Why there is no official market total for England's web agencies

A researcher receives a request for an England-wide total for web-agency work, intended to support a market decision. The brief gives no service boundary, transaction unit or study period. The researcher treats it as a worked decision: define an observable transaction, inspect official records, build a bounded sample, and state what the evidence can support.

The request arrives without a measurable boundary

The researcher first asks whether the buyer wants a supplier population, transaction value or buyer expenditure. Those are distinct measures.

Since the brief asks for a market total, the researcher refuses to turn a broad business count into agency income and instead identifies what kind of transaction the proposed estimate should include.

That boundary matters because a website commission may combine research, content, software integration, hosting and maintenance. Official records classify a business by its principal economic activity, not every service bundled into an engagement. The evidence gap is therefore structural: relabelling a wide business category cannot produce a niche revenue total.

First, the researcher checks the official frame

The official record gives the researcher a frame around possible suppliers, but it does not answer the requested question. It describes registered business units and classifications, not web-agency income, the service mix sold, or spending by English clients on the chosen transaction.

The researcher opens the ONS 2025 business activity dataset. Its register snapshot is dated 14 March 2025; the release appeared on 24 September 2025. It covers registered businesses that are registered for VAT or PAYE and provides UK regions alongside four-digit UK SIC 2007 classes.

The linked quality and methodology material explains how the register is built and classifications are applied. An enterprise is the business unit; a local unit is a site where activity occurs. Adding both counts would duplicate some operations, and neither unit establishes that a business trades as an agency or sold web work.

The official SIC classification supplies the classification vocabulary, but a code is not a service audit. Potential suppliers may fall in classes 6201, 6202, 6209, 6312, 7311 or 7410, each of which also covers activity beyond web agencies. Some sole traders may be missing if they are not VAT- or PAYE-registered.

The ONS business-demography method uses reference-year comparisons between register files to identify active businesses and business births or deaths. Those events do not identify sales leads or purchases of outside web work. The separate market outlook is not a measure of this transaction sample.

Next, the unit is made transaction-sized

The researcher chooses accepted engagements for a defined web service, during a stated period, connected to English buyers. That transaction unit can be evidenced through contracts or invoices, unlike an enterprise count; it also limits the result to the sample and period actually observed.

Before collecting records, the researcher writes down English buyer location, included service, minimum evidence and study period. The candidate populations below answer different questions; only accepted engagements match the transaction measure selected for this scenario. The comparison prevents an entity or location count from being quietly presented as a count of sales.

The comparison now has a practical consequence: an enterprise is not a website, incorporation does not prove trading, and a site does not automatically identify the party signing. Accepted engagements fit the question more closely, but they still describe only the observed sample and period.

Before counting, the researcher distinguishes supplier count, net fee income, buyer expenditure and contract value. Contract value can include hosting, licences or pass-through production, while gross revenue may count subcontracted work twice. These measures cannot be swapped without changing the question.

The researcher verifies each candidate's current service page and legal identity before inclusion. English buyer location, included service, minimum evidence and study period are recorded beside an evidence source. That record makes it possible to explain why an entry was accepted or excluded.

Then, the sample rules are written down

The researcher keeps the sample method tied to the chosen transaction boundary. Each candidate must show a current service offer and evidence of an accepted engagement for that defined service; a directory listing alone cannot stand in for a transaction record.

To keep the screen repeatable, the researcher follows a benchmark protocol and checks it against a selection checklist. The method records how the sample was selected and what evidence supports inclusion, rather than treating the official business count as the sample itself.

Finally, the result stays within its evidence

Once the sample has been reviewed, the researcher reports a sampled value, not an England-wide market total. The result states its measure, period, currency, VAT treatment, inclusions, refunds, subcontracted spend and uncertainty, and explains how the sample was selected.

eligible transactions in the period x observed net fee per comparable transaction = sampled value

For illustration, 40 eligible transactions at a net fee of £2,500 each produce a sampled value of £100,000. These are placeholders, not market evidence. The researcher leaves both inputs unset until records have been checked, then reports the observed sample rather than claiming that the arithmetic estimates the whole England market.

The researcher also reports response rate and selection bias, rather than extrapolating silently from a directory. Website counts, domain registrations, digital-advertising expenditure and global market reports measure different things. An England population share cannot repair the mismatch or convert any of those quantities into a web-agency total.

Rules that survive the estimate

The conclusion carried out of the exercise is simple: identify the unit, measure and period before counting, then keep the claim within the evidence gathered. A national niche total remains unknown; the defensible output is a clearly bounded account of the observed sample and its limits.

When records cannot separate those amounts, describe the recorded value without calling it agency income. The useful outcome is a transparent account of what was counted and what remains unknown; it is not an official business count presented under a narrower name.