The DosageWas Never the Problem

1 · OpeningThe dosage was never the problem

When a drug fails, a clinician has two explanations available. The dose was too low, or the diagnosis was wrong. One of those is far more comfortable than the other, because it leaves every existing belief intact and asks only for more of what has already been prescribed. Roger Martin has spent four decades in boardrooms where the same logic runs on strategy. His observation is unflattering: when a management model stops delivering, executives mostly respond by applying it again, harder, with more conviction than before. The dosage rises. Nobody revisits the diagnosis. This book is a set of fourteen second opinions.

2 · The argumentWhat the book claims

Martin's claim is that executives rarely fail for want of diligence or intelligence. They fail because they are running the wrong model. A model, in his usage, is the framework a leader uses to organise thinking and automate action, and its power comes from operating below conscious attention. That is also what makes it dangerous. A model that has served well turns invisible, and an invisible model cannot be argued with.

So the book proceeds as fourteen paired confrontations rather than one continuous argument. Each chapter names a dominant model, shows why it produces the outcomes managers complain about, then offers a replacement. Competition does not happen between corporations; it happens at the front line, where a product meets a customer. Shareholder value is not maximised by focusing on shareholders. Strategy is not a plan. Execution is not what happens after strategy. Each swap sounds small. Acted on, none of them is.

3 · The authorWho is making this argument

Martin was Dean of the Rotman School of Management at the University of Toronto from 1998 to 2013, and Thinkers50 named him the world's leading management thinker in 2017. His authority here rests less on the academic post than on one consulting relationship. He was strategy adviser to A.G. Lafley across Lafley's tenure as chief executive of Procter & Gamble, a decade in which the company doubled sales, quadrupled profits and added more than $100 billion of market value.

4 · Key insightsFour ideas that repay the reading

Competition happens at the front line, not the head office

Corporations do not compete. Products and services compete, one customer at a time, and the people who understand that contest best are standing closest to it. Martin's line is blunt: it is not corporations that compete, but the products and services they provide. Turn the pyramid over and a hard test falls out. Every layer above the front line has to add more value to the layer below than it costs, and a layer that cannot must go. P&G acted on exactly this in 2019, dissolving its regional presidencies and moving control to category heads.

Put customers before shareholders, so as to serve shareholders

Share price reflects expectations about future performance, not present results. A team that sets out to maximise shareholder value therefore ends up managing expectations rather than the business, and that is a game with no winning move. Focus instead on customer value and let expectations follow. Martin is careful to show this is not an ethical argument dressed as a commercial one. It is a mechanical claim about which way the causal arrow points.

Strategy asks what would have to be true, not what is true

Conventional strategy work starts with a problem and hunts for the answer. Martin's alternative generates several possibilities first, then asks of each what conditions would have to hold for it to be a great choice, and finally tests only the conditions the decision actually hangs on. The mindset shifts sound modest and are not. From what should we do, to what might we do. From what do I believe, to what would I have to believe. From what is the right answer, to what would we need to know.

Planning is not strategy, and comfort is the tell

Planning feels good because it deals in costs, which a company controls. Strategy feels bad because it deals in revenue, which customers control. Leaders drift toward the first. It dissolves the anxiety of a future nobody can know, and that relief is the whole attraction. Martin offers a diagnostic worth keeping: if you are entirely comfortable with your strategy, there is a strong chance it is not very good.

5 · ParallelsWhere the argument echoes

The book's structural claims resonate well beyond management. Three parallels sharpen the argument by testing it on different ground.

In clinical medicine, the gap between treating a symptom and revising a diagnosis maps onto Martin's model problem almost exactly. A physician who escalates dosage without reopening the diagnosis commits the error Martin describes. Medicine has built an institutional answer that management has not: the differential diagnosis, a formal duty to enumerate competing explanations before settling on one. Read that way, the what-would-have-to-be-true process is a differential for strategy, and its real function turns out to be forcing rivals to the favoured answer onto the table, not generating options.

In control engineering, a system that answers error by raising gain, without checking whether its model of the plant is right, will oscillate and eventually go unstable. Engineers call it model mismatch, and the remedy is system identification rather than more gain. That reveals something Martin's framing understates. The failure mode is not merely wasteful. It actively destabilises, because every round of raised conviction burns down the organisation's remaining ability to notice the model is wrong.

Evolutionary biology recognises the same shape in what it calls an evolutionary trap, where a once-reliable cue turns misleading and an inherited response now steers the organism toward harm. The parallel adds a prediction about timing. Traps become lethal not when the model is at its worst, but when the environment has already moved and the model has not yet failed visibly enough to trigger revision. That narrow window is where all fourteen chapters are aimed.

6 · Business use caseWhat a model change costs, and what it returns

The abstraction turns concrete in a case Martin returns to often. By the late 1990s Oil of Olay was a roughly $750 million brand with little growth, selling at $3.99 through drugstores to women over fifty whose stated concern was wrinkles. The reigning model treated it as a mature asset to be optimised, and optimisation recommended what optimisation always recommends: cost discipline and defensive pricing.

The replacement model asked a different question. Not where can this brand defend, but where would it have to play, and how would it have to win, for skincare to anchor P&G's whole beauty business. That reframing produced choices no optimisation would have surfaced. Drop the "Oil of" prefix. Target women aged 35 to 50 meeting the first signs of ageing. Build a technically differentiated product against several ageing signals at once. Package it as prestige, sell it through mass channels including Walgreens and Target, and price it at $18.99. The brand then grew at 10 to 15 percent a year for more than a decade, reaching roughly $2.5 billion and, by fiscal 2009, an estimated $2.8 billion of P&G revenue.

The transferable lesson is not the pricing decision. It is that $3.99 and $18.99 were never two points on the same optimisation curve. They belonged to different models, and the organisation had to be argued out of the first before the second became thinkable at all.

The same test, applied to the CFO's office

For a finance function the parallel exercise is uncomfortable and quite specific. Picture a planning cycle that runs four months, ends in a variance pack nobody acts on, and gets defended each year on the grounds that the numbers are more accurate than last time. Martin's diagnostic asks whether accuracy was ever the objective, or whether it became a proxy the function adopted because accuracy is controllable and relevance is not.

His chapter on corporate functions supplies the sharper test. What value proposition does this function offer the business units, and would they buy it if an outside provider offered the same thing at the same price? A function that cannot answer is running an inherited model rather than a strategy, and it has only two failure modes available: the servile one, doing whatever business units ask and losing all focus, or the imperial one, optimising itself as though the rest of the company were a constraint.

Martin's chapter on knowledge work supplies the operating design that follows. Work moves through three stages, and each rewards a different structure.

7 · PrinciplesFour disciplines to carry into practice

Four operating disciplines follow from the argument. Each is stated as a claim, expanded for anyone who wants to go further, and contracted into something worth remembering under pressure.

8 · RelevanceWhy this argument lands now

Something has shifted since these chapters were first written, and it cuts in Martin's favour. Analytical work that used to justify a management layer is becoming cheap and continuous. When variance analysis, reforecasting, driver decomposition and scenario comparison all come on demand, the layer that existed to produce them has to find another reason to exist or accept that it has none. Martin's demand that every layer earn its keep was a governance principle in 1993. It is now an operating question with a date attached.

The knowledge-work chapter carries the same charge. If judgement that never becomes procedure has to be re-bought every cycle, then functions that hoard heuristics and never codify them are exposed twice over: once to cost, and once to the possibility that somebody else codifies them first.

9 · CritiqueWhere the argument strains

Martin's framework invites engagement from the disciplines it draws on rather than dismissal from outside them. Four lenses bring its genuine claims into sharper relief.

Through the lens of dual-process cognition, Kahneman's work supplies a mechanism Martin never quite provides. What the book calls an invisible model is System 1 pattern-matching, and the implication is harsher than the book allows. Models turn invisible through success, not neglect, because repeated reward is exactly what migrates a deliberate judgement into an automatic one. The evidence that a model works is what makes it unquestionable.

Through the lens of research methodology, the P&G-weighted evidence base raises a selection concern Martin acknowledges without resolving. The cases populating these chapters are ones where a model change preceded a good outcome. Missing are the model changes that failed. Without them a reader has no base rate. Jon Ingham has flagged internal inconsistencies between chapters, which is what one expects when articles written across three decades are collected without being reconciled.

Through the lens of organisational theory, Argyris's split between espoused theory and theory-in-use exposes a gap in the culture chapter. Martin holds that culture changes only when individuals change how they work with one another, which is right. Argyris showed that the defensive routines protecting a theory-in-use are themselves undiscussable. A leader who announces a new model has not yet touched the routine that will quietly restore the old one.

Through the lens of institutional economics, the customers-before-shareholders argument understates what a listed chief executive is actually up against. The causal chain from customer value to shareholder value is sound as mechanism. Quarterly reporting, activist positions and compensation tied to share price still impose real personal cost on anyone who defers current expectations for future customer value. Martin is right about the machinery and light about the politics.

Refracted through these lenses, the friction above does not amount to an objection to Martin's perspective. It is scaffolding a reader can build around the book's argument, places to anchor their own engagement and test the framework against disciplines it touches. What Martin built survives the engagement; what the reader gains is a firmer grip on why it does. The substitution method is genuinely portable and does not depend on the P&G cases for its validity. What the lenses expose is that the book specifies what to change and leaves underdescribed how an organisation changes it against its own defences.

10 · Adjacent readingWhere to go next

11 · ClosingThe second opinion

A second opinion is not valuable because it is always right. It is valuable because it forces the first diagnosis to be spoken aloud, where somebody can look at it. That is what these fourteen chapters do. They do not establish that Martin's replacement models hold in every setting, and several are plainly contestable. They establish that a model was there the whole time, running quietly, and that it can be named. A clinician who cannot state the diagnosis has no business defending the dose. An executive who cannot state the model has no business defending the strategy, and will go on prescribing more of it.

The question it leaves openWhich of your models has been right for so long that nobody in the room can still say what it is?

12 · Carry awayWhat to carry away