How to use ONS and Companies House data to size a UK social listening market

What to take away

Why platform estimates are a weak base for market claims

Platform estimates are useful for direction, not for market size. They count accounts, impressions or keyword mentions, and each vendor defines those differently. None of them knows how many UK companies have a budget line for social listening. Their totals describe conversation, not purchasing power.

If you build a market claim on mention volumes, you inherit the platform's sampling bias. A campaign trending on one network looks larger than a quieter B2B conversation that actually drives software spend. Reading a mention count as a market size is one of the classic measurement mistakes in this field.

UK buyers also sit behind rules that platform counts ignore. UK GDPR and the Data Protection Act 2018 shape what can be monitored, and the ICO enforces them. The data protection duties also change who buys: a media company may need monitoring for compliance evidence, while a retailer may need it for campaign response.

Use platform estimates to show relative movement, such as whether conversation about a category is rising. Use official counts to show how many organisations could plausibly buy. The two answer different questions, so keep them in separate columns of the model and never multiply one by the other to produce a total.

ONS datasets that ground UK digital and social media sizing

Start with the ONS digital economy collection, which holds data on technology adoption in businesses and households. It gives you a defensible count of digital-active firms and works as the base layer for any UK digital market claim, provided you note which survey each table comes from.

Next, take the home internet and social media usage data. It tells you how many UK adults use the internet and social media, and how often. That is your audience denominator, not your buyer count. The first dataset shows how many firms could buy, and the second how many people they might monitor.

ONS data is survey-based and published with confidence intervals. Treat the figures as a range, not a point. If the survey says a proportion of firms use social media, apply that proportion to your company count and keep the interval visible so readers can see how wide the uncertainty is.

Regional splits matter. Greater London, the North West, Scotland and the West Midlands have different business densities, and so do Yorkshire and the Humber, Wales, the South West and the North East. Use ONS regional business counts rather than assuming a national average applies everywhere.

ONS categories also change between releases. Read the methodology note for each dataset before you chain two years together, because a definition change can look like growth. Record the release date next to each figure so a later analyst knows which version you used.

Companies House filings as a count of addressable companies

Companies House is the UK's registrar for limited companies, limited liability partnerships and other bodies. Active companies file accounts and a confirmation statement, which gives you a live population to count from. The Companies Act 2006 sets the filing and disclosure framework you are reading.

Use the free Companies House search and its advanced search by SIC code. Codes 62020 (information technology consultancy), 73110 (advertising agencies) and 73200 (market research and public opinion polling) are the obvious starting filters. Companies House works from a condensed list of SIC codes taken from the ONS full list, so check each code against it.

Those codes overlap. A digital agency may sit under 62020 and also do monitoring, and a PR firm may sit under 70210. Combine codes and accept some double counting risk, then note the overlap in your method record so a reviewer can see how large it might be.

Region filters are built into the advanced search, so you can count companies registered in England, Wales, Scotland and Northern Ireland. Postcode filters then let you drill into Greater London, Manchester, Birmingham, Leeds and Glasgow without leaving the same dataset or changing the counting rules.

Filter by company status. Active companies are your addressable base, while companies in liquidation, dissolved or dormant are not buyers, and including them inflates the market. Filing category is a crude size proxy, since a company filing full accounts is more likely to hold a marketing budget than one filing micro-entity accounts.

Record the date you pulled the count. Companies House numbers move daily as new companies incorporate and others dissolve, so a count without a pull date cannot be repeated. Pull on the same day each quarter so every comparison is like for like.

People with significant control data for ownership analysis

People with significant control data, usually called PSC data, records who owns or controls a UK company. Under the statutory PSC guidance, a person qualifies by holding more than 25% of shares or voting rights, or the right to appoint or remove a majority of the board.

This matters because company counts overstate buyer counts. A group with one marketing function may have five subsidiary companies, each registered separately, and counting all five as separate buyers doubles or triples your market. If the same corporate PSC appears across several registrations, treat them as one buying entity for budget purposes.

PSC data also reveals foreign ownership. A UK-registered company owned by an overseas parent may buy through the parent's procurement, not locally, which changes the addressable spend. Combine PSC grouping with SIC and region filters to get distinct buying entities in a sector and place, which is closer to a market than a raw company count.

There are limits. PSC data does not cover sole traders, and it does not tell you revenue, headcount or marketing budget. It is a grouping tool, not a sizing tool on its own, so pair it with the ONS proportions and a stated price assumption before you quote any figure.

A worked example sizing a UK social listening market

The figures in the worked example are illustrative, but every step uses a named source that anyone can re-run. Take the steps in order, and write down each count, filter and pull date as you go so the calculation can be repeated later.

Suppose the filtered company count is 12,000 active entities across the three SIC codes. PSC grouping removes 2,000 as subsidiaries of other entities, leaving 10,000 distinct buyers. If the ONS digital activity proportion suggests 60 per cent of firms in that group are digitally active, the serviceable base is 6,000 entities.

At an assumed 5 to 15 per cent adoption, that is 300 to 900 buyers. At an assumed £6,000 to £20,000 annual contract value, the implied market sits in a wide range. Present it as a range with the assumptions visible, because a single number invites false confidence and hides the judgement inside the adoption rate.

The buyer count is auditable, while the adoption rate is a judgement you must label. Vendors price by seats, usage or add-ons, so test your contract value against the pricing models in use. For a 2027 view, extend the same method with a stated adoption trend. The method for sizing the social listening market covers that extension.

Timing your figures against the ONS release calendar

ONS statistics are revised. A figure you cite today may be updated next quarter, which breaks your market model if you do not track it. The ONS release calendar lists publication dates for each statistical release, so check it before you publish anything built on those figures.

If a relevant release is due within two weeks, wait for it or state clearly which vintage you used. Provisional figures can move materially, so a model built on provisional data should say so, and a model built on revised data should say which revision it uses.

Set a refresh cadence. Quarterly is usually enough for a market model, because business population and internet usage move slowly. Platform data moves faster but should not drive the model. Pull the latest ONS figures just before each board pack and record the release date alongside the number.

Recording method notes so another analyst can repeat it

A market size without method notes is an opinion. Write down every filter, date and assumption so the next analyst can reproduce or challenge it. Tick off each note below before any figure leaves the team and keep the answers inside the model.

Add a short method note to the model itself, not a separate document that gets lost, so anyone opening the spreadsheet sees the assumptions. State the limitations plainly: company counts exclude unregistered businesses, PSC data excludes sole traders, and ONS surveys carry sampling error. These are not flaws to hide.

Version your model and save each quarter's pull as a dated file so you can show how the number moved and why. Keep observable signs of demand, such as search volume, job adverts and press mentions, as context only. They illustrate demand but never belong in the total.

Common questions