When EBITDA Lies: Industrials & Manufacturing – Capex, Cash Conversion and the Maintenance Myth

Admiral Group for carriers, Howden for brokers and MGAs

Insurance brokerage valuation runs on a multiple of EBITDA; carrier valuation barely uses it. The two halves of the industry need different lenses, and this paper reads one UK carrier and one UK broker through their own accounts to show why.

The insurance paper in the Bloor Finance series, in two parts. Part A reads Admiral Group plc (LSE: ADM), the UK motor and household insurer, against Aviva and Sabre Insurance Group on one basis: UK-adopted IFRS 17, sterling, years to 31 December 2025. Part B reads Howden Group Holdings, the privately owned broker and MGA group, from the accounts it filed at Companies House for the 15 months to 31 December 2025, with US-listed brokers and a dollar-reporting UK peer as contrasts. It is written for boards and finance leaders who sell, buy or lend against insurance businesses, and want to know which number the other side will use.

Prepared by Darren Sack · Research Director · Bloor Finance Desk · Bloor Research International · November 2026

1 · Where the opening paper left off

The opening paper in this series gave insurance one section and one line in its closing table: where the metric lies, “EBITDA is meaningless; model risk sits off the page”; the honest gauge, “Combined ratio; reserve adequacy; AI-governance cost”. Its argument, in its own words:

“Start with what’s not priced. AI underwriting is quietly manufacturing a liability most insurers’ capital bases do not yet reflect: model risk, pricing bias and reserving inadequacy accumulating beneath a loss ratio that looks like it’s improving. High-risk AI underwriting also carries compliance obligations from August 2026 — a cost line that, for most carriers, does not yet exist as a line at all.”

Re-verifying those lines moved one date. The EU’s Digital Omnibus on AI, Regulation (EU) 2026/1744, entered into force on 27 July 2026 and deferred the AI Act’s high-risk obligations for stand-alone Annex III systems from 2 August 2026 to 2 December 2027. The cost line moved; it did not disappear — and for UK carriers it applies to their EU operations. The rest stands, with one refinement this paper adds: the combined ratio is the truth-teller only when the reader knows its basis. Two of the carriers in this paper say they are investing in AI: Admiral has established a GenAI Centre of Excellence, and Sabre calls itself ‘an AI-enabled business’. Neither quantifies the effect on its loss ratio in the reports read.

The paper follows Pallabi Mohanty’s brief for the sector and splits the industry in two. For carriers, EBITDA is close to meaningless, and the lies sit in the float, the reserves, IFRS 17 timing and capital. For brokers and MGAs, EBITDA is the deal metric — and the lies sit in commission timing, interest on client money, profit commissions and the add-backs and earn-outs of a roll-up.

Part A · Carriers

A1 · What does the combined ratio tell you that EBITDA can’t?

The combined ratio is claims plus expenses as a percentage of earned premium (under IFRS 17, insurance revenue). Below 100% the insurer makes an underwriting profit before any investment return. EBITDA cannot say this, because an insurer’s largest costs — claims and the reserves set aside for them — sit above EBITDA, while its largest assets earn the investment income that EBITDA mixes in. No company in Part A reports EBITDA at all.

Admiral reports a Group combined ratio of 80.1% for 2025 (2024: 76.9%). The three-year series shows why the basis matters before the level does. The 2023 figure was 88.7%, as reported in the 2024 annual report. The 2024 figure was first reported as 77.4%; in the 2025 report it is 76.9%, re-presented to exclude the US business after its sale. Admiral’s ratios are struck on discounted claims, net of excess-of-loss reinsurance, and calculated on insurance revenue that includes underwritten ancillaries and an allocation of instalment and administration fees. Its UK Motor ratio was 75.0% (2024: 70.0%), or 75.6% (73.2%) restated for the 2024 change in the Ogden personal-injury discount rate.

One combined ratio, three bases: discounted or not, net or gross, re-presented or as first reported. Ask which before you ask how low.

The peers show the spread. Aviva publishes both: a discounted combined ratio of 90.6% (2024: 92.2%) and an undiscounted one of 94.6% (96.3%), net of reinsurance. The four-point gap is a discounting benefit of £506m (2024: £428m), and Aviva says the undiscounted figure aligns more closely to the way the business is managed. Sabre reports only an undiscounted combined operating ratio: 91.6% in 2023, 84.2% in 2024, 82.3% in 2025. Each figure is honest on its own basis. The three are not comparable until the basis is stated.

A2 · How much of the profit is the float?

An insurer collects premiums before it pays claims, and invests the money in between — the float. Under IFRS 17 the return on the float is reported in two lines that EBITDA would net together: investment return on the assets, and insurance finance expense on the liabilities, which is the unwind of the discount applied to claims when they were first booked.

Admiral’s 2025 investment return was £209.4m (2024: £176.2m). Its net insurance finance expense was £111.5m (£92.5m) — in Admiral’s words for UK Motor, ‘the unwind of the discounting benefit recognised when claims are initially incurred’. In UK Motor, investment income of £183.2m less £102.9m of finance expense left £80.3m of net investment income — against an underwriting result of £795.2m and co-insurer profit commission of £74.5m. Sabre’s equivalent was interest income of £11.7m against a net insurance finance expense of £5.8m.

The line to watch is not the investment income alone but the gap. Discounting lowers this year’s claims cost and raises the finance expense in later years. When rates and claims balances rise together, as Admiral says they have since 2022, both lines grow, and a reader who sees only one of them misreads the profit.

A3 · What is IBNR, and how do reserve releases flatter profit?

IBNR — claims incurred but not reported — is the part of an insurer’s reserves for losses that have happened but have not yet been notified, plus the expected development on claims already reported. It is an estimate, and the estimate moves. When last year’s reserves prove more than enough, the surplus is released into this year’s profit. Under IFRS 17 that release includes both a better best estimate and the unwind of the risk adjustment — the margin an insurer holds above its best estimate for uncertainty.

Admiral’s releases are large and disclosed. Changes to its liabilities for incurred claims credited £423.9m to 2025 insurance service expenses before reinsurance (2024: £556.2m) — about 44% of continuing pre-tax profit of £957.9m. Reinsurers took a share: after their £101.2m (2024: £469.4m), the net release was about £322.7m (£86.8m). In UK Motor, £219.9m of the £335.7m gross release was the release of risk adjustment, about two-thirds (65.5%). The core motor loss ratio before releases was 72.8% (2024: 69.2%); releases took ten points off it, to 62.8%.

A reserve release is real profit. It is also last year’s caution being spent. Read the loss ratio before releases.

The confidence level behind the risk adjustment is the honest gauge, because it says how much caution is left to release. IFRS 17 requires it to be disclosed. Admiral holds its UK Motor risk adjustment at the 94th percentile (2024: 95th), within a policy range of the 85th to 95th. Sabre’s is about the 81.4th (80.6th), against a target of the 80th to 90th. Aviva’s non-life risk adjustment corresponds to the 83rd percentile (80th). Sabre’s 5.5 points of favourable prior-year development, it says, were ‘a combination of normal levels of IFRS risk adjustment run-off and some positive development’.

A4 · How does IFRS 17 change when profit is recognised?

IFRS 17, the insurance contracts standard in force since 2023, decides when insurance profit reaches the income statement. For long-duration contracts the expected profit is held back as the contractual service margin (CSM) — in the standard’s words, ‘the unearned profit the entity will recognise as it provides insurance contract services in the future’ — and released in each period by coverage units. For short-tail business the simpler premium allocation approach (PAA) applies, and there is no CSM.

Admiral applies the PAA to all of its insurance and reinsurance contracts, so its IFRS 17 timing lie is discounting, not the CSM: claims are discounted when they are incurred and the discount unwinds through finance expense later. In UK Motor, Admiral says the booked combined ratio for underwriting year 2025 — undiscounted, and including the risk adjustment — was over 100%, even while the reported Group ratio for the financial year was 80.1%.

Aviva shows the CSM at work. On its own alternative performance measure, the CSM opened 2025 at £7,772m, added £423m of new business and £326m of interest accretion, released £888m into profit, and closed at £7,723m. Its auditor considered the release of CSM to give rise to ‘an inherent risk of fraud in revenue recognition’. The Direct Line acquisition shows a second timing effect: claims bought in an acquisition must be measured under the general model, not the PAA, which ‘creates some timing differences in recognition of profit’. Aviva reverses that effect out of its adjusted operating profit and reports it below the line.

A5 · Solvency ratio and embedded value: what capital is the profit using?

A carrier’s profit is only distributable if the capital behind it is spare. The Solvency UK regime requires own funds above a solvency capital requirement (SCR), calibrated to a 1-in-200-year loss over one year. The solvency ratio is own funds divided by the SCR, and it is the number dividends are set against. Embedded value, the older life-insurance measure of in-force value, does not appear in Admiral’s or Aviva’s 2025 annual report.

Admiral’s solvency ratio was 193% at the end of 2025 (2024: 203%), on eligible own funds of £1.83bn against an SCR of £0.95bn — estimated and unaudited. Admiral’s standard-formula capital carries an add-on for the formula’s limitations with its business, ‘predominantly in respect of profit commission arrangements’: £24m as currently approved, and an updated, not yet approved, add-on in the SCR behind the 193%. Admiral paid £715.4m of dividends in 2025. Aviva’s Solvency II shareholder cover ratio was 180% (203%), after the Direct Line acquisition, with operating capital generation of £2,452m (£1,468m). Sabre’s was 198.7% before dividends and 161.5% after (216.6% and 171.1%).

Part B · Brokers and MGAs

B1 · When is commission income really earned?

A broker earns commission when a policy is placed, but not all of it is safe. Policies are cancelled and commission is returned; profit commissions depend on how a book of business performs years later. IFRS 15 handles both: a refund liability for consideration the broker expects to give back, and a constraint on variable consideration, which may be recognised ‘only to the extent that it is highly probable that a significant reversal in the amount of cumulative revenue recognised will not occur’.

Howden’s policy follows the standard. Where there is a risk of cancellation or return of commission, it defers revenue at inception on its best estimate of the probability of cancellation and releases it over the policy. Profit commission and quota share treaty income are estimated and then constrained; for profit commission the estimate is ‘based on actuarial analysis for the projected ultimate loss ratio with an appropriate risk adjustment’. That is the carrier’s reserving judgement, inside the broker’s revenue line. The amounts of profit commission recognised and of refund liabilities are n/d in the accounts read.

B2 · Is interest on client money operating profit?

Brokers hold premiums and claims money for clients and insurers — client money, or fiduciary funds — and earn interest on it. Howden reports that interest as revenue, so it flows straight into EBITDA, and it rises and falls with interest rates rather than with broking performance.

Howden reported fiduciary income of £91.8m for the 15 months to December 2025 (12 months to September 2024: £67.0m), within total income of £4,231.6m. ‘All income arises from customer contracts except for Fiduciary income.’ Against EBITDA on the IFRS basis, it was about 39% of the 15-month figure of £237.5m (2024: about 17% of £387.2m). Howden also describes the dollar and euro insurance cash and investments, earning floating rates, as a partial natural hedge against the floating interest on its term loans. Marsh McLennan, reporting under US GAAP, states the risk directly: ‘Lower interest rates have led to a decline in our fiduciary income.’

Interest on client money is real cash. It is also a bet on interest rates sitting inside a broking margin.

B3 · Acquisition add-backs and earn-outs: what is “adjusted” in a roll-up?

In a buy-and-build broker, adjusted EBITDA is the number that sets the price, the borrowing and the covenants — and it is built, line by line, from statutory results. Howden says so plainly: ‘Adjusted LTM EBITDA is presented as defined in the Group’s credit agreement’, and it is used for covenant testing.

For the twelve months to 31 December 2025 Howden’s IFRS EBITDA was £150.0m (12 months to September 2024: £387.2m). Adjusted LTM EBITDA was £985.0m (£922.2m). The £835.0m between them is six lines: £434.7m of ‘other cost adjusting items’, £202.9m of losses on New Initiatives not yet at run-rate profitability, £125.7m of IFRS adjustments (including share-based payments and transactional foreign exchange), £30.2m of acquisition costs, £26.1m for pre-acquisition results of businesses bought in the year and closed operations, and £15.4m of expected cost synergies. The ‘other cost adjusting items’ are themselves New Initiative set-up charges of £211.6m — sign-on and retention bonuses, and settlements, legal fees and provisions for significant litigation — transformation projects of £114.3m, integration and restructuring of £62.2m, other items of £13.0m, and £33.6m of ’employee services post-business combination’.

£150.0m of EBITDA under IFRS; £985.0m as the lenders define it. The adjustments are disclosed. They are also 85% of the number.

That last line is the earn-out. IFRS 3 is explicit that a contingent payment ‘automatically forfeited if employment terminates is remuneration for post-combination services’ — a cost of running the business. Howden charges it, then adds it back. Earn-outs that are consideration rather than pay sit on the balance sheet: Howden’s liability for contingent consideration and payments to minority interests was £266.4m at the end of 2025 (September 2024: £268.1m), measured at fair value; fair-value charges on it and the unwind of deferred consideration came to £36.3m over the 15 months. Below the line, the IFRS result for the 15 months was an operating loss of £578.5m and a loss before tax of £1,159.5m, after £336.1m of goodwill and intangible impairment.

The contrasts show the same construction on other bases. Ardonagh, reporting in dollars, gives pro forma adjusted EBITDA of $1.1bn ‘including completed and committed acquisitions to 19 March 2026, together with annualised growth and savings initiatives’. Arthur J. Gallagher, under US GAAP, reports brokerage EBITDAC, as adjusted — before the change in estimated acquisition earnout payables — of $4,446m, a 36.5% margin. The add-backs differ in name. The direction is the same.

B4 · What is a managing general agent, and where does its risk sit?

A managing general agent (MGA) underwrites on behalf of insurers: it prices and binds risks under delegated authority, using an insurer’s capacity and balance sheet, and earns commission plus, often, a profit commission if the book performs. It looks like a broker in its accounts and like an insurer in its economics. Its EBITDA depends on capacity it does not own, and its profit commission depends on loss ratios it does not reserve.

Howden’s MGA, DUAL, shows both sides. Its profit commission is recognised on actuarial projections of the ultimate loss ratio, with discounting for the binder years over which it is earned. And in 2025 the MGA’s risk reached the broker’s balance sheet: of £295.0m of goodwill impairment, part related to DUAL North America after a strategic review that closed certain business lines, and a further £41.1m of impairment included customer relationships there. Howden’s DUAL capacity providers are n/d in the accounts read.

B5 · Insurance brokerage valuation: what will a buyer do to your EBITDA?

A buyer of a broker or MGA will apply its multiple to a number it has rebuilt. In practice that means five tests:

  • Organic or bought? Separate organic growth from acquired revenue on a stated definition. Howden did not produce organic growth for the 15-month period; for the twelve months to September 2024 it reported 15%.
  • Run-rate or promise? Strip out add-backs for losses on initiatives ‘yet to achieve run rate profitability’, expected synergies and pro forma acquisitions until they arrive. Howden’s New Initiatives add-back was £202.9m (£140.0m).
  • Pay or price? Treat earn-outs tied to continued employment as cost, and contingent consideration as debt. Howden’s contingent consideration and minority payments: £266.4m.
  • Rates or broking? Show interest on client money separately, and value it as rate-sensitive. Howden’s fiduciary income: £91.8m for 15 months.
  • Leverage on which EBITDA? Howden’s gross central borrowings were £8,004.6m at the end of 2025 (30 September 2024: £6,172.1m) – about 8.1x adjusted LTM EBITDA, and about 53x IFRS EBITDA for the same twelve months (Bloor arithmetic).
  • Bloor quotes no market multiple in this paper. No published UK broker multiple found for this paper met the series’ sourcing standard, and the adjustments above matter more than the multiple: on Howden’s own reconciliation they are the difference between £150.0m and £985.0m.

How do UK carriers compare on the combined ratio, reserves and capital?

Read on one basis — UK-listed, IFRS 17, sterling, years to 31 December 2025 — the carrier peer set shows one pattern: every company reports a combined ratio, none on quite the same basis, and none reports EBITDA. The table does not rank the companies; it shows which honest gauges each one discloses, and on what basis.

Honest gauges — UK carriers peer set (FY2025; IFRS 17; £m)

Company (profit before tax, 2025)Combined ratio as reported, 2025 (2024)Basis of that ratioPrior-year reserve developmentRisk-adjustment confidence levelSolvency ratioEBITDA disclosed
Admiral Group (957.9 continuing)80.1% (76.9% re-presented; 77.4% as first reported); 2023: 88.7%Discounted; net of XoL reinsurance; ancillary and fee income in revenueGross release 423.9 (556.2); net c.322.7 (c.86.8); c.44% of PBT (Bloor)UK Motor 94th percentile (95th); policy range 85th-95th193% (203%), estimatedn/d
Aviva (1,440 attributable to shareholders)90.6% discounted (92.2%); 94.6% undiscounted (96.3%)Net of reinsurance; both bases published; discounting benefit 506 (428)‘Favourable’ (UK & Ireland); £ n/dNon-life 83rd (80th); life 72nd (68th)180% (203%), Solvency II shareholder covern/d
Sabre Insurance Group (51.0)82.3% undiscounted (84.2%); 2023: 91.6%Undiscounted only; discounted loss ratio 50.4% vs 54.1%5.5 points favourable (0.5 adverse)c.81.4% (80.6%); target 80th-90th198.7% pre-dividend; 161.5% post (216.6%; 171.1%)n/d

All figures are from each company’s 2025 annual report (ESEF filing), years to 31 December 2025, 2024 in brackets, except Admiral’s 2024 as-first-reported and 2023 ratios, which are from its 2024 annual report; Admiral’s 2024 ratio is re-presented to exclude US Insurance. Prior-year development for Admiral is the gross change to liabilities for incurred claims in the insurance service expense note; the share of profit before tax is Bloor arithmetic. n/d means not disclosed in the documents read and is never zero. Brokers are not tabled: no UK broker reports under IFRS in sterling on a comparable basis, so Part B uses Howden’s accounts with Ardonagh (dollars), Arthur J. Gallagher and Marsh McLennan (US GAAP) as contrasts outside the table.

The basis column is the point of the table. Three insurers, three presentations of the same ratio: Admiral discounted, Sabre undiscounted, Aviva both. Each discloses its risk-adjustment confidence level, as IFRS 17 requires, and each is well above its solvency requirement. Two disclose the loss ratio before releases — Admiral’s core motor 72.8%, Sabre’s current accident year 59.6% — but on different bases, and none on a basis a buyer could compare across the three.

What should insurance CFOs track instead?

Two dashboards, because the industry has two economics. For a carrier:

  • Combined ratio on a stated basis – discounted and undiscounted, net and gross of reinsurance. Admiral 80.1% (discounted, net of XoL); Aviva 90.6% / 94.6%; Sabre 82.3% undiscounted.
  • Loss ratio before reserve releases, and releases in pounds and as a share of profit. Admiral core motor 72.8% before releases; gross releases about 44% of pre-tax profit.
  • Risk-adjustment confidence level and its movement. Admiral 94th percentile UK Motor (95th).
  • Investment return and insurance finance expense, shown together. Admiral £209.4m and £111.5m.
  • CSM roll-forward, for life and long-tail business. Aviva: £888m released, £423m of new business.
  • Solvency ratio and capital generation against dividends. Admiral 193%; dividends paid £715.4m.

For a broker or MGA:

  • Adjusted EBITDA with the full bridge from IFRS, every year, on the same definition. Howden £150.0m → £985.0m.
  • Organic growth on a stated definition. Howden: not produced for the 15-month period.
  • Fiduciary income separately, with its rate sensitivity. Howden £91.8m.
  • Profit commission recognised and refund liabilities. n/d in the Howden accounts read.
  • Earn-outs and contingent consideration as debt; employment-linked earn-outs as cost. Howden £266.4m; £33.6m added back.
  • Leverage on statutory and adjusted EBITDA side by side. Howden c.8.1x and c.53x.

The Bloor lens

For a carrier CFO the work is specific. State the basis of the combined ratio every time — discounted or not, net or gross — and publish the loss ratio before releases beside it. Show the risk-adjustment confidence level and say how much of this year’s profit came from releasing it. Put investment return and insurance finance expense side by side. For a broker CFO: publish the bridge from IFRS to adjusted EBITDA, as Howden does, and say which add-backs are promises rather than run-rate. Report fiduciary income separately. Treat employment-linked earn-outs as pay.

Neither anchor is the weak case. Admiral discloses its releases, its confidence level and the basis of its ratios; Howden discloses every line of its £835.0m bridge. That is the point. When the disclosures are this full and the gap is still this wide — releases worth about 44% of a carrier’s pre-tax profit, add-backs worth 85% of a broker’s adjusted EBITDA — the honest reading of insurance EBITDA is the one the reader rebuilds.

Watch for the remaining When EBITDA Lies industry papers as they publish. For boards and owners who want an early diagnostic read of their honest gauges before a sale process, a refinancing or a results season, the Bloor Finance Desk is available for a confidential conversation.

Analysis only; not investment advice.

Frequently asked questions

1. What is a combined ratio?

Claims plus expenses as a percentage of earned premium (insurance revenue under IFRS 17). Below 100% is an underwriting profit. Check the basis: Aviva’s 2025 ratio was 90.6% discounted and 94.6% undiscounted; Admiral’s 80.1% is discounted and net of excess-of-loss reinsurance; Sabre’s 82.3% is undiscounted.

2. What is IBNR?

Incurred but not reported: reserves for claims that have happened but have not yet been notified, plus expected development on known claims. If the estimate proves too high, the surplus is released into profit. Admiral’s prior-year changes credited £423.9m to 2025 before reinsurance.

3. What is embedded value?

A life-insurance measure of the value of in-force business plus net assets. Neither Admiral nor Aviva uses the term in its 2025 annual report. Under IFRS 17 the contractual service margin carries part of the same information within the accounts: on Aviva’s own alternative performance measure it stood at £7,723m at the end of 2025.

4. What is a managing general agent?

An underwriting agency that prices and binds insurance on behalf of insurers, using their capacity, and earns commission and often a profit commission. Its revenue depends on capacity it does not own and loss ratios it does not reserve. Howden’s MGA is DUAL.

5. How is insurance agency valuation done?

Insurance agency valuation — UK readers say brokerage — usually applies a multiple to adjusted EBITDA. A buyer will test organic growth, run-rate add-backs, earn-outs, interest on client money and leverage first. Howden’s adjusted EBITDA was £985.0m against £150.0m on the IFRS basis for the twelve months to December 2025.

9 · Sources & basis

Every figure in this paper comes from a named, dated document. The principal ones are listed below with links, and a full Source Note gives the origin, pinpoint and status of each number. Carrier annual-report figures were read from each company’s ESEF filing; Howden’s from its Companies House filing. Page references are to the printed report.

Standards and regulatory guidance

Note: For references without a supplied direct report URL, the links lead to the publisher’s website rather than the specific publication.

Print Article for this blog Download here