LOCAL WEBSITE HEALTH — ANONYMISED DATASET Scorpion AI Tooling Ltd · published 2026-08-20 FILE local-website-health-2026-08-20.csv (88 rows, UTF-8, comma-separated) WHAT IT IS A technical health score for each of 88 independent small businesses within roughly eight miles of Newcastle-under-Lyme, Staffordshire, England (taking in parts of Stoke-on-Trent and surrounding villages), surveyed July-August 2026. Each row carries its individual criterion scores as well as the total, so the working can be checked. Accompanying write-up: https://scorpionaitooling.co.uk/blog-local-website-health-2026.html HEADLINE FIGURES 88 businesses surveyed. 65 (73.9%) scored less than full marks. Three definitions of "not doing the business justice" -- less than full marks; dropped a point on any criterion or no working site; failed search, upkeep or AI-readability or no working site -- all select the same 65 rows. 72 websites loaded and were scored on every criterion. 16 failed the "live and real" gate, which zeroes the whole rating. Among the 72 that loaded: findable in search 40 scored zero (55.6%) 32 full marks (44.4%) kept up to date 23 scored zero (31.9%) 49 full marks (68.1%) readable by AI 16 scored zero (22.2%) 56 full marks (77.8%) current build 6 scored zero (8.3%) 66 full marks (91.7%) served over HTTPS 3 scored zero (4.2%) 65 full marks (90.3%) 4 partial works on a phone 3 scored zero (4.2%) 68 full marks (94.4%) 1 partial enough content 1 scored zero (1.4%) 71 full marks (98.6%) Of the 40 that failed the search criterion, their notes cite: h1 39, meta description 38, title 36, canonical 28, robots/sitemap 25. The 16 gate failures, tiered — please quote the tier you mean: 16 (18.2% of 88) had a domain that failed the check 13 (14.8%) ...and had no other web presence found 8 (9.1%) ...and also had positive evidence of still trading Of the three with a presence elsewhere: two redirect to a marketplace shop, one has a live maintained site on their web developer's subdomain. Of the thirteen, eight had trading evidence, four had none either way, and one appears to have ceased trading (published phone confirmed disconnected). COLUMNS ref Opaque sequential identifier. Carries no meaning and cannot be resolved back to a business. score Technical health total, 0-10. band Plain-language label for the score range. gated 1 if the site failed the "live and real" gate. A gated row is recorded 0 across every criterion BY RULE, not by assessment — do not read a gated row's zeros as measured failures on those axes. shortlisted 1 if the score is 6 or below. award_reachable Per-criterion scores. For every non-gated row these sum award_https exactly to `score`; this holds on all 72 and is the award_mobile intended way to verify the dataset. award_build award_content award_upkeep award_search award_answers sector Broad category, derived from the business name and the name then discarded. "Other / unclassified" where no category fit. The two smallest categories are present in the rows but suppressed in any published breakdown. distance_band Approximate distance from the survey centre, in miles. "unrecorded" for earlier sweeps that did not capture it. search_note_cites_* Five columns (h1, meta_description, title, canonical, robots_sitemap). NOTE-DERIVED, not measured axes: 1 means the assessor's written note for the search criterion mentions that item. Treat these as a FLOOR, since a note only records what its author chose to write about. Blank on gated rows. Published so the sub-counts quoted in the write-up can be checked against the rows. gate_failure_mode Gated rows only: how the domain failed. other_web_presence Gated rows only: whether the business was found online anywhere else. trading_evidence Gated rows only: whether there was positive evidence the business still trades. "not evidenced" means no evidence was found either way, NOT that the business has closed. SCORING RUBRIC (10 points) reachable 1 Loads on its own domain and serves real business content. GATE: 0 here zeroes the whole score. https 2 HTTPS resolves cleanly (1 if faulty, 0 if http only). mobile 2 Viewport meta present AND real media queries (1 if only one). build 1 Current platform and markup; no pre-2015 jQuery, no table layout. content 1 Enough visible text to say what the business does. upkeep 1 Still maintained: current copyright, no abandoned pages. search 1 On-page basics: sensible title, exactly one h1, meta description, canonical, reachable robots.txt or sitemap. answers 1 Answer-visibility crawlers not blocked and structured data present. Blocked AI TRAINING crawlers are NOT counted as a fault, since blocking them is a legitimate choice. METHOD - Two-way presence check per business: a directory and aggregator sweep cross-checked against a direct business-name-plus-settlement search. No business was recorded as having no website on a directory check alone. - Every DNS failure was re-confirmed against an external resolver (Cloudflare 1.1.1.1) with a known-good control domain in the same query, so a lapsed registration is distinguished from a local resolver fault. - Sites returning 403 to the scoring script were re-fetched with a browser user-agent and their stylesheets walked manually, so a block never counted against a site. - Criteria that could not be judged were named as such, not scored 0. - The scored signals are computed by a script from the fetched page: HTTPS behaviour, viewport, media query counts, platform and library detection, markup structure, text length. Discovery and the written assessment notes were produced with AI assistance and reviewed. Every figure published is a count of recorded criterion scores, not a count of statements in the notes. LIMITS — PLEASE CITE THESE ALONGSIDE ANY FIGURE - One geographic area, rings out to roughly 8 miles from Newcastle-under-Lyme, Staffordshire. Not a national picture. Per-business geography is reduced to a distance band; naming the survey centre does not identify any individual business in an 88-row sample spread across many settlements. - One snapshot, July-August 2026. No trend data, so nothing here supports a claim about conditions improving or worsening. - Not a random sample. Independent SMEs only. Later sweeps deliberately excluded food and hospitality, computer and IT, convenience and grocery, ceramics manufacturers, and professional services. - Businesses with no web presence anywhere are under-represented, because the method finds businesses through listings. This biases the dead-domain figure DOWNWARD. - n=88 supports coarse proportions. It does not support ranking sectors against one another. - The content and search criteria are homepage-only judgements: the script reads one page, so a site could have a correct h1 on inner pages and still score zero here. - A technical score, not a design or usability review. A site can score 10 and still be unpleasant to use. - Trading evidence is not uniform across the gated rows; the column records which it is per row. WHAT IS EXCLUDED Three ratings of our own client work (one pre-build baseline, two periodic re-checks) and our own website. Including sites we built would flatter the figures. SOURCES AND ACCESS Every input came from information the surveyed businesses publish about themselves, freely accessible to anyone with a web browser. - One page per business (the public homepage), requested over the ordinary web exactly as a visitor's browser would. No logins, paywalls, member areas or customer portals were involved. - No access control was bypassed and nothing was guessed or forced. Where a website refused the scoring script's default user-agent, it was re-fetched with a standard browser user-agent, returning the same public page that site serves to every visitor. - Directory, review and social listings were read as published, and used only to establish whether a business was still trading. - DNS resolution checks are queries against public name servers, the same lookup any browser performs before loading a page. - No personal data beyond what a business publishes about itself was collected, and none of it appears in this dataset regardless. PRIVACY No business names, domains, URLs, proprietor names, telephone numbers or street addresses are included. Settlement names are omitted and geography is reduced to a distance band, because a named trade in a named village is re-identifiable even without a business name. The published dataset was checked programmatically against the source records to confirm none of these identifiers survived. LICENCE Creative Commons Attribution 4.0 International (CC BY 4.0). Free to share and adapt, including commercially, with attribution. https://creativecommons.org/licenses/by/4.0/ CITATION Scorpion AI Tooling Ltd (2026), "Local Website Health: 88 Small Businesses Surveyed", dataset. https://scorpionaitooling.co.uk/data/local-website-health-2026-08-20.csv CORRECTIONS If you find an error, or reach a different figure from these rows, please get in touch at hello@scorpionaitooling.uk. Corrections will be noted here with a date.