The Melt Began Long Before the Meeting

On 1 May 2025, Chegg reported that its subscriber base had fallen 31% in a year and that revenue had dropped 30% to $121 million. The New York Stock Exchange had already issued a delisting warning that April, with the stock near 60 cents, and two rounds of layoffs followed within six months, taking out more than half the workforce. Read as a single event, it looks like a company hit by lightning. Read the way Rita McGrath teaches, it looks like the last frame of a film that started rolling on 30 November 2022, when a free chatbot went live and a few students in a few dormitories quietly stopped paying $19.95 a month. Nothing about that was hidden. It simply happened where nobody with a budget was standing.

01 · The argumentWhat the book actually claims

McGrath borrows the term strategic inflection point from Andy Grove and then does something Grove never did: she makes it operational. Her central claim is that inflection points have a structure and a sequence, which means they can be watched for rather than merely survived. The sequence runs through four stages. Something new appears and is wildly overhyped. It fails to meet the hype and is dismissed, often with relief. It then builds quietly among real users, which is the emergent stage nobody photographs. Finally it becomes the ordinary condition of the market, at which point responding is no longer strategy but triage.

The second claim is where the book earns its title. Early evidence of an inflection point does not appear at headquarters, because headquarters is where the aggregates arrive. It appears at the periphery, among the people who touch customers and lose small deals to competitors nobody at head office has heard of. Her chapter title compresses a line she takes from Grove: "snow melts from the edges." The consequence is organisational rather than analytical. A company's blindness is usually not a shortage of information. It is a reporting structure that filters anomalies out before they reach anyone empowered to act on them.

02 · Author's vantageWho is making it, and from where

Rita Gunther McGrath teaches strategy at Columbia Business School. She has circled the same problem for three decades from different angles. Discovery-Driven Growth, written with Ian MacMillan, supplied a planning method for ventures whose assumptions outnumber their facts. The End of Competitive Advantage (2013) argued that durable moats had become the exception, and introduced the arena concept this book reuses. Thinkers50 has ranked her among the world's leading management thinkers more than once. The pedigree matters here, because this book is synthesis rather than discovery.

03 · Key insightsFive things the book gets right that most planning processes get wrong

The dangerous stage is dismissal, not hype

Everyone knows to discount the hype phase. Almost nobody guards against the phase that follows it. When the overhyped thing fails to deliver on schedule, an organisation experiences relief, and relief hardens into a settled institutional judgement that the thing was never real. That judgement then persists through the emergent stage, when the evidence has quietly reversed. The dismissive phase is where competitive position is actually lost, because it is the only stretch in which a company is both wrong and confident. Every stage after it is cheaper to be wrong about, because by then the market is telling you loudly.

The information asymmetry inside a firm is worse than the one outside it

McGrath's edges argument is often read as a call to do more scanning. That undersells it. Her claim is that the relevant information is already inside the company. It is held by people whose job descriptions do not include reporting it, and whose managers are rewarded for smoothing it away before it reaches a slide. What she prescribes is plumbing rather than analysis: level-skipping conversations, senior leaders posted to the frontline on a schedule, a direct channel from the corner office to the street corner. Her warning about the narrow circle around a chief executive, and the sycophancy it breeds, is one of the sharper passages in the book.

The industry is the wrong unit of analysis

Industries are administrative categories, defined by what a firm makes. An arena is defined by the job a customer wants done, and by every route currently available for getting it done. Competitors inside one rarely share a classification code, which is why they never appear in the review deck. A company watching its industry is watching the wrong space, and gets ambushed by an entrant it had no category for. This is the framework's most portable idea. It survives the book's weaker sections intact.

Seeing is the cheap half; disengaging is the expensive half

The most underrated chapter concerns what McGrath calls healthy disengagement, the willingness to pull people, capital and attention out of a declining business while it is still profitable. Foresight without reallocation is a diary entry. Most organisations that named their inflection point correctly still lost, because the declining unit had the revenue, the headcount and the political weight, while the emerging bet had a slide deck. She is candid that this is a governance problem rather than an insight problem, and that it usually requires a chief executive to spend capital they would rather spend elsewhere.

Leading indicators are qualitative, and dashboards are built to exclude them

Reported revenue is the most reliable indicator a company has and the least useful, because by the time it moves the decision has been made for you. The indicators that arrive early are anecdotal, small-sample and easy to dismiss: an odd objection from a customer, a job posting at a competitor, a young engineer who keeps mentioning a tool nobody has approved. McGrath asks leaders to name in advance what would have to be true for a feared future to arrive, then to monitor those specific conditions. That advance commitment is what separates surveillance from hindsight.

04 · Where the argument echoesThree fields that formalised this before management did

The book's structural claims rhyme well outside strategy. Each parallel sharpens something McGrath leaves implicit.

Public-health surveillance solved the timing problem with pre-commitment. Epidemiologists do not wait to recognise an outbreak in real time. They run sentinel sites at the periphery, track syndromic proxies that move before confirmed cases do, and define the epidemic threshold in advance so that crossing it triggers action automatically. Set against this, McGrath's prescription is missing a component. She tells leaders to watch the edges and to trust their judgement about what they see, where epidemiology long ago concluded that judgement at the moment of onset is exactly what cannot be trusted, and pre-set thresholds are the fix.

Physics explains why the edge is not merely first but necessary. A supercooled liquid can sit below its freezing point indefinitely and remain liquid, because a phase change requires nucleation, and nucleation almost never begins in the perfect interior of the bulk. It begins at an impurity, a scratch, a surface. The parallel reveals something stronger than McGrath's metaphor claims. Headquarters is not slow to see the change; headquarters is structurally incapable of hosting its origin. The periphery is where the new state can form at all, which makes the edges a production site rather than an observation post.

Ecology supplies the warning McGrath softens. A lake absorbing phosphorus looks unchanged for decades and then flips to a turbid, algae-dominated state in a single season. What Holling's resilience work established is that these regime shifts are hysteretic: reversing the input does not reverse the state, and recovery costs far more than prevention would have. McGrath frames the post-inflection world as a new normal to be competed in. The ecological reading is harsher and more useful for a board, because it says the old position may be unrecoverable at any price, which changes what the option to wait is actually worth.

05 · The enterprise translationTwo arenas that ran the whole sequence since the book was published

McGrath wrote in 2019, before the fastest inflection point in commercial history. Generative AI has since run her four stages at a speed that makes the framework easier to test than she could have hoped, because the whole cycle fits inside three years rather than three decades.

Chegg is the cleanest case available. The company sold subscription homework help, held a library of more than 100 million pieces of curated academic content, and reached a market capitalisation around $14 billion. ChatGPT launched on 30 November 2022. By February 2023 Chegg was publicly acknowledging that AI chatbots were affecting new customer growth, and in May 2023 the share price fell sharply on that disclosure. The company responded with CheggMate, an assistant built on GPT-4 and trained on its proprietary library. The reasoning was that students would pay for academically rigorous AI rather than accept a free general-purpose tool. Students disagreed. Non-subscriber traffic, historically fed by Google searches, went from down 8% year on year in Q2 2024 to down 49% in January 2025, as AI Overviews kept users on the results page. By Q1 2025 subscribers had fallen 31% to 3.2 million, on revenue down 30%. In April the stock had traded under a dollar for thirty consecutive trading days and the exchange issued a delisting warning; it recovered above a dollar the following month. Layoffs took 22% of staff in May 2025 and a further 45% in October, alongside a change of chief executive. The shares are down more than 99% from a 2021 peak of $108.

The second arena is less tidy and therefore more instructive, because it is still in the dismissive stage as this is written. GLP-1 medications changed how a large and growing population eats. Roughly one in eight American adults has used one, and Morgan Stanley has projected the US user base could reach 31.5 million by 2035. Walmart's US chief executive told Bloomberg in late 2023 that customers collecting the drugs from its pharmacies were buying less food. Three responses appeared to the same signal, and they map onto McGrath's stages with unusual clarity.

Nestlé acted first, launching Vital Pursuit in May 2024, its first major new US brand in nearly three decades, built around portion sizes aligned to a suppressed appetite. Danone followed in 2025 with an Oikos drink aimed at muscle retention among GLP-1 users. Conagra, General Mills and WK Kellogg publicly expected demand for their existing portfolios to hold. WeightWatchers, whose entire arena was the job the drugs now do better, bought a telehealth provider in 2023 to prescribe the very medications displacing it, reported a $345.7 million loss for 2024 on subscription revenue down 5.6%, and filed for Chapter 11 in May 2025 to shed roughly $1.15 billion of debt. Sixty-one years of brand equity did not survive the arena moving underneath it.

06 · PrinciplesFour things to change in the operating rhythm

07 · RelevanceWhy a 2019 book about corners reads differently in 2026

The framework has held. The clock has not. McGrath's illustrations assume inflection points that unfold over years, giving a watchful organisation time to convene, deliberate and stage its bets. Chegg had roughly twenty-nine months from the launch of a free substitute to a delisting warning, and most of the damage was done in the middle twelve. An annual strategy cycle cannot see a corner that arrives inside two planning periods.

That compression changes which of her prescriptions matter most. Arena mapping still works, because arenas are redrawn slowly even when technology moves fast. Discovery-driven planning still works, because its whole point is to convert assumptions into knowledge cheaply. What no longer works is the implicit sequencing, in which a company observes the emergent stage, forms a considered view, and then reallocates. At current speed the observation and the reallocation have to happen concurrently, which means the disengagement machinery must be built before the signal arrives rather than assembled in response to it. Boards that treat McGrath's book as a call for better sensing have read the easy half.

08 · Where the argument strainsFour lenses that sharpen the case rather than dismiss it

The book invites engagement from the disciplines it borrows from. Four lenses bring its genuine claims into relief.

Through the lens of forecasting research, the book has a base-rate problem it never addresses. Philip Tetlock's work established that the hard part of prediction is discrimination, not sensitivity: anyone can raise an alarm, and the skill lies in the ratio of true alarms to false ones. McGrath's cases are selected on the outcome. Every edge signal she describes turned out to matter, which tells the reader nothing about the hundreds of peripheral anomalies that were correctly ignored. A leader who follows her advice without a scoring discipline will generate more warnings and no more accuracy.

Through the lens of strategic theory, the construct is slippery. Clayton Christensen spent years narrowing disruption to a specific mechanism precisely because a loose definition made the theory unfalsifiable, and inflection point is looser still. Richard Rumelt's objection to strategy language that cannot distinguish between two courses of action applies here: almost any large change qualifies afterwards, and the framework offers no threshold that separates an inflection point from a difficult year while it is happening. That the author advises companies on the methods she is recommending does not invalidate the argument, but it does explain why the book reads as more actionable than its evidence supports.

Through the lens of real-options theory, her small-bets prescription is sound and incomplete. Option value depends on cheap abandonment, and large organisations are structurally poor at abandoning things, because every pilot acquires a sponsor within two quarters. The finance literature on this is unambiguous and McGrath's own healthy-disengagement chapter gestures at it, yet the book stops short of the governance mechanics that would make abandonment routine: pre-committed kill criteria, sunset dates written at funding, and a reallocation budget owned outside the business units.

Through the lens of behavioural strategy, the dismissive stage is misdiagnosed as a perception failure. Barry Staw's work on escalation of commitment suggests something less flattering and more tractable. Managers frequently see the signal and suppress it, because the person who names the inflection point is volunteering to have their budget reallocated. Read this way, the edges are not blocked by hierarchy so much as defended by it, and the fix is compensation design rather than information design. The book's remedies are almost entirely informational.

Refracted through these lenses, the friction above are not objections to McGrath's perspective. They are scaffolding the reader can build around the book's argument, places to anchor their own engagement and test the framework against disciplines it touches. What she built survives the engagement, and the reader gains a firmer grip on why it does. The arena construct and the four-stage sequence are durable diagnostic instruments; what the lenses expose is that the book is strong on where to look and thin on how to decide, and that the deciding is where organisations actually fail.

09 · Adjacent readingWhere to go next

10 · ClosingWhat the edges were telling anyone who stood there

The students who stopped paying in December 2022 were not hiding. They were posting about it. Their behaviour was visible to anyone in customer support, anyone reading the forums, anyone who asked a nineteen-year-old how they had done their problem set. What made Chegg's collapse look sudden was not the absence of evidence but the absence of a route by which that evidence could reach a person with the authority to stop defending a subscription business. McGrath's contribution is to insist that this route is a design choice rather than a matter of luck or leadership instinct. The snow was already going. Somebody at the company knew. The question every board should be asking is not whether their own edges are melting, since they are, but what would have to happen inside their organisation for anyone at the centre to find out in time.

11 · What to carry away

13 · ReferencesSources for the business evidence