Sizing the England marketing automation market: the ONS ceiling and a six-row range method

The official ceiling for England

The ONS bulletin UK Business: Activity, Size and Location 2025 counts 2,734,615 VAT and/or PAYE registered businesses in the UK at March 2025. Summing the nine English regions in its regional table gives roughly 2.4 million for England, led by London at about 538,000 and the South East at about 405,000. That is a ceiling on possible buyers, not a market size.

Mistaking the broad business population for the market

The costliest mistake is treating a broad business population as the number of likely buyers.

That turns a contextual count into market demand, although the source does not measure platform need, usable data, category revenue or purchasing intent. Present the count only as a boundary, then define a narrower commercial audience.

The business population estimates are useful for establishing context, but their private-sector coverage and head-office basis do not show which organisations have a marketing automation problem. State these limits beside the baseline; the authority of an official source should not imply buyer demand.

Set the buyer boundary around a real workflow problem, suitable customer data and a need for journey software. The target audience should reflect the offer being considered, while exclusions and any separate charity or public-body population remain visible rather than being silently folded into one total.

Counting sites as independent buying opportunities

Counting every site as a separate opportunity can inflate the apparent market when purchase authority sits centrally. It also changes the unit being modelled: an enterprise is an organisation, while local units describe operating sites that may not make or materially influence a platform purchase.

The ONS business activity dataset gives enterprise and local-unit counts by region, industry and employment band. Use its business activity and location data only after choosing the relevant statistical unit, and retain the edition used. Do not treat a site count as a count of independently authorised buyers.

Write down the geography, SIC classification, employment band and unit before selecting a count. Keep the dataset edition with the working file, so another analyst can identify the population represented and reproduce the choice. If the model changes from enterprise counts to local units, show that as a changed assumption, not a quiet update.

Calling commercial assumptions official rates

Calling eligibility, readiness and reach official rates hides that they are commercial assumptions. A published business count cannot establish whether a workflow is suitable, whether usable customer data exists, or how many accounts a team can reach through its channels during the period under review.

Define the filters separately before applying any proportion. Eligibility concerns a repeated marketing job that automation could handle; readiness concerns usable identifiers and permission states; reach concerns accounts accessible through named channels and the team's capacity to pursue them.

Use discovery records and a documented sample to support any proportions. Do not describe an untested proportion as official statistics; where evidence is absent, call it a provisional assumption. Keeping the filters distinct makes it possible to see whether uncertainty comes from workflow fit, data condition or practical access.

Turning an illustration into a market finding

A worked calculation can demonstrate how assumptions interact, but its output does not prove that the assumed accounts or purchase rates exist. Treat an illustrative result as a demonstration of method, not as measured England revenue, and do not promote a midpoint into an official market figure.

The England market method sets out a labelled illustration and its arithmetic; for a local estimate, preserve the origin of each input and describe the output as illustrative until evidence supports it. Avoid repeating the example as if its assumptions were observations about the whole market.

The six rows of the range

The range has six rows. Each row takes a low, middle and high input. All inputs below are assumed for illustration, apart from the source of the segment count in Row 1, and the output is a method demonstration, not an England market figure.

Multiplying through gives about £9,600 (low), £87,000 (middle) and £342,000 (high) a year for this one segment. The spread, roughly a factor of 35, is the finding: it shows how little the inputs pin down. Replace the 6,000 with the England rows from the dataset, and replace each percentage only with evidence from discovery records.

Combining recurring sales with one-off income

Combining different income types can make repeatable revenue appear larger than it is. Keep software, implementation and managed services on separate lines, include only income the proposed model actually earns, and base annual value on a dated quote or the supplier's own offer.

Match each amount to its charging unit rather than assuming suppliers sell the same thing. The business models use different charging units, while a return on investment review should distinguish repeatable value from one-off fees.

Trusting a range that has not been stress-tested

A low, middle and high output is not useful if nobody can challenge the assumptions behind it. A range can look reassuring while depending on one fragile input, so test whether a modest change in data readiness, reach or support costs would alter the commercial case.

Vary one assumption at a time and record the effect on the result. If a small shift removes the case for proceeding, treat that variable as a priority for investigation rather than smoothing it away inside an average. Rehearse the assumptions against relevant risk scenarios before polishing a forecast.

Where charities or public bodies may also buy, keep them separate from private-company denominators when the source covers those populations on a different basis. Likewise, do not count every location within a centrally purchasing organisation as an independent buyer. Make exclusions clear enough that a reviewer can see their effect on the range.

Leaving no path to recalculate the result

A range loses credibility when its inputs cannot be traced to a dataset, discovery record or dated commercial offer. Save the selected England rows, note the dataset edition, write down exclusions, and record the method and sample behind any observed proportions.

Replace provisional assumptions only when the evidence and its collection method are recorded. Keep the working figures in a budget template, so another analyst can recalculate and challenge them later. The defensible result is a range with traceable uncertainty, not a single national value presented as fact.