Business count by industry and region, UK: the counts, delivered as rows

The question gets answered twice, by two ledgers that measure different things on purpose.

The first ledger is the census. Each year the Office for National Statistics lifts a point-in-time extract off the Inter-Departmental Business Register - the frame of every UK business registered for VAT and/or Pay As You Earn - and publishes it as UK business; activity, size and location, the release that retired the old Business Register and Employment Survey branding. The March 2025 snapshot counts 2.73 million enterprises, up 0.4% year on year, operating 3.2 million local units. It exists to be counted: who is registered, where they trade, how large they run, in what legal form.

The second ledger is the statutory register itself: roughly five million live UK companies, one row apiece, each carrying the industry codes its directors declared at incorporation. Filter to a SIC class and the count stops being a statistic and becomes a list - 63,603 companies under the general building-cleaning code, every one with a name, a registration number and a registered office address.

The two universes disagree by design, and the disagreement is the value. Datadory delivers both layers normalized to one field dictionary, so the market-size denominator and the named universe arrive in the same shipment instead of two separate projects.

Two products carry the question, and they split the work cleanly.

UK Business Register and Employment Survey (BRES) owns the aggregate layer: enterprise and local-unit counts by four-digit SIC 2007 class, geography running from UK country down to parliamentary constituency, employment-size band, turnover-size band and legal status - 28 numbered tables of 414-plus rows per edition, editions reaching back to 1994. Datadory scores it 9 out of 10, carried by a field dictionary that maps fully to the published layout.

UK Companies House - SIC 80-82 Registrations owns the entity layer: one typed row per company with registration number, lifecycle status, legal form, incorporation and dissolution dates, up to four declared SIC codes and a postcode-resolvable registered office. It scores 8 out of 10, against a cross-catalog quality mean of 7.81 across 1,744 datasets.

Both sit on the diversified support services shelf - 21 cataloged datasets (16 primary plus 5 related) spanning statistical surveys, statutory registers and operational feeds. Per-record field dictionaries, coverage statements and sample rows live on the diversified-support-services data hub.

Rows - three support-service classes measured down the geographic cascade, with the register slice underneath:

SIC 2007 class label                 Measure                     UK      GB    Eng&W  England     NE
8010 : Private security activities   VAT/PAYE based enterprises  6980    6905   6690    6480    165
8110 : Combined facilities support   VAT/PAYE based enterprises  4430    4355   4090    3975    110
8121 : General cleaning of buildings VAT/PAYE based enterprises 13545   13295  12415   11895    400

Census totals, snapshot dated 14 March 2025: 2,730,000 enterprises (+0.4% y/y)
                                             3,200,000 local units  (+0.2% y/y)
------------------------------------------------------------------
sic_code  universe       status        companies
81210     all filings    all statuses  63,603
81210     all filings    active only   26,969
80100     all filings    all statuses  44,686
81100     all filings    all statuses  20,302
------------------------------------------------------------------
sic_code  status   incorporated  company_name                          registered_office
81210     active   2016-08-01    LAWRENCE DIVINE LTD                   Peterborough PE2 9SF
81210     active   2015-12-10    ANGEL CLEANING & MAINTENANCE LIMITED  London SE5 0EZ
80100     active   2010-03-23    786 SECURITY SERVICES LTD             Slough SL2 1LB

Read the pair together and the texture falls out. Each census row carries one industry class down a geographic ladder - United Kingdom, Great Britain, England and Wales, England, then any region you name - so a single cut returns the national figure and its regional spread in one pass. Cleaning is the volume story at 13,545 enterprises nationally; combined facilities support is the consolidation story at 4,430; private security sits between them at 6,980.

Then notice the register disagreeing with the census by nearly five to one on the same trade: 13,545 cleaning enterprises surveyed, 63,603 cleaning companies filed. Neither number is wrong. The census counts trading units above the VAT/PAYE thresholds; the register counts every incorporation ever lodged, dissolved shells included. Knowing which question each number answers is worth more than either figure on its own.

To the bottom of the administrative stack, on every axis at once. Geography runs from UK countries through the English regions, counties, districts and unitary authorities, down to parliamentary constituencies - and every level crosses with the four-digit SIC class, the employment and turnover bands, and legal status. The deliverable is therefore a matrix rather than a headline: building-cleaning enterprises in a named district, split by size band and legal form, comes off the same tables that produce the national total.

That depth maps onto real work. A consultant sizing service territories reads district-level counts for candidate sites, then rolls the identical classes up to English-region grain for the board deck without touching a second source. Policy and public-affairs teams brief MPs from the same class at constituency grain. Geography labels ship with ONS GSS codes - the unglamorous detail that makes regional cuts join cleanly to anything else keyed on geography, including the register layer below, where every company carries a postcode-resolvable registered office and the named prospects behind the regional total sit one filter away.

Quote the census when the question is market size - how many enterprises operate in a class and region. Quote the register when the question is screening - which named businesses exist, how old they are, whether they are still standing. Three divergences worth stating in any deliverable:

• Unit counted. The census counts businesses registered for VAT and/or PAYE, so sole proprietors above the turnover threshold appear alongside companies; the register counts incorporated entities only.
• Shape of the universe. Census figures are banded aggregates control-rounded to base 5; the register enumerates with no sampling and no estimates, across eleven lifecycle statuses from active through liquidation and administration.
• Grain. The census stops at the cell; the register goes to the row, with incorporation date, dissolution date and registered office attached to every company.

Used together they interlock rather than overlap: the aggregate denominator from one layer, the named numerator from the other, joined on geography and SIC class.

Enough to treat the count as a series rather than a snapshot. Annual editions of the business-register extract reach back to 1994, each holding the same 28-table layout, so a multi-decade trend line stays computable without rebuilding anything between editions.

The register contributes the other half of time-series work: incorporation and dissolution dates turn counts into cohorts. Group active firms by formation year and the sector's waves fall out with no third-party enrichment; subtract survivors from the historical total and churn becomes arithmetic - of the 63,603 companies that have ever declared the building-cleaning code, 36,634 are gone, roughly 58 percent. That is why dissolved rows ship in a delivery unless asked out: discarded wholesale, they take the survival signal with them.

Files, feeds, or straight into your warehouse. Daily, weekly, or hourly - your call.

Deliveries arrive normalized to the field dictionary: geography labels carrying their GSS codes, SIC classes decoded beside their raw labels, lifecycle statuses resolved to plain values, rounding and suppression notes written down rather than discovered mid-analysis. The sample comes first - name the classes, regions and bands, get real rows in the exact production schema, then decide. Request a sample scoped to your question.

• Market researchers and consultants size territories and TAMs from district-level enterprise counts, then defend the denominator when a client checks it (market researchers).
• Sales and growth teams convert a regional count into a ranked call sheet of named, active firms in the same footprint (sales growth teams).
• Investors and quant researchers read formation rates by trade and region as a small-business sentiment series ahead of official prints (investors quants).
• Data scientists inherit a labeled formation-and-dissolution panel with durable keys for churn and survival models (data scientists).
• Policy and public-affairs teams brief MPs with constituency-grain counts of the employers in the room.
• Journalists and academics cite the primary statistical records behind reporting on business dynamism (journalists academics).

Because the counts were never the hard part - reconciling them is. Two ledgers with different universes. Rounding to base 5 that makes small cells treacherous. Four-code industry arrays that flatten into wrong market shares. Registered offices pointing at accountants rather than depots. Dissolved firms that pollute a prospect list or, thrown away wholesale, delete the churn signal. Each is survivable once; none is fun to re-solve in every new notebook.

Datadory normalizes them before delivery: both layers shaped to one documented schema, coverage stated against the classes and regions you named, the quirks documented instead of waiting in the data. Prototypes built on the sample survive delivery intact - that is the point of shipping rows first.

Start with the two records behind every count on this page: UK Business Register and Employment Survey (BRES) for the enterprise tables and UK Companies House - SIC 80-82 Registrations for the company rows - each with its field dictionary, coverage statement and sample rows on the product page.

This cluster sits inside the diversified support services data guide, which maps all 21 cataloged records in the industry. The ranked shortlist lives on the best diversified support services datasets list and the full inventory on the diversified-support-services data hub.

Go deeper next door: UK SIC 81210 cleaning companies register works the company-level half of the same trade in detail. When you want real rows instead of descriptions, request a sample scoped to your SIC classes and regions - the field dictionary travels with it.