Using ONS and Companies House data for UK local SEO market research

What to take away

• Local SEO market research in the UK can be built on two free official sources: ONS business statistics and the Companies House register.
• Business population estimates give you the number of private sector businesses by region and size band, which is the denominator for any market sizing UK exercise.
• Companies House filings let you count active companies by SIC code and postcode, which is competitor density mapping rather than competitor guessing.
• Business demography data on births and deaths tells you how fast a local market churns, so you know how often to refresh your numbers.
• The method is arithmetic you can show a client: filter, count, adjust for branches and non-trading entities, then multiply by a realistic fee.
• State the limits plainly. Official data lags, covers registered entities rather than trading ones, and says nothing about search demand on its own.

Why ONS and Companies House data belong in local SEO market research

Most agency market sizing is anecdote with a spreadsheet. Someone counts competitors on a map, guesses the average retainer, and presents a number. It survives until a client asks how it was built.

Official data fixes the weakest part of that process. The Office for National Statistics publishes counts of UK businesses by region, industry and size. Companies House publishes the register of incorporated entities. Both are free, both are citable, and both are already used by the finance and property analysts your clients also read.

The practical gain is defensibility. When you say a West Midlands town holds roughly a certain number of firms in your target sectors, you can point at a published dataset rather than a map screenshot. That matters for the local SEO market size conversation you have with a finance director.

The second gain is comparability. ONS figures use consistent geographies, so you can rank Greater London against the North West or Yorkshire and the Humber without rebuilding your method each time. A client expanding from Manchester to Leeds gets a like for like comparison.

The third gain is repeatability. A model built on named datasets can be rerun next quarter with new releases. A model built on manual counting has to be rebuilt from scratch, which means it usually is not.

None of this replaces judgement. It replaces the parts of the estimate that were never judgement in the first place, such as how many firms exist and how many of them are in your sectors.

For a wider view of how these sources fit together, see our notes on local SEO market outlook sources.

What each source is actually good at

ONS datasets are strongest on population, structure and change over time. They tell you how many businesses exist, what they do, how big they are and how the stock moves.

Companies House is strongest on the individual entity. It tells you a named company exists, when it was incorporated, where its registered office sits and what SIC codes it filed.

Search platforms tell you about demand, not supply. Use them for query volumes and category interest, then use official data for the market you are selling into.

The three questions official data answers

How many potential clients are in this area? Business population estimates answer that by region and size band.

How crowded is the supply side? Companies House filings answer that by SIC code and postcode.

How fast does the market turn over? Business demography answers that with birth and death rates.

Those three answers cover most of what a market sizing UK exercise needs before you add pricing assumptions.

Named ONS datasets for sizing local markets

ONS business statistics sit under the business, industry and trade theme. Start with the business statistics published by the Office for National Statistics because that hub links the datasets you will actually cite.

The first dataset to know is the UK business count, often called the Inter Departmental Business Register, or IDBR, based publication. It counts VAT and PAYE registered businesses by region, local authority, industry and employment size band. This is your denominator.

The second is the business demography publication, which reports business births and deaths. It tells you how many new firms appeared and how many disappeared in a period, which is churn.

The third is the changes to business data collection. It covers mergers, acquisitions, disposals and demography, useful when a client asks whether a local sector is consolidating.

The fourth is the business population estimates published by the Department for Business and Trade. These give headline counts of private sector businesses, split into registered and unregistered, with regional breakdowns.

Which geography to use

Choose the geography your client sells in, not the one with the tidiest data. Local authority districts suit town level pitches. Regions such as the North West, West Midlands or Yorkshire and the Humber suit multi site operators.

ONS geographies change. Local authority boundaries have been reorganised in several parts of England, so check the geography column rather than assuming a district code still means what it did five years ago.

For Scotland, Wales and Northern Ireland, the same UK level datasets apply, though some devolved statistics add extra detail. Do not mix a UK total with a nation level subset in the same table without labelling it.

Size bands matter more than totals

A region with many micro businesses is a different sales opportunity from one with a handful of large employers. Micro firms buy small retainers and churn. Larger firms buy more and take longer to close.

Filter to the employment size bands that match your service. If your minimum viable retainer needs a firm with staff, exclude the zero employee band, which is dominated by sole traders and dormant entities.

Record which band you used. A market size that silently includes every unregistered sole trader will look inflated and will not survive scrutiny.

Using Companies House filings to map competitor density

The Companies House register holds every incorporated entity in the UK. Each record carries incorporation dates, registered office addresses, SIC codes and filing history.

Competitor density mapping starts with a SIC code filter. SIC codes are imperfect, but they are consistent. Pick the codes that cover your client's category, plus the adjacent codes that describe the same service under a different label.

Then filter by postcode district. A postcode district such as M4 or B1 gives you a tighter picture than a whole city, and it matches how local search results actually behave.

Count active entities, not all entities. A company in liquidation, a company with overdue accounts and a company dissolved last year are different things. Companies House status fields let you separate them.

Reading the register properly

Registered office addresses are not trading addresses. Thousands of companies use formation agents and accountants as their registered office, which clusters apparent competitors into a few city centre postcodes.

Use the accounts and confirmation statement data where available to judge whether an entity is trading. A company with dormant accounts is not a competitor, whatever its SIC code says.

Check for duplicate group structures. One trading business may sit under a holding company, so a naive count can double it.

Turning counts into density

Density needs a denominator. Divide the number of active competitors in a postcode district by the number of businesses in the same area from ONS data. The result is a ratio you can compare across towns.

A high ratio means a crowded category where differentiation and proof matter more than reach. A low ratio with high business population means an underserved area, which is the more interesting finding.

Plot the ratio rather than the raw count. Raw counts simply track city size and tell you that London is big. Ratios tell you where the fight is easy or hard, which is what local SEO competitor research should establish.

A worked example of density

Suppose a postcode district shows 240 active companies in your target SIC codes and ONS data puts 4,000 businesses in the same area. The density ratio is 0.06, or six competitors per hundred businesses.

Run the same calculation in a neighbouring district with 90 competitors and 1,200 businesses. The ratio is 0.075, higher despite the smaller absolute number.

The second area is more crowded relative to its own business base, so the same visibility effort buys less. That is the sort of conclusion a client can act on.

Combining business population estimates with local search demand

Supply data alone does not tell you whether anyone is searching. Pair the counts with demand evidence before you price anything.

Use search volume tools for category queries, Google Business Profile insights for the client's own profile, and Google Trends for direction of travel. Treat all three as directional rather than precise.

The useful output is a penetration estimate. If a postcode district holds 4,000 businesses and 600 of them search for your category each month, the addressable market is a fraction of the total, not the whole thing.

That fraction is what you sell against. It also explains why a large business population can still be a poor market if the category is rarely searched.

Matching demand to sector

B2B categories with long buying cycles show low search volume and high value per lead. Consumer categories show the reverse. Your model should reflect which one you are pricing.

If your client serves trades, hospitality or retail, local search demand is likely dense and competitive. If they serve manufacturers or professional services, demand is thinner and the sales cycle matters more than the click.

Build separate models for separate sectors rather than one blended number. Blended numbers hide the segment that actually pays.

Where demand signals change the ranking

Two areas with identical competitor counts can rank very differently once demand is included. The one with more searches per business is the better target.

This is why we track local SEO demand signals separately from supply counts, then combine them at the end rather than the start.

Keep the two inputs in separate columns. Mixing them early makes it impossible to see which one drove your conclusion.

Building a local market sizing model step by step

This is the method, in the order you should run it.

1. Define the geography. Choose postcode districts, local authority districts or regions, and write down which one you used.
2. Pull business population estimates for that geography, filtered to the employment size bands your service needs.
3. Pull active Companies House entities for the same geography, filtered to your target SIC codes and excluding dissolved, liquidation and dormant statuses.
4. Calculate density as active competitors divided by business population, and note the ratio for each area.
5. Add demand evidence from search tools and profile insights, expressed as searches or enquiries per business.
6. Apply a realistic conversion and fee assumption, then present a range rather than a single figure.

Step six is where most models fail. A single number implies precision that official data cannot support, so publish a low and high case.

The table you should build

Area	Businesses (ONS)	Active competitors (Companies House)	Density ratio	Demand per business	Notes
District A	4,000	240	0.060	Low	Large base, few competitors
District B	1,200	90	0.075	Medium	Crowded relative to base
District C	2,600	130	0.050	High	Best combination
District D	800	70	0.088	Low	Avoid for now

Keep the table to one page. Clients read the density ratio and the demand column, then ask about District C.

Assumptions to write down

Record the SIC codes you included, the statuses you excluded, the size bands you kept and the date you pulled the data. Anyone should be able to rerun your model from those notes.

Record your fee assumption and where it came from. If it is your own rate card, say so. If it is a market average, cite the source.

Record what you did not count: unregistered sole traders, franchises, public sector bodies, and businesses outside your SIC filter.

A quick checklist before you present

• Geography named and consistent across every input
• ONS release named with its period
• Companies House filters stated, including excluded statuses
• Density ratio calculated with a matching denominator
• Demand evidence dated and described as directional
• Fee assumption sourced or labelled as internal
• Low and high case both shown

If any box is unticked, the model is not ready for a client meeting.

Limits of official data and how to state them

ONS releases lag. Business population estimates typically describe a period months before publication, so a figure you cite today may already be out of date for a fast moving sector.

Companies House coverage is registration based. It includes non trading companies, companies registered for a future project and entities whose registered office is an accountant's address. None of those are competitors in any meaningful sense.

SIC codes are self reported and often wrong. A firm may file a general code for years after changing what it does, so a filter can both miss competitors and include unrelated ones.

Neither source measures search behaviour. Demand evidence comes from platforms with their own sampling and privacy thresholds, which is why it should be described as directional.

How to phrase the caveats

Say what the number is, what it counts and what it excludes. One sentence each is enough.

Example: this figure counts active companies registered in the district under the selected SIC codes, and excludes dissolved entities and unregistered sole traders.

Avoid hedging that undermines the whole exercise. State the limit once, then move to what the number is still useful for.

Regulatory context worth knowing

If you publish market claims in marketing material, the Advertising Standards Authority and the CAP Code expect evidence to substantiate them. The Competition and Markets Authority takes a similar view of misleading commercial claims.

If your research involves personal data, for example contact lists pulled alongside company records, UK GDPR and the Data Protection Act 2018 apply, and the Information Commissioner's Office is the regulator. Company data is not personal data, but sole trader records can be.

Keep client data handling inside your normal compliance process. Market research does not create an exemption.

Reporting market research findings to clients

Lead with the decision, not the method. Clients want to know which areas to target and why, not which dataset you opened first.

Put the method in an appendix. One page of sources and filters is enough for the finance lead who wants to check your work.

Show the range. A low and high case with the assumptions between them is more persuasive than a single confident figure.

Name your sources in the report. ONS and Companies House are recognisable names, and citing them raises the perceived quality of the work at no cost.

Structuring the findings

Open with the target areas ranked by opportunity, using the density ratio and demand column. Then show the sizing range for the top two or three.

Follow with the competitive picture: how many active competitors, how concentrated, and what that implies for budget. Then state the limits in plain sentences.

Close with what would change the conclusion, such as a new ONS release or a shift in search demand. That gives the client a reason to revisit the work next quarter.

Keeping the model alive

Diarise the ONS release calendar and the business demography publication. Rerun the model when new figures land rather than rebuilding it.

Track your own client results against the density ratio you predicted. Over time that comparison becomes your best evidence for the next pitch.

Store the raw extracts with their pull dates. If a client asks in a year why a number changed, you will have the answer.

Common questions

Which ONS dataset should I start with for market sizing UK work? Start with the UK business count by region and industry, then add the business population estimates for headline totals. Both give you a denominator you can defend.

Can I use Companies House alone to map competitors? You can map registered entities, but you will overcount. Filter by status, exclude dormant accounts, and remember that registered office addresses are often accountants rather than trading premises.

How often should I refresh the numbers? Align refreshes with ONS releases and annual Companies House filing cycles. Quarterly is usually enough for agency work, monthly only if a client is actively expanding.

Do these sources cover Scotland, Wales and Northern Ireland? Yes. ONS business statistics and the Companies House register are UK wide, though some devolved publications add extra detail you may want to cite separately.

What is a reasonable density ratio? There is no universal threshold. Compare areas against each other within the same sector, and treat the ratio as a ranking tool rather than pass or fail test.

Should I charge for this research? Many agencies include a light version in a pitch and charge for a fuller model. If you charge, scope the geography and sectors in writing before you start.