Variant Perception

Where We Disagree With the Market

The market is debating the wrong knob. Consensus anchors the FY2026 debate on whether Q1 2026's trading-income windfall normalises and whether the forensic overhang clears — both legitimate, both partially priced into the 12.8x trailing P/E versus StoneX's 17.1x. The report's evidence sits one layer underneath: comp-to-revenue has held at 61% through 27% revenue growth and three years of acquisition integration, and average clearing client balances compounded from $11bn to $16bn through a rate-cut cycle — operating-leverage and float-engine signals consensus treats as background noise rather than as the decisive variables. The forensic discount is also misallocated: seven plaintiff firms allege off-book inflation in Market Making (~12% of FY2025 revenue, ~17% of Adjusted PBT), while consensus applies the discount to the whole franchise — Clearing's 49.5% Adjusted PBT margin and float-funded annuity are not in the population at risk. The variant view is not "the stock is cheap"; it is that the wrong line items are doing the explanatory work in consensus modelling, and Q2 2026 plus the FY2026 audit close are the windows where the right ones become observable.

Variant Perception Scorecard

The 67 reflects a real but contained variant view. Consensus clarity is high — five primary covers all rate Buy / Strong Buy with target convergence at $55-67, FY2026 EPS consensus is published at $4.81 across 8 analysts, and the trailing-multiple gap to StoneX is the explicit framing in three sell-side notes captured in research. Evidence strength is solid on the operating-leverage and clearing-balance disagreements, weaker on the segment-misallocation forensic point because the class-action allegations are unproven. The first resolving event is roughly three months out (Q2 2026 print, early August), with a second confirming event in the FY2026 audit close (March 2027) — short enough to keep the variant view actionable.

Consensus Map

The Disagreement Ledger

Disagreement #1 — Operating leverage is the under-modelled variable. Consensus would say "the comp ratio is sticky in a people business and revenue scaling alone won't bend it." The report's evidence is that 61% is already the result of three years of integration of ED&F Man Capital Markets (2022), Cowen Prime (2023), Hamilton Court (2025), Aarna (2025), Agrinvest (2025) and Winterflood (2025) — every one of those would normally widen the comp ratio because acquired traders bring portable books. That comp ratio held flat despite aggressive hiring is an unusual fact, and the same mechanism — fixed-cost absorption against rising commission and float income — is what would compress it from here. If the market is right, comp ratio rises above 63% as Webb Traders / Valcourt integrate and consensus EPS holds at $4.81; if we are right, it bends below 60% by FY27 and EPS prints $5.50-6.20. The cleanest disconfirming signal is the Q2 2026 segmental tables — if Group comp/revenue prints above 62.0%, the variant view is wounded.

Disagreement #2 — The forensic discount is segmentally mis-allocated. Consensus would say "you cannot ring-fence the discount because a restatement at the parent level taints the whole entity's reported financials." That is fair on legal precedent but understates the segment economics: Clearing's $528M of revenue and 49.5% Adj PBT margin runs on regulatory-segregated client funds, audited capital ratios (Total Capital Ratio 253%), and an investment-grade credit standing that the May 2025 senior notes issuance reaffirmed at tighter spreads than the prior tranche. The class-action complaint targets Market Making OTC trades and intercompany flows — not Clearing margin balances. The variant view: even a contained MM restatement leaves 70%+ of Adj PBT untouched, and an MTD ruling that scopes the dispute to MM accelerates a partial rerating. The disconfirming signal is the consolidated complaint scope at MTD stage — if the plaintiff firms successfully add Clearing or Solutions populations to the discovery scope, the whole-franchise discount is correctly priced.

Disagreement #3 — Consensus is double-counting Q1 normalisation. The simple arithmetic: $1.48 Q1 EPS, $4.81 FY26 consensus, residual $3.33 over Q2-Q4 implies a $1.11 average — a 25% step-down from Q1 and below the Q4 2025 print of $1.16. The model already assumes the windfall fades. Management's caveat then triggered a second mark-down even though the data point that triggered the caveat is the same data point that consensus already discounted. Consensus would say "anchor on FY24 trading-income run-rate and let recurring lines drive the rest" — but that anchor (a) ignores Winterflood's Q1 2026 $32.7M consolidated contribution and (b) treats the $16bn clearing balance run-rate as a single quarter rather than the new base. The disconfirming signal is whether Q2 2026 EPS prints above the $1.11 implied — under-print confirms consensus, over-print confirms variant.

Disagreement #4 — Bermuda redomicile tax arbitrage is calendar, not modelled. Consensus would say "no responsible model bakes in an unmodelled ETR change." Fair, but the May 21 vote is high-confidence to pass (proxy advisory commentary published; consent solicitation on 2029 notes already underway with a May 15 deadline; activist resistance not surfaced in research), and the implementation timeline is well-flagged for H2 2026. The economic content is mechanical: every 100bps of ETR compression on FY27E PBT of ~$520M is ~$5.2M of net income. A 5-7pp step-down is consistent with peer redomiciles and the current US/UK ETR delta. The disconfirming signal is post-implementation 6-K disclosure of the new effective rate — if it lands above 23%, the lever is smaller than implied.

Evidence That Changes the Odds

How This Gets Resolved

What Would Make Us Wrong

The cleanest way the variant view fails is at the comp ratio. Three years of flat comp/revenue is the unusual fact the disagreement rests on, but it has not yet been tested through an integration as large as Webb Traders combined with Valcourt — both of which are equity-derivatives-heavy market-making businesses where acquired traders historically command higher comp shares than commodity-clearing populations. If Group comp/revenue prints at 62-63% in Q2 2026 6-K segmental tables, the operating-leverage argument loses its evidence and the variant collapses to a trading-income debate the bears already win on management's own caveat. The honest reading: this is a one-quarter test, and one negative print does not refute three years of evidence — but two consecutive prints above 62% would.

The forensic mis-allocation argument fails if the consolidated complaint scope expands at MTD stage to include Clearing or Solutions populations. Plaintiff firms in this kind of fraud-on-the-market case routinely amend complaints to capture broader populations once discovery opens; the original allegations focus on Market Making, but the FY25 20-F's balance-sheet substantiation weakness applies firm-wide, and a clever plaintiff lawyer could re-frame the discovery scope to include receivables wherever they sit. If MTD is denied and the discovery scope is firm-wide, the whole-franchise discount is correctly priced and our segmental ring-fencing argument is wrong. The double-counting argument on Q1 normalisation also fails if Q2 prints below $1.00 EPS — at that point consensus is right that the $4.81 is the ceiling, not the base.

The Bermuda tax-rate argument fails if the post-implementation ETR lands at 22-24% rather than the 18-20% we imply. Recent UK-to-Bermuda redomicile peers have shown ETR step-downs of varying magnitude depending on US CFC and global minimum-tax interactions, and Marex's commodity client base creates US-source income that may not benefit from Bermuda treatment. We do not have direct evidence of the post-implementation ETR target — that is the lever's fragility and the reason it ranks fourth not first.

Finally, the operating-leverage argument cannot survive a clearing credit event of meaningful scale. The Q1 2026 $34M natural-gas client default was 2.7% of FY25 closing equity — small enough to be a warning shot, large enough to demonstrate the mechanism. A second concentrated default above $50M in any single quarter — particularly in commodity-cycle stressed populations like Brazilian coffee or Middle East energy — would force a re-underwriting of the float-engine durability that anchors disagreements #1 and #2.

The first thing to watch is the Q2 2026 segmental tables — specifically whether Group comp/revenue prints below 61.5% on a Webb-consolidated base. That one number adjudicates the strongest disagreement before the forensic and Bermuda paths even open.