web application firewall blocking the crawler rather than a site that is down for human visitors. A further 22 returned no response within the timeout, 1 returned a 503, and 1 returned a 404. Because a 403 to an automated request does not describe the experience of a real browser, these sites are reported here as excluded, not as failing.

Server response time by vertical

Time to first byte is the interval from the start of the request to the first byte of the homepage document, measured from a single request per site including any redirects. The good, needs improvement, and poor bands use Google’s documented time to first byte guidance of under 800 milliseconds for good and over 1,800 milliseconds for poor.

Infrastructure and delivery by vertical

Original synthesis

1. Protocol modernization has outpaced response time

Logic: compare the share of sites on a modern transfer protocol with the share meeting the good time to first byte threshold. 90.3% of reachable sites served the homepage over HTTP/2 or HTTP/3, yet only 20.9% met the good time to first byte threshold. Inputs: this study’s protocol field and time to first byte field across the reachable set. Limitation: protocol adoption is often set by hosting defaults, while time to first byte depends on server processing, database calls, and origin location, so the two can move independently.

2. Sites on a CDN returned a faster median response

Logic: split the reachable set by whether a content delivery network was detected in response headers, then compare median time to first byte. Sites with a detected CDN had a median time to first byte of 1,018 milliseconds against 1,361 milliseconds for sites with no detected CDN. Inputs: the CDN detection field and time to first byte field. Limitation: this is an observed association, not a controlled experiment. Sites that adopt a CDN may also invest in better hosting, so the gap cannot be attributed to the CDN alone.

3. A four point readiness bar clears fewer than one in five sites

Logic: count sites that simultaneously met four basic checks: HTTP/2 or better, a compressed homepage, a mobile viewport tag, and a good time to first byte under 800 milliseconds. Only 57 of 320 reachable sites (17.8%) cleared all four at once. Inputs: the four fields named above. Limitation: this is a floor of basic hygiene, not a full performance audit. It does not measure lab or field Core Web Vitals, which were out of scope.

Methodology

Sample. 373 small business websites in the United States, drawn across six service verticals (marketing agencies, home services, financial advisors, law firms, dental and medical practices, and accounting and tax firms) and eight cities (Austin, Charlotte, Phoenix, Denver, Columbus, Portland, Tampa, and Nashville). This is a purposive sample of real small business sites, not a probability sample of all United States small businesses, so the figures describe this sample and should not be read as national point estimates.

Measurement. Each homepage was requested once over HTTPS with a desktop browser user agent, following redirects, with compression enabled. Recorded fields: negotiated HTTP protocol version, time to first byte, total transfer time, HTML document transfer size, content encoding, HTTP/3 advertisement via the alt-svc header, content delivery network signatures in response headers, presence of a mobile viewport meta tag, count of render blocking scripts and stylesheet links in the document head, and use of native image lazy loading. Measurements were taken from a single network location at one point in time and reflect a cold request rather than a distribution of real user sessions.

Inclusion and exclusion. Only sites returning an HTTP status in the 200 to 399 range with a homepage document were included in performance figures. The 53 sites that returned a 403, a 404, a 503, or no response were excluded and reported separately.

Thresholds. The good, needs improvement, and poor bands for time to first byte follow Google’s documented guidance on web.dev of under 800 milliseconds for good and over 1,800 milliseconds for poor.

Limitations. This study measures server response time and page delivery signals. It does not measure lab or field Core Web Vitals such as Largest Contentful Paint, Interaction to Next Paint, or Cumulative Layout Shift, which require field data from the Chrome User Experience Report or a full lab run and were out of scope. Time to first byte from a single synthetic request can differ from the 75th percentile field value a site would show in the Chrome User Experience Report. Content delivery network detection relies on response header signatures and will undercount providers that do not expose identifying headers. Render blocking counts are parsed from static HTML and do not account for scripts injected at runtime.

Last updated. 2026-07-31.

Source quality ranking

Tier 1, primary. The CO Consulting first-party crawl dataset described above (original measurement). Google web.dev documentation for the time to first byte thresholds.

Tier 2. None used. No third-party market research figures are presented as findings in this study.

Tier 3. None used.

For journalists and analysts

Most quotable statistics

• Only 20.9% of small business homepages in this sample met Google’s good server response time threshold of under 800 milliseconds.
• The median small business homepage returned its first byte in 1,206 milliseconds.
• Fewer than one in five sampled small business sites (17.8%) cleared a basic four point performance readiness bar.
• 35 of 320 sampled small business homepages (10.9%) had no mobile viewport meta tag.
• Accounting and tax firm sites were the slowest vertical, with a median time to first byte of 1,389 milliseconds.

Data limitations for reporting

This is a purposive sample of 373 United States small business sites, not a national probability sample. Figures describe the sample. Time to first byte was measured from a single request per site from one location and is not a field percentile. The study does not measure Core Web Vitals.

Downloadable dataset, recommended fields

Domain, vertical, city, reachable flag, HTTP status, negotiated protocol version, time to first byte in milliseconds, total transfer time in milliseconds, HTML document weight in kilobytes, compression flag, CDN name, HTTP/3 advertised flag, viewport tag flag, render blocking script count, stylesheet link count, and lazy loading flag.

Press summary

A first-party crawl of 373 United States small business websites across six service verticals and eight cities finds that most small business homepages are slow to respond. Among the 320 sites that returned a homepage, the median time to first byte was 1,206 milliseconds, and only 20.9% met the good server response threshold of under 800 milliseconds that Google documents. Basic infrastructure was more modern: 90.3% used HTTP/2 or HTTP/3 and 85.9% served compressed responses. Even so, only 17.8% of sites cleared a four point readiness bar of modern protocol, compression, a mobile viewport tag, and a good response time at once. Accounting and tax sites were the slowest vertical and marketing agencies the fastest. Sites on a content delivery network returned a faster median response than those without one. The study measures server response and delivery signals and does not measure Core Web Vitals. Figures describe the sample and are not national estimates.

Suggested headlines

• Only 21 percent of small business websites hit Google’s server speed mark, study finds
• The median small business homepage takes 1.2 seconds to start loading
• Small business websites have modern plumbing and slow servers
• Fewer than one in five small business sites clear a basic speed and mobile bar
• Accounting firms run the slowest small business websites in a six vertical study