Using ONS and Companies House data for UK SaaS personas

What to take away

Which ONS datasets support UK SaaS persona work

The Office for National Statistics publishes several releases that map onto SaaS buyer questions. Each answers a different part of the persona: how big the market is, what those firms spend on, and who works inside them.

Start with the business services statistics. The ONS business services release covers computer programming and consultancy among other service industries, with turnover and output movements over time. If your product sells into professional services or software delivery teams, this is your baseline for whether the sector is expanding or contracting.

For market scope rather than sector health, use the ONS digital economy data. It covers the IT and internet industry, which is where most UK SaaS vendors sit and where most of their buyers sit too. Use it to size the addressable population before you start slicing.

Workforce questions need a different release. The ONS employment and employee types data covers employment by type, including self-employment and part-time work. That matters for SaaS because seat counts and licence models depend on whether a firm hires contractors, full-time staff or both.

If you need firm-level counts rather than industry aggregates, the UK Business Counts dataset is the one to pull. It reports numbers of local units and enterprises by SIC code, region and employment size band.

That is the single most useful ONS product for persona segmentation, because it lets you say how many firms of a given size exist in a given region.

You can also pull the Inter-Departmental Business Register summaries, which underpin those counts. The IDBR is the sampling frame behind most ONS business surveys, so figures derived from it are consistent with published statistics.

One caution. ONS business statistics lag. A release published this quarter may describe activity from two or three quarters ago. Treat them as structural evidence, not as a live demand signal.

What each dataset answers

Querying Companies House filings for firmographics

ONS data tells you how many firms look like your buyer. Companies House tells you which ones they are. The Companies House register holds incorporation documents, annual accounts, confirmation statements, director appointments and filing history for every registered UK company.

The free search and the advanced search both accept filters that matter for segmentation: company status, incorporation date, SIC code, and location. Advanced search also lets you exclude dissolved companies in one step, which saves hours of cleaning later.

Read the confirmation statement first. It gives you the registered office address, the SIC codes the company has declared, and the statement date. That date tells you how current the record is, which matters when you are building a list you intend to keep.

Then read the accounts. Small companies often file abridged or micro-entity accounts, so you will not always get turnover. You will usually get employee numbers, fixed assets and net worth, which are enough to place a firm in a size band.

Director appointments are underrated. They show tenure, multiple directorships and resignations. A cluster of recent resignations at a firm that also changed its registered address is a signal worth noting, though not a conclusion.

Filing history shows you behaviour over time: whether accounts are filed early, on time or late. Late filing is a published fact, not an inference, and it is the kind of detail that makes a persona segment feel real rather than modelled.

Use the register for due diligence as well as list building. Before you name a company as a customer example or a target account in a persona document, check that it is still active and that the SIC codes you are citing are the ones it has actually declared.

A worked example

Suppose your product helps finance teams at UK software firms reconcile subscription revenue. You want a persona for finance managers at SaaS companies with 50 to 249 employees, headquartered in Greater London or the South East.

Segmenting by SIC code, region and company size

SIC code targeting for software is the fastest way to cut a large register down to a relevant set. The UK SIC 2007 classification is what Companies House and the ONS both use, so your filters line up across both sources.

Pick two or three, not eight. A persona that spans eight SIC codes is a market, not a persona. If you cannot say which code your best customers sit in, run a quick check against your own CRM before you query anything.

Region segmentation in the UK is not uniform. Greater London and the South East hold the largest concentration of software and consultancy firms, and the ONS regional breakdowns will show that clearly. Scotland has a distinct cluster around Edinburgh and Glasgow.

The North West, West Midlands, Yorkshire and the Humber, East of England and South West England each carry smaller but real populations.

Use regions that match your sales coverage, not every region in the file. If you have no field sales in Scotland, a Scottish persona segment is a content exercise, not a pipeline one.

Company size segmentation UK style usually runs on employment bands rather than revenue, because employment bands are what UK Business Counts publishes. The common bands are 0 to 9, 10 to 19, 20 to 49, 50 to 249, and 250 or more.

The 50 to 249 band is where many UK SaaS vendors find their sweet spot: large enough to have a budget owner, small enough to buy without a nine-month procurement cycle. The 250-plus band behaves differently, with formal vendor onboarding and often a security review.

If you are still working out which segment deserves the effort, the groundwork in SAAS marketing audience research pays off before you spend on campaigns. It is cheaper to narrow a segment than to abandon one.

Segment filters worth saving

Building a persona that survives scrutiny

A persona survives scrutiny when every claim in it can be traced to a source, a date and a query. That is the whole test. If a reviewer asks where a number came from and you cannot answer in one sentence, the persona is not finished.

Write the persona around decisions, not demographics. The useful lines are: what triggers a purchase, who signs it off, what budget cycle applies, and what would make the buyer say no. Official statistics support the first two; your own customer conversations support the last two.

Use the ONS employment data to sanity-check role titles. If your persona says the buyer is a head of finance at a 120-person firm, the employment figures tell you whether that firm size is common in your target region or vanishingly rare.

Use Companies House to ground the firmographic half. Named examples, even anonymised ones, make a persona concrete. A segment described as "mid-sized software firms in the South East" is forgettable. One described with a real size band, a real SIC code and a real filing pattern is not.

Keep the buyer and the firm separate. A persona document that mixes company attributes with human motivations produces muddled targeting. Split it into a firm profile and a buyer profile, and let each cite its own sources.

The best qualitative input remains direct conversation. SAAS marketing customer interviews give you the language buyers use, which no dataset will supply. Statistics tell you who to talk to; interviews tell you what to write.

If you are starting from nothing, the structure in SAAS marketing buyer personas is a reasonable scaffold to hang the official data on. Build the skeleton first, then populate it.

Remember the legal frame. Any personal data you hold on named directors or contacts falls under UK GDPR and the Data Protection Act 2018, and the Information Commissioner's Office enforces it. Company data is public; personal data you derive from it is not automatically fair game.

Documenting sources and refresh cadence

Every figure in a persona needs a citation line. Write it as: source name, dataset, release date, filter applied, date accessed. That single line is what turns an assertion into evidence.

The UK Statistics Authority sets the standards for how official statistics should be used and presented. Its guidance is worth reading once, because it explains why you should not compare figures from releases with different definitions.

Refresh cadence should match how fast the underlying data moves. ONS business releases update on a published schedule, so a quarterly check is usually enough. Companies House records change continuously, so any list you keep should be revalidated at least twice a year.

Set a review date on the persona document itself. A persona with no review date quietly becomes wrong. Twelve months is a reasonable default for firmographic sections; six months if you are selling into a fast-moving segment.

Version the document. Keep the previous version so you can show what changed and why. When a colleague challenges a number, the version history answers faster than memory.

Keep a source register alongside the persona. One table, one row per figure, with the URL, the release date and the query. It takes an hour to build and saves days over the life of the document.

For market sizing specifically, be clear about what you are measuring. The distinctions in SAAS marketing market size matter, because a firm count is not a revenue figure and neither is a total addressable market.

Before you brief anyone external, the discipline in define your target audience is the same discipline described here: name the segment, cite the source, set the review date.

Source register format

Common errors when using official statistics in personas

The most common error is mixing definitions. A firm count from one release and a turnover figure from another may not describe the same population. Check the coverage notes before you put two numbers in the same sentence.

Another is treating a lagging statistic as a leading indicator. ONS business data describes what has happened, not what is about to happen. Use it to size and segment, not to forecast demand next quarter.

A third is over-filtering. Stack SIC code, region, size band, incorporation date and filing status together and you can produce a segment of eleven companies. That is a list, not a persona.

Confusing local units with enterprises is a quiet but frequent mistake. A company with four offices counts four times in local unit figures. If you are counting companies, count enterprises.

Ignoring dissolved companies inflates every list. Companies House keeps dissolved records visible, and they will appear in exports unless you filter them out.

Finally, do not present a modelled number as a measured one. If you extrapolated, say so. The UK Statistics Authority's point about proper use applies to your own documents as much as to government releases.

Quick self-check before publishing

Common questions

Which ONS dataset should I start with?

UK Business Counts, because it gives firm counts by SIC code, region and employment size band in one place. Add the business services release for sector context once your segments are set.

Can I use Companies House data for marketing lists?

Company data on the register is public, but personal data you extract from it, such as director names, falls under UK GDPR and the Data Protection Act 2018. Check your lawful basis before you contact anyone.

How often should I refresh a persona?

Review the firmographic sections every twelve months, or every six if your segment moves quickly. Revalidate any Companies House derived list at least twice a year, since filings change continuously.

What if my best customers do not fit one SIC code?

Check what they have actually declared on their confirmation statements rather than what you assumed. If they span several codes, build two personas instead of one blurred one.

Is a persona without customer interviews still valid?

It is valid as a firm profile but weak as a buyer profile. Official statistics tell you who to target; interviews tell you what to say to them. Use both.

Do I need to cite sources inside the persona document?

Yes. A citation line per figure, with source, release date, filter and access date, is what makes the document defensible when someone questions a number.