Every Exit Score is built from public Companies House filings by two independent models: one ranks how strongly the ownership signals match businesses that went on to sell, the other checks the business is worth buying. Here is how they work, how we tested them, and where they stop.
3 full reports free. No card required.
We scored businesses as they looked before they sold, using sale years the model never saw in training, and compared them with lookalike companies of the same age, size, sector and year that did not sell.
Measured on owner-to-owner sales (individual buyers and management buyouts) found in the Companies House PSC register. How the test works
Two questions every acquirer asks, scored separately and then combined into one number from 0 to 100.
Owner and director age, succession gaps, board changes and control structure, ranked against every scored UK company.
Financial health and operating performance from multi-year filed accounts. Below 35, a company is suppressed whatever its timing.
Every scored company receives an Exit Score between 0 and 100. The score is a weighted composite of two independent pillars: Exit Timing (45%) and Business Quality (55%).
Exit Timing ranks how strongly a company’s ownership signals match the pattern of businesses that go on to sell: retirement, disengagement, succession gaps. Underneath, the model estimates the probability of a succession-type sale (an individual buyer or management buyout) completing within the next 48 months, the window in which a searcher who starts today would realistically close. The published score is not that probability. It is the company’s rank within the scored universe. A timing score of 90 means stronger succession signals than 90% of scored UK companies; it does not mean a 90% chance of a sale. It is trained on owner-to-owner transitions observed in the Companies House PSC register, not on strategic corporate M&A. Business Quality measures whether the company is financially sound, operationally mature, and worth approaching in the first place. It answers: is this a business an acquirer would actually want?
The two pillars are scored independently, then combined. A company can score highly on timing (owner is 68, sole director, no succession plan) but poorly on quality (declining margins, thin asset base). The composite score reflects both dimensions.
We enforce a hard quality floor: any company scoring below 35 on Business Quality is suppressed from search results regardless of its timing score. A retiring owner does not make a weak business worth acquiring. This floor removes the noise that demographic-only filters leave behind.
Buy-and-build platforms and strategic acquirers. Screen a whole sector for owner-managed bolt-ons whose owners show succession signals, with group structures already consolidated and shells removed.
Search funds and ETA. Find quality-floored businesses in the EBITDA range you can fund, ranked by how strongly the owner’s profile matches businesses that sold.
Business brokers. Prioritise outreach to owners whose ownership signals match the businesses that go on to sell, before they look for an adviser.
The 45% timing pillar detects behavioural and structural signals that precede ownership transitions. These are patterns observed in real UK business sales, not arbitrary thresholds.
The average age of a company’s directors is one of the strongest predictors of exit. A board averaging 65+ with no younger directors is materially more likely to be in an exit window than one averaging 45. We score on age brackets, not exact ages, because Companies House records birth month and year only, not full dates.
A single director with no recent appointments represents concentrated key-person risk. If that director is also in a late-career age bracket, the company has no internal succession path. This is the most common profile among businesses that sell reactively rather than through a planned process.
A pattern of director resignations without replacement, or rapid recent turnover, can signal an approaching transition. We track the rate of board changes relative to company age. A 30-year-old company with three resignations in two years looks different from a startup with similar churn.
Beyond the boardroom, we read the ownership register itself. The age of the controlling shareholder is one of the strongest signals in the model: in observed owner-to-owner sales, controlling owners in their seventies sell at roughly twice the rate of owners in their fifties over a four-year window. A majority owner in a late-career bracket marks a genuine succession window. So does an owner who has stepped back from running the business: where none of the declared owners sits on the board, companies in our training data sold noticeably more often than their owner and board ages alone would predict.
Founder-led and family-held businesses transition differently from professionally managed ones. We detect founder tenure, shared surnames across the board and ownership register, and the presence or absence of a younger generation inside the company. A founder in a late-career bracket with no family successor is among the highest-conviction succession profiles the model sees.
When the longest-serving director has been in post for 25+ years with no co-directors of similar tenure, the business identity is concentrated in one person. This is not inherently negative (many excellent businesses are owner-operated), but it increases the probability that the next major event is a transition.
The timing score is produced by a gradient-boosted model trained on thousands of completed owner-to-owner business sales detected in the Companies House PSC register. These are individual buyers and management buyouts, with corporate and strategic M&A deliberately excluded, because strategic acquirers buy at every owner age and would dilute the succession signal the score exists to capture. Each training company is observed as it looked before its sale, and compared against companies of the same age, size, sector, and calendar year that did not sell, so the model learns genuine timing signals rather than “bigger and older companies get acquired more.”
Validation is strictly forward-looking: the model is tested on sale outcomes from years it never saw in training, and its accuracy is checked within every company-age and board-size band separately, so no single structural attribute can masquerade as a timing signal. In those held-out years, the observed sale-rate gradient by owner age matches what the model predicts, including the roughly two-fold difference between owners in their fifties and owners in their seventies over the four-year window.
The headline result of that test: 62% of the businesses that went on to sell were already scoring 75 or higher (the Pursue threshold) before their sale. Among businesses of the same age, size and sector that did not sell, only 30% scored that high, and in the very top band (90+) companies sold at roughly 2.5 times the rate of their comparators. In other words: even between two companies that look the same on paper, the score tells you which one is quietly heading for an exit. If anything these figures understate the separation, because the comparison group is deliberately built from lookalike companies rather than the average UK business.
The published score is a percentile rank against the full scored universe: a timing score of 90 means stronger succession signals than 90% of scored UK companies. The Pursue verdict requires the top quartile of that distribution. Clearing it triggers a further independent review layer that screens out lifestyle vehicles, subsidiaries, and holding structures before a company ever carries the label.
See which companies in your sector score 75+. Every match comes with the full score breakdown, financials and risk flags. 3 full reports free. No card required. Start free trial · See a sample report
The 55% quality pillar combines two independently scored components: Financial Health and Operational Performance.
Financial Health assesses the balance sheet: liquidity ratios, cash position, debt-to-equity levels, net debt trends, asset encumbrance from charges, and filing discipline. We score the trajectory, not just the snapshot.
Operational Performance assesses how effectively the business converts assets into profit: net profit and EBITDA margins, growth rates, return on assets, profit per employee benchmarked against sector peers, cash conversion quality, earnings stability over multi-year periods, and business maturity.
Both components use multi-year financial data where available. We weight recent performance more heavily but penalise volatility: a business swinging between strong and weak years is less attractive than one with steady, moderate performance.
Most UK private companies do not report turnover. We reverse-engineer revenue from ten independent balance sheet signals, cross-checked by a machine learning model trained on 157,000 company-years with known revenue.
A gradient-boosted decision tree (XGBoost) takes these ten estimates as input. When methods converge, confidence is higher. When they diverge, the model has learned which signals to trust for that sector and filing type.
Every estimate includes a confidence percentage, a calibrated probability that actual revenue falls within ±35% of the central estimate. When confidence is low, the report shows a range only with no central estimate.
Median absolute percentage error across the search-fund target range (EBITDA £250k–£5M, 2,988 companies in the holdout set):
EBITDA derivation. Where net profit is available, we derive EBITDA by adding back depreciation, amortisation, and interest. A cascading XGBoost model (68 features) estimates EBITDA directly when the components are not individually available.
Valuation orientation. Indicative valuation range based on sector-appropriate multiples applied to normalised EBITDA. A market orientation, not a formal valuation.
Spike-year detection. One-off events such as large asset revaluations or exceptional gains and losses are identified and flagged. Contaminated metrics are suppressed rather than displayed with misleading values.
Risk flags are surfaced separately from the Exit Score. They do not reduce the score. They sit alongside it so the buyer makes their own judgement.
Flags fall into three severity tiers: critical (issues that could block a deal), warning (areas requiring investigation), and positive (structural strengths that derisk the acquisition). Each flag includes a plain-language explanation and, where applicable, a suggested diligence action.
Common triggers: key-person dependency (sole director, no management depth), customer or supplier concentration patterns inferred from balance sheet structure, deferred capital expenditure, data quality concerns (limited filing history, abbreviated accounts), financial anomalies, and sector-specific risks drawn from a knowledge base covering regulatory exposure and industry-specific structural factors.
UK business ownership is rarely a single entity. We detect holding structures, suppress non-trading shells, and estimate consolidated group financials where subsidiary data is available.
Detection. Four independent methods: SIC code classification (642xx, 643xx, 7010), PSC ownership chain analysis, name pattern matching, and financial profile detection (a deterministic model that catches shell entities missed by the other three methods).
Two types, different treatment. Group holding companies (parents of active UK trading subsidiaries) are genuine acquisition targets and appear in search with a group holding label. Financial profile holdings, meaning shell entities, treasury vehicles and dormant intermediaries, are suppressed from search entirely. Revenue estimation is suppressed for both.
Group consolidation. When a company is the holding parent of UK subsidiaries, the report includes a consolidated view of the group. Where the parent files its own group (consolidated) accounts, we use those filed figures directly because they are the authoritative source and always take priority. Only where consolidated accounts are not available do we build an estimated consolidation: revenue, EBITDA, profit, net assets, net debt, and headcount aggregated across all active subsidiaries. The holding entity’s own revenue is excluded to avoid double-counting intercompany management fees.
Subsidiary perspective. When an entity is itself part of a group, the report shows the same consolidated view from the subsidiary’s perspective, with the parent, all active siblings and the entity itself aggregated into a single group picture.
Pre-exit restructuring signal. A holding company incorporated within the last three years that owns established trading businesses is a recognised pattern for pre-exit restructuring. We detect this and factor it into the exit timing score.
A SIC code and a registered office are what a company told Companies House, often years ago and often through its accountant. Before a company carries the Pursue label, we check both against what the business actually does and where it actually trades.
Why it matters. SIC codes are chosen at incorporation and rarely updated, and the registered office is frequently an accountant’s or formation agent’s address, sometimes in a different region from the trading site. A buy box built on those two fields alone misses real targets and returns false ones.
Activity. Companies that clear the Pursue threshold go through an independent review layer. It reads the full register record and, for most companies, the business’s own website and public listings, then places it in our activity taxonomy. Where the evidence points to a different trade from the SIC code, we correct the activity and sector.
Trading location. The same research establishes where the business trades from: its own website, a regulator or accreditation register, a trade-body listing or local coverage. Company-data aggregators do not count, because they only republish the registered office. A web-sourced location must be cited, and a named town must actually lie in the region we assign; where the web is silent, registered charges can supply the trading site. The resolved location replaces the registered office in region filters and the report, radius searches measure from it when it pins down a single place, and the search card flags any disagreement before you unlock.
How far to trust it. This is desk research, not a site visit, and it is not infallible. Where the evidence does not show where a company trades, we keep the registered office rather than guess. The web research covers nearly all Pursue companies above £250k EBITDA and most of those below it. Roughly one Pursue company in twenty turns out to trade from a different region from the one it is registered in.
Scores are built from Companies House bulk data products, published under the Open Government Licence v3.0. We process the full register: company profiles, officer appointments, annual accounts filings (including XBRL-tagged financial data), persons of significant control records, and company charges. On top of that, the web research described above can correct a Pursue company’s activity and trading location, and revenue estimation draws on ONS sector benchmarks.
What we have: Balance sheet data for most filed accounts, officer dates, director birth month and year (not full dates), filing dates and types, PSC ownership chains, company charges, SIC codes, registered office addresses, company status history.
What we do not have: Profit and loss data for micro-entities and most small companies filing abbreviated accounts (the majority of UK SMEs), real-time trading data, management accounts, customer lists, contracts, employee-level data beyond declared headcount. Director ages have a ±1 year margin. Revenue is estimated, not filed, for the vast majority of scored companies.
XBRL parsing only. ExitRadar reads accounts filed in XBRL (iXBRL) format. The vast majority of UK companies in the search-fund target range file in XBRL, but a small minority (typically older micro-entities, dormant entities or paper filers) submit PDF-only accounts that we cannot machine-read.
No metric in an ExitRadar report is shown in isolation. Every financial figure, score component, and operational metric is benchmarked against the company’s full sector peer set: companies in the same activity classification and EBITDA band.
Reports include a comparable companies table ranking the target against sector peers on Exit Score, estimated revenue, EBITDA margin, headcount, and average director age. Sector averages are displayed as a baseline.
This contextualisation matters because a 12% EBITDA margin means something different in construction (where 8% is typical) than in B2B software (where 25% is the floor).
ExitRadar reports are pre-approach intelligence. They identify which companies are worth your time and give you enough context to approach with confidence and specificity. They are not a substitute for due diligence.
Management accounts, site visits, customer concentration analysis from real data, and direct conversation with the owner remain essential. Our methodology gets you to the table with a well-informed view. What happens at the table is yours.
Every report carries the full score breakdown, financial provenance, comparable companies and a tailored approach strategy. 3 full reports free. No card required.