Vistoplex Digital Research
What an audit of 500 UK small-business websites across ten sectors reveals about performance, Core Web Vitals, security, accessibility and conversion basics — with the full sample register, coding manual and data published.
This report measures twelve website criteria across 500 named UK small-business websites in ten sectors, audited on 13 July 2026. 465 websites were successfully audited; every homepage was fetched and coded by the same procedure, and every included site was performance-tested with Lighthouse under identical mobile and desktop lab conditions. It is a purposive, non-probability sample and should be read as a descriptive one-day snapshot of visible, machine-detectable website characteristics — not as a census of all UK SMEs, a quality ranking of any named business, or a compliance assessment.
Executive summary
The basics are broadly in place: 98.7% of the 465 audited websites served their homepage over working HTTPS, 97.2% declared a mobile viewport, and 95.3% offered a visible contact route. The gaps appear one layer up. Only 2.7% of tested sites met the lab thresholds for Google's two layout-and-loading Core Web Vitals on mobile, 9.7% of homepages presented no clear call to action in their delivered HTML, and 31.0% contained at least one accessibility problem detectable by static inspection alone.
The performance gap between devices is stark: the median Lighthouse performance score was 45 on mobile against 74 on desktop. The median mobile largest-contentful-paint was 7.22 seconds against Google's 2.5-second threshold. Sector differences were also material: median mobile scores ranged from 36 (Electricians) to 51 (Garages & vehicle repair).
These results describe what an audit can detect in a homepage's delivered HTML and lab performance on one day. They do not establish that a business is unprofessional, that a feature is absent deeper in the site, or that fixing any single item would improve commercial performance.
Quote the research as “Vistoplex’s audit of 465 UK small-business websites (500 sampled).” Do not generalise the percentages to every UK small business.
Key findings
Lab LCP ≤ 2.5 s and CLS ≤ 0.1 under Lighthouse mobile emulation. INP requires real-user interaction and is not claimed; field CrUX data was not used.
No explicit action phrase (“get a quote”, “book online”, “contact us”) was present in the delivered homepage HTML.
Static checks only: missing image alt text, missing lang attribute or unlabelled form fields. Because rendered-page issues were not tested, this is a floor, not a ceiling.
Identical pages, identical day, identical lab machine — the difference is simulated mobile hardware and network throttling, which is how most real visitors arrive.
Sector medians are descriptive; sample sizes per sector are around 46 and are published per sector below.
Ranged from 6.1% (Restaurants & cafes) to 85.1% (Dental practices). The criterion measures homepage discoverability of any pricing route, not price completeness.
From 2.1% (Builders & construction) to 67.4% (Hair & beauty salons). Quote-led trades have little reason to adopt booking widgets; appointment-led sectors increasingly have.
“Enforced” means the plain-HTTP address redirected to valid HTTPS. The residue is small but notable a decade after HTTPS became the default expectation.
Aggregate results
All percentages use the same denominator of 465 included websites. Excluded sites are not treated as failures. Definitions for every criterion are in the coding manual below.
Source: Vistoplex homepage audit of 465 included UK small-business websites, audit date 13 July 2026. Non-probability sample; no population confidence interval is claimed.
Of the 290 sites displaying a copyright year, 49 (16.9%) showed 2023 or earlier. 10 of 465 homepages (2.2%) still carried visible COVID-19 wording in mid-2026.
119 of 465 sites had at least one image without alt text; 15 of 465 did not declare a document language. Contrast, focus order and ARIA correctness were not tested and would only raise the issue rate.
Performance & Core Web Vitals
Each included site was tested with Lighthouse under mobile emulation with simulated 4G throttling, then separately under desktop conditions, on the same machine on the same day. These are laboratory diagnostics; Google’s official Core Web Vitals assessment uses field data from real users, which was not available for this study.
| Measure | Mobile | Desktop |
|---|---|---|
| Median Lighthouse performance score | 45 | 74 |
| Median largest contentful paint | 7.22 s | 2.33 s |
| Score 90 or above | 8 of 4501.8% | — |
| Score below 50 | 264 of 45058.7% | — |
12 of 450 tested sites (2.7%) met both lab thresholds on mobile: LCP ≤ 2.5 s and CLS ≤ 0.1. Adding Lighthouse’s interactivity proxy (total blocking time ≤ 200 ms) reduces the pass group to 4 of 450 (0.9%).
The median mobile page weight was 2.29 MB. Weight alone does not determine speed, but it is the simplest lever most small-business sites control.
The median UK small-business homepage in this sample loads acceptably on a desktop and struggles on the device most of its visitors actually use.
Sector analysis
Sector results are descriptive and should not be read as statistically representative of every business in each sector. Per-sector sample sizes are shown in every row.
| Sector | Sampled | Included | Median mobile score | Median desktop score | CWV lab pass (mobile) |
|---|---|---|---|---|---|
| Builders & construction | 50 | 48 | 40 | 70 | 3/456.7% |
| Plumbing & heating | 50 | 46 | 49 | 75 | 1/442.3% |
| Electricians | 50 | 50 | 36 | 60 | 0/450.0% |
| Accountants & bookkeeping | 50 | 44 | 43 | 76 | 0/410.0% |
| Dental practices | 50 | 47 | 49 | 73 | 1/472.1% |
| Hair & beauty salons | 50 | 46 | 50 | 79 | 1/462.2% |
| Restaurants & cafes | 50 | 49 | 48 | 80 | 0/490.0% |
| Estate & letting agents | 50 | 49 | 43 | 74 | 2/474.3% |
| Garages & vehicle repair | 50 | 44 | 51 | 73 | 2/444.5% |
| Veterinary practices | 50 | 42 | 41 | 75 | 2/424.8% |
| Sector | Clear CTA | Online booking | Pricing | Reviews | Schema | A11y issue |
|---|---|---|---|---|---|---|
| Builders & construction | 44/4891.7% | 1/482.1% | 6/4812.5% | 31/4864.6% | 41/4885.4% | 14/4829.2% |
| Plumbing & heating | 42/4691.3% | 6/4613.0% | 10/4621.7% | 32/4669.6% | 35/4676.1% | 14/4630.4% |
| Electricians | 47/5094.0% | 5/5010.0% | 10/5020.0% | 35/5070.0% | 45/5090.0% | 11/5022.0% |
| Accountants & bookkeeping | 43/4497.7% | 6/4413.6% | 12/4427.3% | 27/4461.4% | 40/4490.9% | 14/4431.8% |
| Dental practices | 45/4795.7% | 29/4761.7% | 40/4785.1% | 38/4780.9% | 40/4785.1% | 15/4731.9% |
| Hair & beauty salons | 35/4676.1% | 31/4667.4% | 28/4660.9% | 19/4641.3% | 39/4684.8% | 14/4630.4% |
| Restaurants & cafes | 43/4987.8% | 31/4963.3% | 3/496.1% | 11/4922.4% | 32/4965.3% | 16/4932.7% |
| Estate & letting agents | 46/4993.9% | 21/4942.9% | 17/4934.7% | 38/4977.6% | 33/4967.3% | 24/4949.0% |
| Garages & vehicle repair | 38/4486.4% | 17/4438.6% | 11/4425.0% | 22/4450.0% | 33/4475.0% | 10/4422.7% |
| Veterinary practices | 37/4288.1% | 25/4259.5% | 27/4264.3% | 21/4250.0% | 40/4295.2% | 12/4228.6% |
Online booking was most visible where the product is an appointment—hair & beauty salons (67.4%) and restaurants & cafes (63.3%)—and nearly absent in quote-led trades such as builders & construction (2.1%). A missing booking widget in a quote-led trade is not a defect; the sector split is the finding.
Dental practices led pricing visibility at 85.1%, while restaurants & cafes sat at 6.1%. Sectors where jobs are bespoke systematically avoid publishing prices; sectors selling repeatable services increasingly publish them.
No sector’s median mobile score reached 90. Even the fastest sector (garages & vehicle repair, median 51) sits well below what the same sites achieve on desktop, suggesting the mobile gap is a property of how these sites are built—heavy themes, unoptimised imagery, render-blocking scripts—rather than of any one industry.
The useful benchmark is not “add every feature.” It is whether the website makes the right next step visible for the sector’s actual buying journey.
Methodology
| Research question | Which of twelve specified website characteristics were machine-detectable on each business’s homepage, and how did each site perform under identical Lighthouse lab conditions? |
| Sampling method | Purposive, structured, non-probability sample: for each of ten sectors, organic web-search results for “<sector> <town>” queries across a rotation of UK towns and cities in all four nations were collected, then filtered to exclude directories, aggregators, review platforms, social profiles, national chains and franchises. The first 50 qualifying businesses per sector formed the register. The sample is not a census or random sample of all UK small businesses. |
| Sample frame | 500 named businesses: 50 per sector across Builders & construction, Plumbing & heating, Electricians, Accountants & bookkeeping, Dental practices, Hair & beauty salons, Restaurants & cafes, Estate & letting agents, Garages & vehicle repair, and Veterinary practices. The complete register appears below and is downloadable as CSV. |
| Included sample | 465 websites where the homepage was successfully fetched and coded. 35 sites were excluded (unreachable, timed out, blocked access, or broken HTTPS with no usable fallback) and were not coded as negative results. |
| Audit surface | The homepage HTML as delivered by the server to a desktop browser user agent, following up to six redirects. A plain-HTTP request was made separately to test HTTPS enforcement. Content injected by JavaScript after delivery was not visible to the audit and this is disclosed as a limitation. |
| Performance testing | Lighthouse (performance category), one mobile run with simulated 4G throttling and mobile emulation and one desktop run per site, executed locally in isolated headless Chrome processes on a single machine. Single runs on one day; lab variance applies. |
| Audit date | 13 July 2026. |
| Primary coding | Every homepage was coded by the same published rule set (see coding manual). No per-site human judgement was applied to criterion decisions, which makes the classifications reproducible from the retained evidence. |
| Evidence retention | The delivered homepage HTML of every included site and the raw Lighthouse JSON metrics were retained internally. |
| Quality assurance | All published numerators and denominators were recalculated from the raw result files by a separate script, and a random sample of evidence files was manually spot-checked against their criterion classifications before publication. |
| Reliability limitation | Automated coding applies the same rule identically to every site, but rules can misclassify individual sites (for example, a CTA phrase present only in an image). Rule-level definitions and known failure modes are published in the coding manual. |
| Uncertainty | Formal population confidence intervals are not presented because the study uses a purposive non-probability sample. Selection and exclusion bias may be more important than ordinary sampling error. |
Coding manual
The coding manual is public so that readers can understand exactly what counted and reproduce the classifications from the same inputs.
| Criterion | Counted as confirmed | Did not count | Interpretation |
|---|---|---|---|
| HTTPS enforced | The site's plain-HTTP address redirected to a valid HTTPS address, and the homepage was delivered over HTTPS with a certificate that passed standard verification. | Sites reachable only over plain HTTP, sites with failing certificate verification, and sites that serve HTTP without redirecting, even if an HTTPS version also exists. | Measures whether an ordinary visitor typing the bare domain ends up on a secure connection. It is not a full TLS-configuration audit. |
| Clear homepage call to action | The delivered homepage HTML contained visible text matching a defined list of action phrases (for example “get a quoteâ€, “book onlineâ€, “make an appointmentâ€, “contact usâ€, “request a callbackâ€, “book a valuationâ€). | Pages whose only next step was a bare navigation menu, a phone number with no invitation to act, or CTA text injected purely by JavaScript after load. | Measures whether the delivered homepage states an explicit next step. It does not judge the CTA's design, placement or persuasiveness. |
| Contact route | A visible link to a contact or enquiry page, a clickable telephone link, or a working contact form present in the homepage HTML. | A postal address alone, or contact details present only in hidden structured data. | Measures visible access to a way of contacting the business from the homepage. |
| Contact form on homepage | A form element with email, phone or name fields (or a recognised form-plugin signature) present in the delivered homepage HTML itself. | Search boxes, newsletter-only forms without contact fields, or forms that exist only on a separate contact page. | A stricter subset of the contact route criterion. |
| Online booking route | A visible booking link or an embedded booking tool (for example Calendly, Fresha, Treatwell, OpenTable, ResDiary, BookMyGarage or an equivalent “book online†route) detectable in the homepage HTML. | A generic contact form, a phone number, or wording such as “call to bookâ€. | Measures whether a visitor can start a booking without calling or emailing. Sector norms differ and are reported per sector. |
| Pricing visibility | A visible link labelled prices, pricing, fees or price list, or visible from-pricing (for example “from £49â€) in the homepage HTML. | Quote-request routes without prices, finance information, or pricing hidden behind a login. | Measures homepage discoverability of any pricing information, not the completeness or accuracy of the prices. |
| Testimonials or review signals | Visible testimonial content, a reviews/testimonials link, an embedded review widget, or a link to a recognised review platform (Trustpilot, Google reviews, Checkatrade, Reviews.io, Feefo) in the homepage HTML. | Unsupported claims such as “trusted by hundreds of customers†with no review content or route. | Measures the presence of social-proof signals, not their authenticity or volume. |
| Structured data (schema markup) | At least one JSON-LD script block or microdata itemtype attribute in the homepage HTML. | Open Graph or Twitter meta tags alone, which are social-preview tags rather than schema.org structured data. | Counted permissively: any schema.org structured data qualifies, including blocks auto-generated by SEO plugins. LocalBusiness-type markup is reported separately. |
| Cookie and privacy signals | Privacy: a visible link containing “privacyâ€. Cookies: a recognised consent-management platform signature (for example CookieYes, Cookiebot, OneTrust, Complianz, iubenda) or visible cookie-notice wording in the homepage HTML. | Privacy or cookie text present only inside PDF documents, or consent tools loaded in ways not detectable from the delivered HTML. | Measures the presence of standard compliance signals. It is not a legal assessment of UK GDPR or PECR compliance. |
| Social links | At least one visible link to Facebook, Instagram, X/Twitter, LinkedIn, YouTube, TikTok, Pinterest, Threads or Nextdoor in the homepage HTML. | Share buttons that post the page without linking to the business's own profile were not distinguished; this is disclosed as a limitation. | Measures the presence of a social route, not activity or follower counts. |
| Detectable accessibility problem (static) | One or more of: image tags missing alt attributes, a missing lang attribute on the html element, or visible text inputs with no label, aria-label, placeholder or referencable id. | Issues that require rendering or interaction to detect (contrast, focus order, keyboard traps, ARIA misuse) were NOT tested. | A deliberately narrow static check. It undercounts real accessibility problems and is not a WCAG conformance audit. |
| Outdated content signals | A visible copyright year of 2023 or earlier (reported against sites displaying any year), or visible COVID-19 wording still on the homepage (reported separately). | Sites showing no copyright year at all were excluded from the stale-copyright denominator rather than counted either way. | Proxy signals only. A stale footer year does not prove the business is inactive. |
| Performance and Core Web Vitals (lab) | Each included site was tested with Lighthouse (performance category) under mobile emulation with simulated 4G throttling, and separately under desktop conditions. “Passes lab Core Web Vitals thresholds†means lab LCP ≤ 2.5 s and lab CLS ≤ 0.1 on the mobile run. | INP cannot be measured in a lab load without user interaction and is not claimed. Field (real-user CrUX) data was not used. | Lab measurements on one machine on one day. Google's official CWV assessment uses field data; lab results are a diagnostic proxy and are labelled as such throughout. |
Public sample register
The register publishes business names, locations, homepage URLs and inclusion decisions. It deliberately does not publish per-business criterion scores and does not identify any named business as lacking a feature or performing poorly. The register below is loaded from the published CSV, which is the canonical version of record.
Loading the sample register… If it does not load, download the register CSV.
Limitations
Open research materials
These files make the sample, aggregate calculations, sector analysis and coding framework inspectable without publishing business-level negative classifications.
Munar, R. (2026), UK Small Business Website Statistics 2026, Vistoplex, Version 1.0, audit dated 13 July 2026, https://vistoplex.com/research/uk-small-business-website-statistics-2026/