THE COMMERCIAL GIGABIT GAP: THE UK PREMISES THAT CANNOT ORDER A GIGABIT BROADBAND CONNECTION Published dataset - README =============================================================================== File uk-commercial-gigabit-gap-2026.csv Rows 1,011 data rows plus one header row Unit one row per published geography: 361 local and unitary authorities and 650 Westminster parliamentary constituencies. These are TWO complete coverages of the same country, not 1,011 separate pieces of it: the authority rows cover every UK premise and so do the constituency rows. Filter on geography_level before aggregating, or a whole-file sum counts every premise twice. Snapshot January 2026, compared against July 2025 Encoding UTF-8, LF line endings sha256 9b567123f27fa6615f1b66d6660140d192866e7aacf8c56b0fe6208ca7f4b303 Personal data none. The dataset contains no individual, no contact detail and no premise-level or postcode-level record; every figure is an area total. WHAT WAS MEASURED ------------------------------------------------------------------------------- Ofcom publishes fixed broadband coverage twice: once for all premises, and once for residential premises only, as separate files at every geography. It does not publish the difference below nation level. The difference is the commercial premise base, and this dataset is that difference, computed for every local authority and every Westminster constituency, alongside the residential figures it should be read against. WHAT A COMMERCIAL PREMISE IS, AND IS NOT ------------------------------------------------------------------------------- A commercial premise is what is left when Ofcom's residential-only count is taken away from its all-premises count. It is dominated by properties Ordnance Survey AddressBase classifies with a code beginning 'C', but Ofcom's premise base also admits classifications for dual-use property, places of worship, stately homes and mail-only addressees. Ofcom reports only two premise types and the split is exhaustive, so those records must sit in one of the two and Ofcom does not publish which. The commercial figure is therefore the complement of the residential one and is not confined to purely commercial property. This is stated rather than quantified, because Ofcom publishes no allocation. It is a property class. It is NOT a count of trading organisations, and must never be reported as one. Ofcom counts firms a different way - an address match to the Dun & Bradstreet Business Universe - and gets a different and higher coverage figure. Of the difference Ofcom says only: "Coverage remains highest for micro businesses, which are often based in residential areas and can make use of residential services." Being based in a residential area is not the same as holding an address AddressBase classifies as residential, and Ofcom publishes no figure for how many trading premises sit in the residential count. METHOD ------------------------------------------------------------------------------- 1. Both Ofcom archives were downloaded once, hashed, and are re-hashed on every run. The analysis reads the CSVs straight out of the archives and makes no network request. 2. The subtraction was proved before it was used. Ofcom publishes all three premise types at UK and nation level, split urban and rural, across eleven speed rows. The identity Commercial = All minus Residential was tested on every published cell in both releases and holds exactly, to the unit, with no failures. Had one cell failed, the study would have stopped. 3. The same subtraction was then applied to the 361 local-authority rows and the 650 constituency rows in each release. Joins between the all-premises and residential files are complete in every case. 4. The derived rows were summed back to Ofcom's published UK totals, from each geography independently, for premise base, gigabit count and unmatched count, for all three premise types. Every one reconciles exactly. 5. The six-month change is anchored on the 650 constituency codes, which are identical across the two releases. Two local authorities were recoded (Barnsley E08000016 to E08000038, Sheffield E08000019 to E08000039) and are joined on an explicit code map. SERIES BREAKS AND REVISIONS ------------------------------------------------------------------------------- - The premise base is rebuilt from a refreshed AddressBase at each snapshot, so the six-month change is a change in the published measure, not a tracked panel. - The January 2026 release upper-cases every area name; the July 2025 release does not. Every join in this work is on ONS code, never on name. - The constituency CSVs are Windows-1252 encoded while the rest are UTF-8. - The constituency name column is renamed between releases, and the July 2025 'below the USO' columns are replaced in January 2026. Neither is used here. - The January 2026 archive is a reissue (v2). Ofcom's correction covers the 121 all-premises postcode files only. This study uses no postcode file. LIMITS ------------------------------------------------------------------------------- - Coverage is availability - whether an end user could order a service - not installed service or take-up. - Ofcom's data covers broadband products. Ofcom's methodology says that dedicated leased lines are not included in its coverage data. A commercial premise counted here as unable to order a gigabit broadband connection may hold a leased line. No adjustment is attempted, because per-premise leased-line take-up is not published and any correction would be a guess. - Unmatched premises are unclassified and are not counted as gigabit-capable. The commercial unmatched rate is several times the residential one, so the file carries if-all-unmatched-capable columns computed on the assumption that every unmatched premise turns out to be gigabit-capable. Those columns are the LOWEST the shortfall could be. Moving both segments at once is not the hardest test of the gap between them; moving only the commercial side is, and the study publishes that case at UK level. - Ratios in this file are computed from the four count columns, not by dividing the two rounded percentage columns, so recomputing them from the percentages as printed gives a slightly different answer. The gap_pp columns are the other way round: they ARE the difference of the two published percentage columns. - Ofcom publishes rurality only in its UK-and-nations file. There is no urban or rural split for any authority or constituency, so none is given here. - No percentage in this study is computed on a denominator below 10. The smallest commercial premise base in the file is 589 (Isles Of Scilly). - Any ranking uses a published floor of 2,000 commercial premises, which keeps 340 of 361 authorities and 570 of 650 constituencies. A low percentage in a tiny premise base is noise. - The January 2026 snapshot will be superseded by the next Connected Nations release, expected around November 2026. Refresh or date this file clearly at that point. COLUMNS ------------------------------------------------------------------------------- geography_level local_authority (361 rows) or westminster_constituency (650 rows). ons_code ONS geography code. Join key. Codes are for the January 2026 release. area_name Ofcom's area name for the January 2026 release, which upper-cases every name. nation England, Scotland, Wales or Northern Ireland, from the first letter of the ONS code. ofcom_snapshot The snapshot the main figures describe: January 2026. all_premises Ofcom's all-premises base for the area. Published by Ofcom. all_premises_matched Published by Ofcom. Ofcom defines this column as the "Count of all premises that could be matched to at least one fixed provider record with a non-zero predicted download speed. Unmatched premises are unclassified" (Ofcom, About this data - fixed coverage and full-fibre take-up (v2)). all_premises_unmatched all_premises minus all_premises_matched. Unmatched premises are unclassified and are not counted as gigabit-capable. all_premises_with_gigabit Published by Ofcom. all_premises_without_gigabit all_premises minus all_premises_with_gigabit. all_premises_without_gigabit_pct Denominator: all_premises. residential_premises Ofcom's residential-only premise base. Published by Ofcom. residential_premises_matched Published by Ofcom. residential_premises_unmatched residential_premises minus residential_premises_matched. residential_premises_unmatched_pct Denominator: residential_premises. residential_premises_with_gigabit Published by Ofcom. residential_premises_with_full_fibre Published by Ofcom. residential_premises_without_gigabit residential_premises minus residential_premises_with_gigabit. residential_premises_without_gigabit_pct Denominator: residential_premises. residential_premises_without_gigabit_pct_if_all_unmatched_capable As above, but assuming every unmatched residential premise turns out to be gigabit-capable after all. This is the LOWEST the residential shortfall could be, not the highest. Denominator: residential_premises. commercial_premises DERIVED: all_premises minus residential_premises. Dominated by premises Ordnance Survey AddressBase classifies with a code beginning 'C', but see WHAT A COMMERCIAL PREMISE IS, AND IS NOT above: it is the complement of the residential count and so also carries the dual-use, place-of-worship, stately-home and mail-only records Ofcom does not publish the allocation of. NOT a count of trading organisations. Ofcom publishes this figure for the UK and the four nations but not for any area in this file. commercial_premises_matched DERIVED: all_premises_matched minus residential_premises_matched. commercial_premises_unmatched DERIVED: commercial_premises minus commercial_premises_matched. commercial_premises_unmatched_pct Denominator: commercial_premises. commercial_premises_with_gigabit DERIVED: all_premises_with_gigabit minus residential_premises_with_gigabit. commercial_premises_with_full_fibre DERIVED, by the same subtraction. commercial_premises_without_gigabit commercial_premises minus commercial_premises_with_gigabit. commercial_premises_without_gigabit_pct Denominator: commercial_premises. commercial_premises_without_gigabit_pct_if_all_unmatched_capable As above, but assuming every unmatched commercial premise turns out to be gigabit-capable after all. This is the LOWEST the commercial shortfall could be, not the highest. Denominator: commercial_premises. gap_pp_commercial_premises_minus_residential_premises The published commercial_premises_without_gigabit_pct column minus the published residential_premises_without_gigabit_pct column, in percentage points, so the subtraction can be checked on the two columns as printed. Negative where commercial premises fare better than homes. gap_pp_commercial_premises_minus_residential_premises_if_all_unmatched_capable The same gap computed on the two if-all-unmatched-capable columns, i.e. both segments moved at once. This is NOT the hardest case for the finding: that is the commercial column moved on its own, which the study publishes at UK level. ratio_commercial_premises_to_residential_premises_without_gigabit How many times more likely a commercial premise is than a home, in this area, to have no gigabit broadband option. Computed from the four count columns (commercial_premises_without_gigabit / commercial_premises, divided by residential_premises_without_gigabit / residential_premises), NOT by dividing the two rounded percentage columns, so it can differ in the second decimal place from that division. The gap_pp columns are the other way round: those ARE the difference of the two published percentage columns, so that every printed row is internally consistent. commercial_premises_jul2025 DERIVED from the July 2025 release by the same subtraction. commercial_premises_with_gigabit_jul2025 DERIVED from the July 2025 release. commercial_premises_without_gigabit_pct_jul2025 Denominator: commercial_premises_jul2025. commercial_premises_without_gigabit_pct_change_pp January 2026 minus July 2025, in percentage points. Negative means the shortfall is smaller at the later snapshot. residential_premises_without_gigabit_pct_jul2025 Denominator: the July 2025 residential premise base for the area. residential_premises_without_gigabit_pct_change_pp January 2026 minus July 2025, in percentage points. gap_pp_jul2025 The gap at July 2025. gap_pp_change_pp Change in the gap over the six months. previous_ons_code_jul2025 Filled only for the two authorities recoded between releases (Barnsley, Sheffield). Empty otherwise. six_month_join_method ons_code, or ons_code_remapped for those two authorities. meets_min_size_floor_2000_commercial_premises yes where the area has at least 2,000 commercial premises. No ranking in this study includes an area marked no. comparison_snapshot The snapshot the change columns compare against: July 2025. SOURCES, LICENCE AND ATTRIBUTION ------------------------------------------------------------------------------- Ofcom, Connected Nations update: Spring 2026 - fixed coverage and full-fibre take-up data (January 2026 snapshot). https://www.ofcom.org.uk/phones-and-broadband/coverage-and-speeds/connected-nations-update-spring-2026 Ofcom, Connected Nations 2025 - fixed broadband coverage data (July 2025 snapshot). https://www.ofcom.org.uk/phones-and-broadband/coverage-and-speeds/connected-nations-20252/data-downloads-2025 Ofcom, About this data - fixed coverage and full-fibre take-up (v2). https://www.ofcom.org.uk/siteassets/resources/documents/research-and-data/multi-sector/infrastructure-research/connected-nations-spring-2026/about-this-data-fixed-broadband-coverage-and-full-fibre-take-up-2026.pdf Ofcom, Connected Nations 2025: Methodology annex. https://www.ofcom.org.uk/siteassets/resources/documents/research-and-data/multi-sector/infrastructure-research/connected-nations-2025/cn2025---methodology-annex.pdf Ofcom, Connected Nations UK report 2025 (Table 2.3, page 10, for the SME figure). https://www.ofcom.org.uk/siteassets/resources/documents/research-and-data/multi-sector/infrastructure-research/connected-nations-2025/connected-nations-uk-report-2025.pdf Ofcom, Connected Nations Spring 2026 update tables. https://www.ofcom.org.uk/siteassets/resources/documents/research-and-data/multi-sector/infrastructure-research/connected-nations-spring-2026/connected-nations-spring-2026-update-tables.xlsx The licence sentence below appears in the two 'About this data' documents. The other Ofcom documents cited carry no licence statement of their own; Ofcom's site terms of use and the attribution condition below apply to them. Ofcom's SME coverage figures, quoted once in this study so the two measures are not confused, derive from the Dun & Bradstreet Business Universe, which Ofcom states it uses subject to three attributions, carried here for the same reason: (c) Dun & Bradstreet Inc., 2025. All Rights Reserved (c) Crown copyright. All rights reserved. Licence number 10002057 (c) CACI 2025 Ofcom's licence wording, quoted exactly and carrying no version number: "We are providing this data on an open basis via the Open Government Licence, which gives users various freedoms about how they choose to use the data, subject to conditions." Ofcom's terms of use add: "The material must be acknowledged as Ofcom copyright and the title of the document/publication specified." Credit line for any reuse of the source data: Source: Ofcom, Connected Nations update: Spring 2026, fixed coverage and full-fibre take-up data (January 2026 snapshot), and Connected Nations 2025, fixed broadband coverage data (July 2025 snapshot). Ofcom copyright. Ofcom provides this data on an open basis via the Open Government Licence. This derived dataset is published under Creative Commons Attribution 4.0 International (CC BY 4.0), https://creativecommons.org/licenses/by/4.0/. Attribute as: Servnet, The commercial gigabit gap, https://www.servnetuk.com/research/uk-commercial-gigabit-gap-2026. Derived from Ofcom Connected Nations data; Ofcom copyright. REPRODUCING IT ------------------------------------------------------------------------------- python3 scripts/research/uk-commercial-gigabit-gap-2026/analyse.py It reads only the two archives already stored under data/research/uk-commercial-gigabit-gap-2026/raw/ re-hashes them, and rewrites this file and the JSON the page renders. It makes no network request and it is safe to run repeatedly.