Variant Perception

Figures converted from GBP at historical FX rates — see data/company.json.fx_rates. Ratios, margins, and multiples are unitless and unchanged.

Where We Disagree With the Market

The sharpest disagreement: the market is pricing Elixirr as if it sits inside the same AI-substitution trade that has cost Accenture roughly 40% of its market value over the past year — and the evidence says the exposure runs the other way. ELIX's adjusted EBITDA margin expanded from 28.0% to 29.6% through FY25, the same period in which Accenture's margin compressed; it has no junior pyramid to cannibalise, has already deployed 45 internal AI agents, and grew AI-attributed revenue 260% YoY. The 32% P/Sales de-rate against the stock's own six-year mean (2.71x vs 3.97x) reads as sector pricing applied without business-model differentiation. We carry one supporting variant — that the 640bp statutory operating-margin compression from FY23 to FY25 is mostly mechanical acquired-intangible amortisation, not structural decay — and one institutional variant — that the persistent valuation discount is implementation friction ($2.3M five-day capacity), not earnings doubt. The combined view: the tape consensus is wrong about why the stock is cheap, and the H1 FY26 trading update (mid-July 2026) is the higher-information signal both advocates have under-weighted because the Stan verdict points everyone at the September print.

Variant Perception Scorecard

The 62 variant strength reflects three things: the AI-asymmetry disagreement is materially evidenced (Numbers, Business, Story tabs all stack the same way), but resolution gates on a single print four months out, and the sell-side already partially shares the bull setup — what is genuinely non-consensus is the price-action consensus that the discount should hold. Consensus clarity is high on the tape (17.8% drawdown from the September peak, fresh death cross 2026-03-05, P/Sales 32% below own mean) and bifurcated on the sell-side (19 buys, +53% upside cited but a $14.30 broker midpoint). Evidence strength is the most defensible leg — the upstream forensic, business and historical work converge.

Consensus Map

The map deliberately separates price-action consensus from sell-side consensus. Sell-side reads bullish (19 Buys, broker midpoint $14.30 = +45% upside); the tape reads cautious-to-negative (-17.8% from peak, fresh death cross). When those two diverge, the marginal-investor consensus is the price action — and that is what we are taking the other side of on issues 1 and 2.

The Disagreement Ledger

Disagreement 1 — AI as asymmetric upside. Consensus would say ELIX is a sub-scale consultant in a sector that AI is structurally compressing — every listed peer has de-rated and ELIX has come along for the ride. Our evidence disagrees on a specific, model-level point: ELIX has no graduate-analyst pyramid to defend, so AI cannibalises competitor revenue but absorbs the junior workload that ELIX never had. The FY25 datapoint that the market is over-discounting is direction-of-margin: 160bp of expansion at ELIX in the same year ACN's margin compressed and ACN's stock dropped 40%. If we are right, the market would have to concede that the consulting AI trade is bifurcated — pyramid firms compress, partner firms expand — and the 5y-mean P/Sales of 3.97x becomes the floor, not the ceiling. The cleanest disconfirming signal is the H1 FY26 adj. EBITDA margin: anything below 27% with AI revenue mix flat says the asymmetry is rhetoric.

Disagreement 2 — Margin compression is mechanical amortisation. Consensus says the statutory operating margin slide from 26.3% to 19.9% in two years is fundamental erosion. Our reading: capex is still 0.16% of revenue, D&A is overwhelmingly amortisation of acquired intangibles (Forensics B4: green flag, "asset-light consultancy, nothing to capitalise"), and the gap between adjusted EBITDA margin (held 28-30%) and statutory operating margin widened in lockstep with M&A cadence. If we are right, the market would have to concede that the right denominator for valuation is adjusted EBITDA less a normalised SBP charge ($6.3M FY25), not statutory operating income — and the multiple math changes by 200-300bp. Disconfirming signal: H1 FY26 statutory margin at 18% or below would force us to retire the "transient amortisation" framing and accept structural compression.

Disagreement 3 — Discount is liquidity, not earnings. Consensus reads the 32% P/Sales de-rate as a verdict on quality. Our reading: institutional ownership at 1.6% per TipRanks (vs ~28% insider) and a five-day capacity of $2.3M mean the marginal buyer pool is too thin to set a price at the analytical fair value. The 28 April 2026 secondary placing was explicitly structured "to broaden the institutional shareholder base" — an admission that liquidity is the binding constraint, not valuation. If we are right, the market is correct on fundamentals and we are still wrong on payoff — the discount persists until float expands. Disconfirming signal: a clean H1 FY26 print and tighter spreads, with no rerate — the institution-implementation thesis would survive only if liquidity remains the constraint.

Evidence That Changes the Odds

The two heaviest items are #1 (adj. EBITDA margin direction) and #3 (the gap composition). They are linked: if the gap is dominated by amortisation that fades, statutory margin should re-expand and the de-rate is unjustified; if the gap is driven by recurring SBP and judgment-based fair-value movements, the bear is correct and we are over-paying for adjusted earnings. Item #4 is the existential risk to the trade — being right on fundamentals while liquidity keeps the price unpaid.

How This Gets Resolved

The resolution path is asymmetric on time and information. Signals 1, 2 and 3 land inside a five-month window (mid-July trading update through September interim) and resolve disagreements 1 and 2 cleanly. Signal 6 — liquidity — is the slow signal that determines whether we get paid even if the analytics are right. A PM following one signal should follow signal 3 (mid-July trading update) because it pre-tells signals 1 and 2 by ten weeks and gives the most time to act.

What Would Make Us Wrong

The strongest counter is the bear's denominator argument. If FY26 organic growth comes in below 10% while the adj. EBITDA premium to operating income widens further, our reading collapses: the AI-asymmetry framing was a one-year mirage, the margin expansion was Hypothesis-mix specific, and the market was correct to apply the sector AI de-rate. The whole "no junior pyramid to cannibalise" argument becomes a story without a product when organic growth itself is the cyclical variable. We have one year of supportive data; the bear has two years of statutory compression and a widening add-back stack. A second consecutive year of margin compression and a narrowing AI-revenue contribution would force us to retire the variant and concede the bear's read.

The second counter is structural rather than tactical: ELIX has never been tested through a real consulting recession. It IPO'd into the post-COVID rebound, expanded through the rate-hiking cycle, and is now riding the AI wave. The senior-led model that recruits in good times can disperse in bad ones — partners with vested equity options and no client demand have outside options (start a boutique, take their book to McKinsey). If H1 FY26 prints alongside a broader consulting downturn — Accenture, Capgemini and the Big Four warning together — the asymmetric-tailwind variant would be testing precisely against the conditions that have never been seen. We do not have the data to underwrite that outcome.

The third counter is the implementation tax. Even if we are right on items 1 and 2, item 3 says the discount may not close until liquidity expands materially. FTSE 250 inclusion is mathematically out of reach in the June 2026 review (cap rank ~351 typically requires ~$1.3B). The first plausible inclusion window is January 2027, conditional on a material rerate that the variant itself is meant to drive. If we are right on fundamentals and the discount persists for two years because the marginal-buyer pool stays thin, the trade is correct and the carry is wrong.

The fourth counter is the founder-selling pattern. If the 28 April 2026 placing turns out to include Newton stock, the market will read that as a more reliable forward indicator than any of the upstream tabs we have leaned on. We are taking the side of the analytical case against the founder's revealed preference; we should be honest that "the man who built it is selling at depressed prices" is a hard signal to outweigh.

The first thing to watch is the H1 FY26 trading update on or around 14 July 2026 — specifically whether it includes any quantitative organic-growth colour or an "ahead of expectations" framing, and whether the AI revenue mix is named as a contributor.