Hermès (RMS): Moat Analysis

Revised 8 August 2026. Full revision against the H1 2026 half-year report published 29 July. Every condition re-read, not only those expected to have moved. Readings are H1 2026 unless stated. Resale figures are from the series established 7 August 2026.

This document asks one question: how strong is the moat, and is it still intact. It contains no price, no position and no target. The moment it does, it stops being able to answer that question.

Verdict

Field	Reading
Strength	Exceptional. A deliberately controlled scarcity with no substitute at the apex, not replicable with capital because the constraint is trained craft and time, attacked repeatedly across a century and never breached
Condition	Intact. No discounting, no outlet, no loss of the allocation mechanism; the gap between supply and demand is being held exactly as the model requires
Security	High. The only party who can damage the moat is the family itself, and no outside actor, competitor or regulator can reach the mechanism
Pricing authority	Unconstrained. No regulator, competitor or channel partner has a say; the apex formats are deliberately priced below the clearing level, leaving power in reserve
Demand anchoring	Level 2. The underlying demand, visible positional signalling, is anchored in a permanent human drive, so it returns after any cyclical fall. The one caveat is medium migration: the drive is permanent, the physical-object medium is not (see What cannot be seen)
Verdict	Exceptional / Intact / Level 2 demand. Passes all three gates: deep moat, undamaged, on demand that always returns
Class A conditions	None triggered. No discounting or outlet, cadence unchanged, family control intact, no external compulsion, no apex substitute
Class B gauges	None triggered. Leather growth, margin direction and the resale premium all read clean, several at record levels
Decision	Passed on all three gates. A genuinely exceptional moat, undamaged, serving demand anchored in a permanent human drive that returns after any downturn. Held under monitoring and eligible for a Layer 2 read on any dislocation, because a drawdown here is a pause, not a permanent loss of the market

In one line: the mechanism sits entirely inside the company, nobody outside can reach it, nothing has moved it, and the demand it serves is anchored in a human drive that always returns after a downturn.

Hermès reads exceptional on strength, high-intact-unconstrained on condition, and level 2 on demand anchoring. That is the strongest reading available on all three axes, and it is the combination the Layer 2 gate requires. How those judgements are made, and the rules that stop them being adjusted after the fact, are set out in the annex.

1. What the company does

Hermès was founded in Paris in 1837 making harness and saddles. When the motor car killed the carriage trade, the company decided its asset was the ability to work leather, not the saddle. The saddle stitch on a Birkin is the harness stitch. It does not unravel when one thread breaks.

The business is organised into métiers rather than divisions, and one of them carries almost all of the economics. Leather Goods and Saddlery is close to half of revenue and most of the profit. Inside it, the work is done by a handful of apex formats: the Birkin, the Kelly and their variants. Everything else is arranged around them. The other métiers are Ready-to-wear, Silk and Textiles, Perfume and Beauty, Watches, Jewellery, and Art of Living.

How the money is actually made

Most analysis reads Hermès as a brand with high margins. It is better read as an allocation system.

The apex bags are not sold on request. They are not on the shop floor and cannot be bought online. A sales associate grants access to clients with a purchase history. Associates earn commission on the other categories rather than on the bags. The retail price is set below the price that would clear demand. That is why a secondary market exists at a premium, and why getting one at retail feels like an achievement rather than a purchase.

So the company monetises the queue as well as the object. A client working toward an allocation buys scarves, shoes, jewellery and homeware. Many never receive the bag. The other métiers are therefore not mainly diversification. They are how unmet demand for the apex object turns into revenue at high margin. That also explains why their growth rates can be dull while the system is exceptional.

The supply limit is real

Each bag is made by one artisan, largely by hand, in fifteen to twenty-five hours. Training takes eighteen months to three years in the company's own schools before an artisan works unsupervised, so output cannot grow faster than people can be trained. The company has chosen a pace of roughly six to seven percent volume growth a year, delivered through new workshops. The twenty-fifth leather workshop opened at Loupes in April 2026. Charleville-Mézières follows in 2027, Colombelles in 2028, Les Andelys in 2030.

Two features of that expansion matter. The limit is genuine, because craft capacity cannot be bought. And the pace is chosen, because the company could go faster by changing the method and has refused to. That means the scarcity does not rest on management restraint alone. Part of it is physical.

Distribution is owned

More than three hundred stores in forty-five countries. No wholesale channel of consequence, no outlets, no discounting. The company controls the point of sale, which means it controls allocation, and allocation is the mechanism. A competitor selling through department stores could not run this system even if it wanted to.

Where revenue came from in H1 2026

Revenue was €8.16bn, up 6.1% at constant currency and 1.6% as reported. The currency headwind was over €360m. The Americas grew 15.3%, Japan 11.0%, Europe excluding France 8.8%. Asia Pacific excluding Japan grew 2.4%. The region containing the Middle East fell 4.2%. Leather grew 9.8% and accelerated into Q2, from 9.4% to 10.2%, while the group decelerated. Silk grew 9.7%. Ready-to-wear grew 2.0%, Watches 0.2%, Perfume and Beauty fell 4.5%.

2. The moat

Six mechanisms, ranked. Two foundational, three reinforcing, one optional, plus governance, which protects all of them.

Keep one distinction throughout. The mechanisms are the demand driver, the controlled gap, the apex position and the ownership of supply and distribution. Margins, pricing power and the resale premium are not mechanisms. They are what the mechanisms produce.

Layer	Mechanism	Why it works
Foundational	Permanent demand driver	Stratified societies always produce demand for visible, hard-to-get markers of position
Foundational	The controlled gap	Supply held below demand, price below clearing, shortfall allocated by relationship
Reinforcing	Apex position	Nothing above it to migrate to. The closest substitute for a Birkin is a Kelly
Reinforcing	Craft capacity	Output is tied to trained hands. Capital cannot lift the ceiling
Reinforcing	Owned supply and distribution	No intermediary with inventory and a sales target, so no discounting
Optional	Ancillary métiers standing alone	Explicitly not load-bearing
Governance	Family control via H51	Turns discipline from a condition into a structure

Foundational: the permanent demand driver

Positional signalling is a structural feature of hierarchical societies rather than a modern marketing artefact. Every stratified society on record produced objects whose job was to make rank visible: burial goods, sumptuary law, livery, plate, furs, horses, land. The objects themselves varied enormously from one society to the next, while the demand underneath them stayed constant. Status objects appear in societies with neither money nor markets, so the drive predates any commercial system. What makes a marker work is not cost alone but difficulty of acquisition. An object anyone with money can buy signals only money, while an object needing money plus access plus time signals something money alone cannot buy, which is why rationed goods outperform merely expensive ones as status markers. The demand itself is judged permanent for as long as societies stay stratified and consumption stays visible.

There is a boundary condition here that everything below inherits. The demand is permanent, but the medium carrying it is not, and it has already migrated several times, from land and retinue to plate and furs to the modern branded object. It could migrate again, toward experience, toward discretion, or toward something digital that a handbag cannot carry. So the foundational claim is conditional: demand for what Hermès sells is anchored in a permanent human feature, as long as the physical luxury object remains a working medium for it.

Foundational: the controlled gap

This is the layer that does the work, and it gets mislabelled as brand often enough that the difference is worth spelling out.

A brand moat says people prefer this product and will pay more. That is real and common, and a competitor can erode it with a better product or a louder story. It does not produce an object that resells above retail.

The mechanism here is different. Hermès makes fewer apex bags than the market wants, prices them below clearing, and allocates the shortfall by relationship. Three things follow that a brand alone cannot deliver. The gap is publicly measurable, because the secondary market prices it, and a resale premium is an advertisement for scarcity that the company neither pays for nor controls. The purchase becomes an achievement rather than a transaction, which attaches value to the process as well as to the object, and a competitor selling a comparable bag on demand cannot copy that, because the thing being valued is the difficulty. And the queue itself is monetisable, at margins those categories could not command on their own merits.

One honest qualification. The system does not remove speculation, it channels it. The gap between retail and resale is an arbitrage the mechanism creates, and some allocations end up resold rather than carried.

The critical structural fact is who is in a position to break this. Supply is set by Hermès. Price is set by Hermès. Allocation runs through stores Hermès owns. There is no intermediary whose permission is needed, no standard-setter who can license an alternative, no regulator with power to make a private manufacturer produce more or sell to a particular person. That last claim has been tested in court rather than assumed. A class action filed in March 2024 argued that conditioning Birkin access on prior purchases was unlawful tying. In September 2025 the district court dismissed the federal claims with prejudice. It rejected the market definition, found no plausible market power over the object, and held that prioritising top-spending customers does not by itself breach antitrust law. The plaintiffs appealed to the Ninth Circuit and Hermès filed in May 2026. Not finally resolved, but the most direct external attack available has failed on the merits at first instance.

That arrangement buys an unusual amount of security, because there is no outside actor to worry about, and it carries an unusually high maintenance requirement for exactly the same reason. The equilibrium is managed rather than locked. It has to be held in place by a supply decision taken every year, against a permanent incentive to serve the demand sitting visibly unserved. A moat that depends on a rival's weakness dies when the rival improves. A moat that depends on your own refusal to maximise volume can only die by your own hand. That moves the durability question to governance.

Reinforcing: the apex position

The substitution test asks what the buyer's second choice actually is. Most luxury moats fail it quietly, which is why it gets run here rather than assumed.

The right benchmark is Ferrari, the closest structural analogue this business has. Both ration supply, allocate by relationship, price below clearing and refuse to maximise volume. Yet run the substitution test on Ferrari and a real menu appears: Lamborghini, Bugatti, Pagani, Rolls-Royce. Even at the apex, Ferrari competes for a contested wallet.

So scarcity discipline on its own does not empty the space above a product, which is why the Hermès answer cannot simply be assumed from the fact that the company rations supply.

Applied to the apex leather object, the answer is different in kind. The closest substitute for a Birkin is a Kelly, made by the same company. No other house occupies the position, and none has assembled one in four decades of trying. The secondary market confirms this independently. Hermès core formats hold value at or above retail while almost every other handbag depreciates on leaving the store, including houses with comparable heritage and higher prices. That is a market clearing price for a product with no equivalent. It is exactly the artefact a strong brand alone does not generate. Language confirms it too, in a smaller way. "The Birkin of" is used in ordinary speech to mean the best example of any class of thing. No competing handbag has that, and it was accumulated over forty years rather than bought.

The scope has to be stated carefully, because the strong version of this claim badly overstates it. Hermès holds no monopoly on luxury spending, on status signalling, or on handbags. It holds something close to a category of one inside a narrow subcategory: the top tier of leather goods used as status expression. Inside that boundary the substitution test is near-airtight. Outside it, in ready-to-wear, beauty and watches, the company competes on ordinary terms, and the divisional growth rates show it.

This layer counts as reinforcing rather than foundational because the gap produced it rather than the other way round. Decades of restraint are what emptied the space above the apex format, and an empty space does not by itself make a company restrained. It is also the layer a rich competitor can least easily attack, because the route runs through decades of restraint their own shareholders will not fund. And it is exactly why medium migration is the risk that matters: a category of one inside a category that stops mattering is not a defence.

Reinforcing: craft capacity

There are two arguments here, and they tend to get run together.

The first is predictive. The company has run continuously since 1837, through the destruction of its founding market, two world wars, the Depression, fast fashion, the digital shift, repeated luxury downturns, a pandemic and a hostile stake-building campaign. That record is the best available reason to expect it survives the next shock, though it rests on judgement rather than on anything demonstrable.

The second argument is causal, and it carries more weight. Output is limited by trained hands. Training is internal, and training capacity is limited by the number of existing artisans available to teach. That puts a ceiling on expansion that capital cannot lift.

A competitor with unlimited money can buy leather, machines, stores and advertising. It cannot buy accumulated craft capacity and it cannot compress the time required to build it, so it would own a company that was zero years old. This is the one input that behaves more like a physical law than a commercial advantage, and it explains why six to seven percent is a ceiling the company works beneath rather than a slogan it has adopted.

Reinforcing: owned supply and distribution

A scarcity strategy without control of the channel is not a strategy. An intermediary with inventory and a sales target will discount, and a discount destroys the gap. Hermès owns both ends. Upstream it has bought tanneries and integrated into hides, which matters because premium and exotic skins are themselves scarce. Downstream it sells through its own stores, with no wholesale of consequence, no outlets and no discounting.

So allocation decisions are executed by employees rather than negotiated with partners. No third party has the ability or the incentive to clear inventory in a soft quarter. When demand weakened in parts of Asia and the Middle East in H1 2026, the response available was to sell less. That is exactly the response the mechanism requires, and exactly the response a wholesale-dependent competitor cannot take.

The honest counterpart: this layer is expensive and slow. Owned distribution eats capital, and tanning brings animal welfare and environmental exposure. Those are real costs. They are the price of the mechanism working.

Optional: the ancillary métiers standing alone

Their role in monetising the queue is core to the thesis. What is optional is something narrower: could they eventually generate demand on their own terms, so that a weakening apex gap would not be a thesis-level event. High confidence: they are profitable and benefit from the halo. Moderate confidence: some, notably jewellery and Art of Living, are building genuine followings. Low confidence, treated as scenario: they become a second engine.

Two things keep it optional. H1 2026 showed Ready-to-wear at 2.0%, Watches at 0.2% and Perfume and Beauty down 4.5%, which is what these look like when the halo does the work rather than the product. And they compete in real markets against large houses, where Hermès has neither a controlled gap nor a category-defining object. Those are the two things that make the leather business exceptional.

Governance: the layer that protects the layers

This layer comes last in the order and close to first in importance, because it is the mechanism that protects the mechanism.

The gap has a single failure mode, and it is internal. Every year management faces visible unserved demand and an easy way to convert it into revenue. Producing more would raise this year's earnings and shrink the gap that produces the pricing power. The damage would not show for years.

That is exactly the trade a quarterly-reporting company under activist pressure is built to make. Four houses made it. Hermès has removed the ability to make it. The legal form is a French SCA, where Emile Hermès SAS is the active partner and holds powers a normal board does not. The family holds roughly two thirds of capital and over three quarters of votes. H51, created after the LVMH episode, pools slightly over half the capital under a pact running to 2031, with rights of first refusal and a dividend-funded mechanism to buy out family members who want out. Changing the key provisions needs a supermajority the float cannot assemble.

What separates this from ordinary good management is the difference between discipline as a condition and discipline as a structure. A long-tenured chief executive with the right instincts is a condition, and conditions expire with the person holding them. A control structure that makes indiscipline procedurally impossible outlasts whoever happens to be running the company. Hermès has the second, which is the most important defensive feature in this file and the main reason security reads high rather than moderate.

Three qualifications are owed here. The pact ends in 2031 and needs renewal; the incentive to renew is overwhelming, but a pact with an end date is not a perpetuity. The structure prevents sale without guaranteeing competence, so it removes market pressure risk while concentrating generational risk. And control protects against indiscipline but not against miscalibration. Management executing six to seven percent into a market growing more slowly would close the gap while behaving impeccably.

Evidence of strength: the competitive record

You cannot prove supply discipline works by looking at Hermès alone. That is survivorship reasoning. The mechanism only becomes provable by pointing at houses that ran the alternative and showing what happened to them.

Case	What it tests	Result
Burberry	Does licensing dilute the signal	Yes. Roughly fifteen years to rebuild
Coach	Does an outlet channel survive contact with the brand	No. The outlet became the business
Gucci	Can desirability rest on creative direction	No. Two boom-and-bust cycles in twenty-five years
Prada	Does reaching down-market stay contained	Probably not. Required retrenchment
Chanel	Can price substitute for allocation	No. Volumes and profits fell when the cycle turned
Ferrari	Does scarcity discipline alone remove substitutes	No. A real menu exists at the apex
LVMH	Can the ownership structure be captured	No. The best-equipped attacker failed

The four discipline breakers are the load-bearing evidence, because the mechanism is identical each time and the outcomes are documented over decades rather than quarters. Burberry licensed the check through the 1990s, especially in Japan, so the signal appeared on products it neither made nor priced. The signal became ubiquitous, then downmarket. Buying the licences back took fifteen years and two chief executives.

Coach introduced outlets to clear excess stock, and they grew until they were the main channel. A customer who knows the product will be discounted within months will not pay full price today. That is the whole mechanism in one sentence. Full-price equity never recovered and the group eventually renamed itself. Gucci made desirability depend on the creative director. That produces spectacular upside and equally spectacular downside on a cycle tied to one person's tenure. Two complete cycles in twenty-five years. Prada expanded distribution and accessible price points into China in the early 2010s, then had to retrench. The precise mechanism there is documented less cleanly than the other three and is recorded at moderate confidence.

Four houses, four different mechanisms, one result. Licensing, discounting, creative dependency and over-distribution are independent routes to the same place: a scarce signal turned into an available product. In every case recovery took years or decades, which is the evidentiary point: breaking supply discipline does structural damage rather than cyclical damage.

None of them were badly run. They had capable management, real heritage and access to capital. What they lacked was a structure making the discipline non-optional, which is why governance is rated as it is here.

Chanel is the closest thing to a controlled experiment available. It raised prices aggressively between roughly 2020 and 2023 to manage demand, rather than rationing supply. That captures more revenue per unit near term and needs no allocation apparatus. When the cycle turned, volumes and profits fell in a way the Hermès core métier did not, and secondary market support for the repriced formats weakened.

The comparison works precisely because Chanel is not weak. It is private, unbuyable, has comparable heritage, and faces no quarterly pressure. It chose price over rationing deliberately, as a strategy rather than under duress.

LVMH tested whether the ownership structure could be captured. From around 2010 the largest operator in luxury built a stake reaching the high teens, using derivatives that hid it until disclosure. It could not get control. The family consolidated shares into H51 with voting agreements, the regulator ruled against the acquirer on disclosure, and the position was unwound. The best-equipped possible attacker, with unlimited capital and a record of buying and improving luxury houses, mounted the most serious attack available and failed against the ownership structure.

Evidence of strength: pricing power

What has been demonstrated is exceptional by any standard. Mid to high single-digit increases on core formats every year, globally and simultaneously, through booms, recessions and a pandemic. No discounting, no outlets, no evidence of volume response. January 2026 took the Birkin 25 in Togo to $13,500 in the US, up about 6%, and the Mini Kelly 20 in Epsom to $11,400, up about 9%.

Current authority over the price is unconstrained. No regulator sets or reviews it. No competitor forces it. No channel partner negotiates it, because there are none of consequence. The tying claim went to allocation rather than to price, and it was dismissed at first instance.

The more interesting part is the pricing power the company chooses not to use. Hermès prices its apex formats below the level that would clear demand, and the evidence is the secondary market, since an object that resells above retail was underpriced at retail. The premium is therefore less a measure of desirability than a measure of pricing power deliberately left in reserve.

Two things follow from that. Reported pricing power understates real pricing power, by roughly the size of the premium. A company taking six percent a year on an object reselling at a large multiple is not at the limit of what the market will bear. It is choosing a rate that keeps the queue intact. And a compressing premium would be compression of the unused reserve rather than of the authority actually being exercised, which are two different things. Realised authority is measured by whether the January increases were absorbed, and they were. The reserve is measured by the premium, which stands at 2.47 to 3.12 times retail on the rationed formats. A file that reads a movement in the second as evidence about the first has confused the buffer with the balance.

The reserve is not infinite, and the mechanism depends on it. If the premium on the genuinely rationed formats went durably to parity, the object would have become an ordinary expensive good, the achievement component would disappear, and the ancillary monetisation would weaken with it. That is written into the register as B4.

Evidence of strength: the financial fingerprint

The figures below are included as evidence that the mechanisms are real, rather than as financial analysis in their own right.

Recurring operating income was €6.6bn in 2025, 41% of sales. That sits at the top of the global luxury set, above Ferrari, the relevant LVMH division and Richemont. It is the quantitative signature of an object sold below clearing price at a cost of goods bearing no relation to the price.

Return on invested capital is exceptionally high, reflecting a model whose main input is trained human craft rather than capital. The balance sheet is debt-free with substantial net cash, so capacity expansion is funded internally regardless of financing conditions. The company paid €328m to employees in early 2026 for 2025.

The pattern underneath those figures matters more than the level. Revenue growth has run well ahead of volume growth over long periods, which is the arithmetic proof that price rather than units is doing the work, and margins expanded while it happened, which rules out growth bought with promotion.

One peculiarity, noted rather than judged. The company pays dividends and reinvests but does not buy back shares of consequence, so cash accumulates. The sympathetic reading is that this is the conservatism of the whole enterprise. The critical reading is that reinvestment capacity is structurally smaller than cash generation, so surplus compounds below the cost of equity. Both are true. It resolves eventually through larger dividends, a policy change, or an accumulating pile.

One caution about relying on any of this. All of it lags, because these figures measure the moat's output rather than its mechanism, and if the gap ever starts closing in earnest the margin will be among the last things to show it. The register below is built on inputs for that reason.

Alternative explanations

A moat claim is only worth anything if the competing explanations fit the data worse, so the obvious ones are set out here and tested.

A post-pandemic boom now normalising. Contributes, and explains part of the recent deceleration. Cannot be the whole story, because the margin structure and the relative outperformance predate the pandemic and survived the normalisation.

Temporary Chinese demand. Inconsistent with the record. The margin structure and pricing cadence predate the Chinese boom, and core leather has grown through Chinese slowdowns while China-exposed peers wobbled. The 2.4% Asia Pacific figure explains the growth rate. It does not explain a 41% margin, absorbed price increases, or double-digit growth in three other regions.

A broad luxury cycle. Inconsistent with the cross-sectional evidence. In the same downturn, a price-led strategy produced falling volumes at one house and a fashion-led model produced a severe decline at another, while the Hermès core métier kept growing. A common cycle cannot explain divergence that large.

It is simply a very strong brand. The easiest to test. Strong brands do not produce objects that resell above retail, because a brand that could clear inventory at a higher price generally does. A pure narrative premium should produce depreciation at resale, as it does for almost every other handbag. The apex formats trade above retail instead.

Superior execution. Contributes, but cannot explain why competitors with excellent execution and larger resources sustain structurally lower margins in the same categories.

Narrative fitted after the fact. Worth stating, because a company this successful attracts explanations that were never predictive. The test is whether the mechanism was stated in advance and executed consistently. Supply restraint has been declared policy for decades, the capacity cadence is disclosed ahead of time, and the company repeatedly declined to accelerate when accelerating would have been rewarded.

The explanation preferred here is a deliberately maintained gap between supply and demand for an object serving a permanent positional need, monetised directly and through the queue, protected by ownership of the craft constraint and the point of sale, and made durable by a control structure that removes the ability to close the gap for short-term gain.

It is the only account consistent at once with the margin level, the resale premium, the absorbed price increases, the disclosed cadence, the cross-competitor divergence, the ancillary economics, and the multi-decade persistence of all of them.

How strong is it

Everything above establishes that a moat exists. It does not by itself establish how good one it is, and "exceptional" is a word that gets handed out too easily. So it is settled against three questions rather than left to impression, all of them about how hard the moat is to attack rather than about how well the company is currently trading. The scale itself is in the annex.

The first question is whether a substitute exists, which means asking what the buyer's second choice actually is rather than whether the product is well liked. There is none at the apex, and the closest one is another format made by the same company. That is a stronger answer than scarcity discipline alone produces, which is why Ferrari is in this file: it runs the same discipline and still faces a genuine menu.

The second is whether someone with money could replicate it. The inputs are a four-decade apex position, craft capacity that grows only as fast as people can be trained, and a control structure that makes restraint enforceable. None of the three is purchasable, and the second cannot be compressed by spending more. A rival could copy the workshops and hire the artisans and would still own a company that was zero years old.

The third is whether it has been attacked and what happened when it was. It has been attacked repeatedly, by different routes, and survived each one. Four houses broke the same discipline through licensing, discounting, creative dependency and over-distribution, and each needed years or decades to recover. Chanel ran the price-led alternative at comparable scale and diverged sharply when the cycle turned. The largest operator in luxury mounted the most determined capture attempt in the industry's history and failed against the ownership structure.

All three come back clean, which puts Hermès at exceptional.

3. What could break it

3a. Who can break it

The register follows from one question: who takes the decision that ends this moat, and would their action be visible in the numbers?

Mechanism	Actor	Visible in the numbers?
Controlled gap	The company	Yes, in cadence, workforce and capex
Craft capacity	The company	Yes, a method change would be announced
Owned distribution	The company	Yes, discounting is binary and obvious
Apex position	A competitor	Only late, and through resale data rather than disclosure
Governance	The family	No. Pact documents and filings
All of them	Court or regulator	No. It arrives as a ruling rather than a result

Two rows read "no". Those are the conditions this file rates as most consequential, and they are the ones a set of results cannot answer. That is the reason a clean quarter must never be read as a clean register.

Note what is missing from this table: there is no row where a customer takes the decision. That is the structural comfort in this name. The actor is almost always the company itself.

3b. Class A: mechanism conditions

These describe events rather than measurements. Each has an actor and a date, and none of them can fire because of a recession. A trigger here ends the thesis. The reasoning behind that rule, and behind the way these conditions are written, is in the annex.

#	Condition	Actor	Observable event	Reading now
A1	Cadence exceeded, or production method decouples output from trained hands	Company	Volume growth running ahead of workforce growth, or an announced process change	Clean. Workforce +5.5% to 27,107, capex +8.9% to €344m, leather +9.8%. Cadence unchanged, workshops disclosed to 2030
A2	Discounting, wholesale, outlet or licensing appears in any channel	Company	Binary. It appears or it does not	Clean. None of them. Gross margin rose to 71.1% from 70.7%
A3	Family control lost	Family	H51 lapses at 2031 without renewal, a governance amendment reaches the supermajority, or the bloc fragments	Clean. Pact in force, family holds ~2/3 of capital and >3/4 of votes. Treasury purchases of 94,846 shares for €160m are 0.09% of shares and immaterial
A4	External party compels supply, price or allocation	Court, regulator, legislature	A ruling or a rule	Clean, one matter open. Tying claim dismissed with prejudice September 2025. Ninth Circuit appeal pending; Hermès filed May 2026
A5	Reputational damage reaching the allocation system	Company	Clients leaving the queue, or earning a bag becoming embarrassing rather than desirable	Clean. Second-order, for reasons set out below
A6	A substitute appears at the apex	Competitor	Another house's format reselling structurally above retail. Not a launch, not a review, not a good collection	Clean. No competing format holds value at this scale

These six are not equally well calibrated, and recording that is more useful than presenting a register in which every line looks equally solid.

A1 and A2 are calibrated. Burberry and Coach broke in the same way and both showed how long the damage takes to surface, which is the empirical basis for this whole register. At Burberry the mechanism was broken and internally acknowledged by 1997, while revenue doubled between 2000 and 2006 and Harvard was writing case studies about the hottest brand in luxury in 2003. At Coach the outlet channel grew visibly in the disclosures for years before comparable sales turned. In both cases the lag between mechanism damage and financial damage ran to somewhere between three and eight years, which is why this register is built on events rather than on results.

A4 is half-calibrated. The failure mode itself is well documented elsewhere, where an administrative decision by an outside party altered a business while its reported figures were the best in its history. What those cases also show is that the instrument can sit at the wrong height. A condition written against legislation misses a break that arrives through administrative permission, and the height of A4 is set with that in mind.

A3, A5 and A6 are uncalibrated. There is no precedent in the comparison set where a family control structure like this one broke. The reputational precedents, Balenciaga and Dolce & Gabbana, are the wrong type, because those were desire brands with no allocation mechanism; whether a client four years into a relationship walks away is genuinely unknown, and sunk cost probably works the other way than it does for an impulse purchase. And there is no case anywhere of an apex position in luxury being taken, so the A6 threshold is reasoned rather than observed.

In practice that changes how each one is read. Where a condition is calibrated, the threshold can be trusted. Where it is not, any movement in that direction is a reason to look, including movement that stays below the threshold. This file does not claim precision it has not earned.

3c. Class B: gauges

These are measurements rather than events. A number has two causes, the mechanism and the environment, and the number alone cannot tell you which of them moved. So a gauge never rejects a name on its own; it obliges an investigation, set out in 3e.

Each one is a ratio rather than a level, because a level answers the wrong question. In 2020 recurring operating margin printed 21.5% with nothing whatsoever wrong.

#	Gauge	What it isolates	Expected direction in a shock	Suspicious if
B1	Leather growth vs delivered cadence (workforce, capex)	Is the gap closing from below?	Both fall, spread holds	Leather slows while workforce keeps growing
B2	Gross margin	Is anyone discounting?	Flat. The response to weak demand is to sell less rather than cheaper	Falls while volume holds
B3	Recurring operating margin	Nothing about the moat	Falls, and sharply	Stays high. Then ask what was given up
B4	Resale premium, three tests	Size of the gap	Both tiers fall together	Only the rationed tier falls
B5	Revenue growth vs volume growth	Is price still doing the work?	Spread holds	Spread disappears
B6	Divergence vs macro peers	Cycle or company-specific	Everyone falls together	Hermès falls alone

The current readings. B1: leather +9.8% against workforce +5.5% and capex +8.9%. Spread intact. B2: gross margin 71.1%, up from 70.7%. A house defending volume shows the opposite direction. B3: 41.0%, level with FY2025. B4: below. B5: leather revenue growth well above the six to seven percent volume cadence. B6: no divergence identified.

B3 earns its place by being the gauge most likely to be misread. A condition phrased as recurring operating margin falling below some level would have fired hard in 2020, at 21.5%, and the cause was vertical integration and fixed payroll, which is the mechanism doing exactly what it is built to do. A company that held its margin through that half would have done so by giving something up.

B4 needs setting out in full. The resale premium is the only material input the company does not disclose, so it is measured directly. Five formats are fixed before any price is collected, split into two groups. The rationed tier is the Birkin 25 Togo, the Mini Kelly 20 Epsom and the Kelly 25 Epsom, which are formats the capacity expansion did not reach. The common tier is the Birkin 30 and 35 in Togo, which it did.

That split is the whole design. If both tiers move together, the cause is macro or currency. If the common tier falls while the rationed tier holds, the cause is the supply expansion doing its job. If the rationed tier falls, no supply explanation is available and the residual is demand.

Only store-fresh examples in neutral colours with standard hardware count. Leather is fixed per format, because the Kelly 25 in Epsom and in Togo trade about forty premium points apart at identical retail.

Three tests, each anchored to a company decision rather than a chosen level:

Absolute price. Hermès raises retail by six to nine percent a year. A resale price in euros growing at or above six percent means the increase is being absorbed. Growth between zero and six percent is a mechanical effect of a rising denominator rather than a demand signal, so it registers as a watch condition. A fall is the trigger. Needs a series; no reading until the second measurement.

Multiple. Anchored to parity, which is the definitional end point rather than a chosen one. Trigger at 1.5x on the rationed formats, because a gauge that fires when the mechanism is already gone performs no function.

Spread between tiers. This is what the market pays for hard to get over easy to get, which is the rationing itself expressed as a number. Floor at 0.3x, less than half the measured level, which catches a convergence where both tiers stay above 1.5x while the distinction between them vanishes.

Reading of 7 August 2026, 173 observations across three vendors in three regions. Rationed tier 2.47 to 3.12x. Common tier 1.96 to 2.12x. Spread 0.59 to 0.84x.

Three features matter more than the levels. Every source returns the same ordering from smallest format to largest, without exception. Three inventories, three continents and three buyer populations produce one monotonic decline, which is the signature of a supply expansion feeding the accessible end of the range while the rationed end stays rationed. The most rationed format is the strongest, and the three sources agree within ten basis points despite different condition standards, printing 2.96, 2.89 and 2.99. And realised prices are not sitting below asking prices; at two vendors, sold examples cleared at or above asking for comparable inventory. A market in retreat shows the opposite.

Two things the measurement found that were not sought. The Birkin 35 has effectively left the store-fresh market. One vendor held three listings against forty-eight for the Birkin 25, another one pristine against 946 in lower grades. That is an absence of readings rather than a weak reading, and it is dropped as a fixed measurement point. And the gap between asking and realised prices is widest where demand is loosest, so that spread is adopted as a secondary gauge, because it should move before the level does.

The publicly circulating auction premium is not used. It averages across all conditions, colours and configurations, so it is determined by several hundred worn bags rather than the handful of unworn ones. A five-year-old bag carried daily selling below retail is what one would expect of any product. Reading that as evidence about the supply-demand gap confuses wear with desirability.

3d. Comparator sets

Both sets are fixed here, in advance. Choosing them after a bad print means choosing the peer that supports the conclusion you already have.

Macro peers. They share the demand shock, so they separate cycle from company: LVMH Fashion and Leather Goods, Richemont, Chanel.

Mechanism peers. They test the mechanism, and mostly do not share the shock: Ferrari for scarcity discipline, Richemont hard luxury as a migration destination, the quiet luxury cohort as the frontier if apex desire migrates within physical objects rather than away from them.

Two different lists doing two different jobs. The mechanism peers are the ones most easily left out, because they are usually not competitors in the ordinary sense and nothing in the reported figures points at them.

There is a trap in reading this. If Hermès lags because the rest are discounting, lagging is evidence the mechanism is working. So divergence raises a question, never a verdict. Always read A2 alongside it.

3e. The attribution test

Run this every time a Class B gauge triggers. All three must come back clean for the cause to be recorded as cyclical.

1. Is there a nameable external cause with a date?

2. Do the macro peers move with it?

3. Is the mechanism side unchanged? Did the company respond to weak demand by selling less, or by discounting?

   
   

Each answer is recorded with the cause named and dated, not just the conclusion, because a cyclical attribution only holds while the cause it names is still there. The escalation rule that follows from that is in the annex.

The attribution as it currently stands. Asia Pacific at 2.4% and the Middle East region at -4.2% are attributed to regional geopolitical disruption and its knock-on effect on tourist flows. The corroboration is sequential recovery inside the half: the Middle East from -5.9% in Q1 to -2.4% in Q2, France from -2.8% to +6.2% on returning tourists. Meanwhile the métier carrying the mechanism accelerated while the group decelerated. All of that describes a demand condition rather than damage to the mechanism.

What would bring this document out again: any disclosure bearing on the conditions above. Any of the named events: a Ninth Circuit ruling, a change to the articles, a board change at Emile Hermès SAS, or anything touching H51 ahead of 2031. And anything unforeseen where the question of whether to revise even arises.

4. What cannot be seen

Two risks carry real weight and have no actor, no date and no usable observation. They are not conditions. Listing them here stops "nothing triggered" from being read as "everything is watched".

The first is migration of the medium. The demand for positional signalling is permanent while the object expressing it is not, and it has moved before, so it could move again toward experience, toward discretion, or toward something digital. There is a related version that works differently, in which the object stays exactly where it is and gets revalued downward, read by a rising generation as an old-money marker that no longer confers position. Migration takes the signal somewhere else, whereas reinterpretation leaves the object in place and drains the meaning out of it.

Both are slow, both are invisible to internal metrics, and standard competitive analysis is structurally blind to them because product quality is not the contested variable. There is no announcement date and no rival to point at. This is the only way this moat fails without the company failing first.

The second is miscalibration of the cadence. Management executing a disclosed six to seven percent into a market growing more slowly would close the gap gradually while behaving impeccably, and every half-year it would look identical to a cycle. The recovery test in 3e is what separates the two, though it is a weaker guard here than elsewhere, because a cadence running too fast is company-specific and still shows up as demand weakness.

Two structural limits sit alongside those. A3, A5 and A6 have no precedent to calibrate against, so their thresholds are reasoned rather than observed. And rejection under this mandate is close to permanent, which makes a badly drawn Class A condition expensive and argues for writing them narrowly.

5. Assumptions

#	Assumption	Status
1	Stratified societies keep generating demand for visible positional signalling	High confidence
2	The physical luxury object stays an operative medium for it	The central uncertainty. Slow, no current evidence of migration
3	Hermès keeps holding supply below demand by choice	High confidence. Cadence disclosed and unchanged
4	Family control and the discipline it enforces persist	High confidence to 2031. Requires renewal after
5	Craft capacity stays a genuine constraint capital cannot bypass	High confidence
6	No external party gains power to compel supply, price or allocation	High confidence. Tested and upheld at first instance

6. Change log

Date

Change

8 Aug 2026

Full revision against the H1 2026 half-year report of 29 July. Every condition re-read, not only those expected to have moved. The resale premium is measured directly for the first time, 173 observations taken 7 August across three vendors in three regions, and B4 is stated with derived thresholds as a result. Zero Class A conditions triggered, zero Class B gauges triggered. Verdict Exceptional / Intact / Level 2 demand.

Annex: how this assessment is made

These are the rules the document is written under. They are set out separately so the file above can stay about Hermès, and so the rules cannot quietly change to suit a conclusion.

Judging strength

Strength is settled before anything else, because the three fields behind the condition axis only tell you whether a moat is undamaged. They say nothing about whether there was much of a moat to damage, and a mediocre business that nobody happens to be attacking would come back clean on all three, which would be an absurd result. So strength is a gate rather than a score. Three questions, all of them about how hard the moat is to attack rather than how well the company is trading. Is there a substitute? Could someone with money replicate it? Has it been attacked, and what happened?

Exceptional means all three come back clean: no substitute at the apex, no route to replication for a buyer with capital, and a record of surviving attack rather than simply never having been attacked. Strong means one of the three is soft, most often because the position has never actually been attacked and its resilience is therefore assumed. Ordinary means two or more are soft.

Only a name judged exceptional goes any further, and only then if its demand anchoring is level 1 or level 2. Anything below exceptional is declined however clean its conditions read, because a quiet moat is not the same thing as a deep one; and an exceptional moat on level-3 or level-4 demand is declined too, because a drawdown there cannot be waited out, since the market may not return.

The verdict:

The verdict is two judgements held apart, because collapsing them into a single grade hides the thing that matters most. One axis is strength: how deep was the moat to begin with, scored exceptional, strong, or ordinary. The other is condition: is it still whole, scored intact, impaired, or broken.

Keeping them separate is deliberate. A once-exceptional moat that has been damaged and a mediocre moat that happens to be undisturbed can look identical under a single letter, and they are not the same asset. The first can recover its strength if the damage heals; the second has no depth to return to.

Three questions feed the condition axis, answered in a fixed order so it cannot be talked into a softer reading afterwards. Can anyone outside the company reach the mechanism? That is security, high, moderate or low. Has anything altered the mechanism itself? That is condition, intact, impaired or broken. Does anyone else have a say in the price? That is pricing authority, unconstrained, constrained or contested.

Below all of it sits a rule the verdict cannot override. If a foundational or a reinforcing layer is broken, the name is rejected whatever the axes would otherwise say. Damage to a layer the file calls optional does not move the verdict, because the thesis does not rest on that layer. The obvious danger is that a broken layer gets reclassified as optional the moment it breaks, so the guard is procedural rather than a matter of judgement: a layer counts as optional only if it was already written that way in a revision dated before the damage, never in the revision that reports it.

Hermès reads exceptional on strength and intact on condition, the strongest available reading on both, and level 2 on the demand axis judged in the next section, so it passes all three gates the Layer 2 read requires.

Judging demand anchoring

Strength and condition together answer whether the company keeps its share of the market. They do not answer whether the market returns after a fall, and that is a separate question that decides whether a drawdown can be waited out at all. A moat can be fully intact while the demand beneath it shrinks permanently, so demand anchoring is judged on its own axis, on two tests: biology, is the underlying demand rooted in a permanent human or physical necessity or drive, and precedent, is there a history of similar demand collapsing and not returning.

The levels run from one to four. Level 1 is anchored in a physical or biological necessity that never stops, waste, water, illness, basic food. Level 2 is anchored in a permanent human drive that can fall cyclically but always returns, status signalling, transacting, connection, travel as experience. Level 3 is anchored in an established infrastructure or habit with a replacement cycle, probably durable but with no biological guarantee, and exposed to a structural shift. Level 4 is anchored in a moment, a build-out or a specific medium that may never return to its peak, a technology wave, or a status medium that a generation abandons.

The demand must be peeled to the exact layer the company serves, because a permanent drive can be served through a medium that is not permanent. Status is level 2; a specific physical object carrying status can carry a level-4 medium-migration risk. Only level 1 and level 2 demand is eligible, because only there does a drawdown reliably reverse. A level-4 name is declined whatever its moat, because when its demand falls there is no way to know, and no need to know, whether the fall is a pause or a permanent plateau.

Hermès reads level 2. The demand for visible positional signalling is anchored in a permanent feature of stratified societies and returns after every downturn, which is why the cyclical dips in Asia and the Middle East are read as pauses rather than as erosion. The one qualification is the medium: the drive is permanent, but the physical luxury object that currently carries it is not guaranteed to remain the medium, which is the migration risk carried in What cannot be seen, not a defect in the demand anchoring itself.

Why the conditions are split in two

A Class A condition describes something someone did. It has an actor and a date, and it has one cause: somebody decided. A Class B gauge is a number, and a number has two causes, the mechanism and the environment. You cannot tell from the number alone which of them moved. That is the whole reason for the split, and it has nothing to do with which conditions matter more. Gross margin sits in Class B and Coach died of it, while reputation sits in Class A and is second-order here.

So a Class A trigger ends the thesis on its own, with no interpretation and no mitigating circumstances, and good results are not a defence. A Class B trigger never ends anything by itself. It obliges the attribution test, and the worst it can return on its own is heightened attention with the cause unexplained.

Each Class A condition is written against the position as it stands today rather than as a change from an unstated starting point. Burberry was already licensed when its clock started, so a condition phrased as "if licensing appears" could never have fired there.

Each Class B gauge is a ratio rather than a level, and each carries the direction it is expected to move in a shock. Some of them are supposed to get worse. A margin that holds up through a demand shock at a vertically integrated company raises a question rather than offering comfort, because something was given up to hold it.

When a cyclical explanation expires

"It is cyclical" is always available and rarely falsifiable on the day, so it needs an expiry date rather than a counter.

A fixed number of quarters does not work, since five years of a dead economy is twenty quarters in which a gauge might legitimately stay red. So the attribution holds only while the cause it names is present and verifiable. When the cause lifts and the gauge stays red, the matter escalates to a Class A judgement. When the peers recover and this name does not, it escalates immediately, whatever the calendar says.

Divergence in the recovery is the sharpest moment available, because that is where the shared cause drops away.

Revision

The document is revised whenever something might have changed: a disclosure, a named event, or anything unforeseen. If the question of whether to revise arises at all, the answer is yes. Revising is cheap.

Class A is walked in full every time, including revisions prompted by disappointing results. It is binary and takes minutes, and the reason it cannot be skipped is that a moat can be broken while the reported numbers are the best a company has ever produced. Reading Class A only when the figures are poor would miss exactly that case.

The assignment of a condition to Class A or Class B is fixed before a trigger, never during the revision that reports one.

And the file is never revised for the first time while a price is under consideration. A moat assessment written with a price in view is an assessment under pressure to find whatever the price implies.

A note on what this document is and is not. This is written for the author's own investment process and published in that form. It reflects the analysis, assumptions and judgment of the author at the date of writing and nothing more. Nothing here is investment advice, a recommendation tailored to any reader, or an offer to buy or sell any security. Investing involves risk, including loss of capital, and past performance is no guarantee of future results. Any reader considering an investment should do their own work and consult a qualified adviser who knows their situation. The author may hold, or may come to hold, positions in the securities discussed.