# UK local search study 2026: data v1.4

Published 29 September 2026. Report: https://searchdecoded.co.uk/research/uk-local-search-2026/
Source: frozen v1.4-human-reconciled dataset and analysis. Source checksums are in data-dictionary.json.
results.csv contains 1,000 result observations. analysis.json contains full descriptive calculations, including category, top-three, sector, domain, Reddit and sensitivity measures.
Collection: 28 September 2026, 14:21:33–15:09:06 UTC. Five terms × 20 deliberately selected places. Same signed-in desktop session, UK country and English preferences, pws=0, one UK IP vantage point. Google showed non-personalised results; location and session effects are not eliminated.
Locations: Preston, Manchester, Newcastle upon Tyne, Leeds, Sheffield, Nottingham, Birmingham, Norwich, London, Brighton, Bristol, Exeter, Glasgow, Edinburgh, Inverness, Cardiff, Swansea, Wrexham, Belfast, Derry.
Results exclude ads, Maps and grouped modules. Rank is eligible organic order, not absolute on-screen position. 52 queries required page two; 60 positions came from that page.

Categories: independent-local is an observable business proxy, not certified legal SME status; large-chain uses documented scale or chain/group evidence; commercial-intermediary is a for-profit third-party discovery, review, booking or lead service; publisher-editorial is independent editorial content; forum-community includes standalone social profiles; other includes institutional/non-commercial roles and unresolved cases. A provider's own content follows its provider class. Independent/local includes national online sellers: three Prestige-group domains account for 24 positions. Separate domains are not necessarily separate companies.
Twelve uncertain cases affect 13 positions: ten unresolved-business positions and three unresolved-directory positions. Dickins has an unknown provider role within those ten business positions. The conservative commercial denominator is 216, not 215, when that role is included. Sensitivities are coding scenarios, not confidence intervals.
This is a purposive snapshot, not all Google UK results, traffic share or evidence of change over time. No directory lead/conversion economics or website-quality audit was conducted. AI-assisted coding and targeted human review of 47 cases are not independent blind double-coding.
The original method is retained in methodology.md. Subsequent authorised review created v1.1–v1.4 without changing search collection. Raw SERP captures, human excerpts and internal review records are private and are not included here. The public CSV excludes those fields. All original and derived study data are licensed CC BY 4.0: https://creativecommons.org/licenses/by/4.0/. Credit searchDecoded, link to the report and CC BY 4.0 licence, and indicate whether you made changes. Retain supplied attribution and notices; do not imply endorsement. This does not license third-party content, company names/trademarks, website copy, design or branding.