Novo Nordisk Business Model: 6 Lessons Every Incumbent Can Learn from the Obesity Market

This piece was first written in December 2025, when the debate about Novo Nordisk was still framed as a race for the better molecule. It has been rewritten with the figures from the half-year report of August 4, 2026, and with a different question at its centre. The question is not whether Novo was too slow. It is whether the classical pharmaceutical business model still fits the job its customers actually need done. That job is to stay healthy and keep the weight off, which is a continuing state and not an event. The piece runs the case through the four elements of a business model, takes the counterargument seriously, and ends with six lessons that have nothing to do with pharmaceuticals.

The Novo Nordisk business model in numbers

The facts first, because the interpretation should have to survive them.

Novo Nordisk launched Wegovy in America in June 2021 and had the weight-loss market to itself for more than two years. The Economist reports that American Wegovy sales reached $4.3bn by 2023. That year Eli Lilly launched Zepbound, took $4.9bn in 2024 and pulled ahead in 2025. By December 2025 Novo’s market value stood at roughly DKr1.4trn, about two-thirds below its June 2024 peak, when it had been Europe’s most valuable company.

From the company’s own half-year report:

• Volume growth of the branded obesity market in the United States, moving annual total to May 2026: 87%.
• American sales of injectable Wegovy in the first half of 2026: down 17% at constant exchange rates, and down 22% in the second quarter alone. The company’s own explanation is lower realised prices, partly offset by higher volumes.
• Adjusted gross margin: 79.3%, against 83.1% a year earlier, attributed first to lower realised prices.
• From January 1, 2027, the American list price of Wegovy falls to $675, a cut of roughly half.

Group revenue is holding up, and on August 4 the company raised its outlook for adjusted sales growth at constant rates from a range of minus 4% to minus 12% up to a range of 0% to minus 6%. The price pressure is a United States phenomenon so far, with China following. That is not a contradiction. Different geographies sit at different points on the same curve, and the most competitive market arrives first.

The usual reading of these numbers is that a competitor won a product race. I think that reading is too small, and the rest of this piece explains why.

The job never ends, but pharma is built for episodes

Start with the customer rather than the company, and ask what job they are hiring the product to do.

Nobody wants an injection, and nobody wants a molecule. What a person wants is to live well, which for a large group of people includes getting their weight down and keeping it there, because obesity is the entrance to type 2 diabetes, cardiovascular disease, kidney and liver damage, sleep apnoea and worn-out joints. Losing weight has an end. Staying well does not. The job is a continuing state.

The pharmaceutical business model was not built for continuing states. It was built for episodes, and rightly so. Infections, fractures, operations, acute failures: a problem appears, an intervention resolves it, the episode closes. Around that shape an entire industry logic assembled itself. Discover a molecule, prove it in trials, obtain approval, hold exclusivity for a defined period, sell through an institution that decides and pays, recover the development cost before the window shuts.

That logic is not broken. It is fitted to a different problem shape than the one in front of it. Four places where the fit fails:

The window. The model monetises a period of exclusivity. The job runs for decades. The company’s clock and the customer’s clock are not the same clock. With an episode nobody notices, because episode and window are both short.

The buyer. The whole architecture is built to persuade an institution with a dossier. Here a private individual decides and frequently pays.

The unit of account. What gets invoiced is a package. What is needed is a maintained state.

The recipient of the value. The largest part of the value consists of things that do not happen: the diabetes that does not develop, the infarct that does not occur, the dialysis that is never needed. Those savings land years later, in somebody else’s budget. The model has no mechanism to invoice them.

The customer relationship a drugmaker is not allowed to have

At this point the obvious advice would be: own the customer relationship. Become a consumer company. Novo’s chief executive, Maziar Mike Doustdar, said as much when he told The Economist that the company must build a “consumer mindset” and think more like Amazon.

The advice is unusable, and not because of culture or speed. A drugmaker is structurally barred from the relationship the job requires.

A manufacturer may make the product. It may not prescribe it, may not decide who takes it, may not titrate it, may not manage the patient over time. A licensed prescriber stands in between, by design, precisely so that the party earning on consumption is not the party deciding on consumption. In the European Union it goes further: Article 88 of Directive 2001/83/EC obliges member states to prohibit advertising of prescription-only medicines to the general public, and Article 86 defines advertising broadly enough to catch almost any inducement aimed at the patient. Direct-to-consumer advertising of prescription medicines is lawful in only two countries in the world, the United States and New Zealand, and even there the prescription runs through a physician.

Read the consequence carefully, because it is the centre of this case. The element that would carry the continuing part of the job, the relationship, is not an element a pharmaceutical company is permitted to own. What looks like a failure of imagination is in large part a constraint of licence.

This is why I keep insisting that a business model is a system of interdependencies rather than a list of blocks. You cannot simply swap a stronger block into the customer-interface slot if the law does not let you hold that slot. The question is not what Novo should want to be. It is what a drugmaker is allowed to be. However, this regulation was useful in the past, but it will be challenged by personalized, digital and data driven medicine.

1985: the pen was the permitted form of customer relationship

There is a precedent, and it belongs to Novo Nordisk itself.

The company was founded in 1923 to make insulin, a molecule that long ago stopped being scarce or distinctive. In 1985 it launched NovoPen, the first commercially available insulin pen, which the company today calls a design icon. The pen took the glass syringe out of a person’s daily life. The molecule did not change. What changed was the thing the patient touched every day.

I wrote about this in 2009, in a post on design thinking and disruptive business model innovation. The point then was that the drug was the means and the delivery system was where the benefit sat. The point now is sharper. The pen was the only form of customer relationship a manufacturer is permitted to have: a relationship through the product, not through the treatment. Novo did not move the regulatory line in 1985. It made the decisive move inside the line, at the one place where a manufacturer legitimately touches a person’s daily life.

That is what makes this case worth studying rather than judging. The company is not being asked to learn something foreign. It is being asked to repeat something it invented, under conditions where the equivalent position is harder to find and, this time, already occupied by somebody else.

Value proposition: nobody buys semaglutide

The value proposition is the promise of a benefit to an identifiable human being. It is never the product; the product is only the means. That is the reference argument on this blog, and it has rarely been so literal.

Novo’s implicit promise remains product-shaped: a clinically proven GLP-1 medicine that helps you lose weight. Notice the verb. Losing is a process with an end state, and the promise quietly stops there.

The job does not stop there. Ask what someone is hiring the medicine for and the answer is a condition, not an outcome: to be a person who stays at this weight, without organising their life around it. That formulation contains supply, price, privacy, side-effect management, food, movement, habit and continuity, and the molecule is one component among them. This is jobs to be done applied at the level of the business model rather than the product feature.

When the customer changes, the value proposition changes by definition. Here the customer changed from a payer’s formulary committee to a person paying out of pocket, and the promise has not been rewritten to match. That is why I talk about customers rather than markets: a committee and a forty-year-old woman with a credit card are not two segments of one market. They are two different businesses.

The evidence that the promise is not being kept is not clinical, it is commercial. Real-world persistence on these therapies is far below trial conditions. In a cohort of 125,474 American adults, 64.8% of patients without type 2 diabetes discontinued within twelve months, against 46.5% of those with diabetes, and the dominant reasons reported in this literature are cost and coverage rather than side effects. Persistence has improved substantially as supply normalised, which matters and belongs in the record. But a value proposition whose benefit exists only while treatment continues, sold through a model that loses most customers within a year largely on price, is a promise the business model is not equipped to keep.

Value architecture: the customer account belongs to somebody else

Novo’s value creation architecture is linear and optimised for control: research, manufacturing, wholesalers, insurers, physicians, patients. Every stage assumes a mediated market.

The Economist records what happened when it met unmediated demand. Novo planned for demand at three times the level of Saxenda, its older obesity drug; five weeks after launch Wegovy had recorded the equivalent of four years of Saxenda prescriptions. Supply could not keep up, the drug went onto America’s official shortage list, and compounding pharmacies were legally permitted to sell copies at steep discounts. Novo estimates roughly a million Americans still use them. Lilly, meanwhile, launched LillyDirect in January 2024 at $399 for low-dose vials against a list price near $1,100; Novo followed with NovoCare Pharmacy in March 2025 at $499 a month. Fourteen months.

Now the structurally interesting part. Novo has since built the continuing relationship the job requires. Its half-year report describes a subscription for self-pay patients with three, six and twelve-month plans, available since late in the first quarter of 2026 through Ro, WeightWatchers and LifeMD, with Hims & Hers and Sesame expected to follow. The company calls it the first subscription programme for FDA-approved Wegovy.

Look at who is in that sentence. Every one of those partners can do what Novo cannot: they employ or contract prescribers, they consult, they titrate, they follow up, they hold the account and the payment relationship. Novo has created the recurring relationship and delivered it through intermediaries who own it. In the language I use for defensibility, that is a borrowed block on someone else’s land, and it is borrowed at exactly the point where the long half of the job is done.

The response elsewhere has been capacity, cost and focus: 9,000 job cuts announced in September 2025, about 5,000 in Denmark, and a halt to development of anything not tied to diabetes or obesity. Dave Snowden’s Cynefin framework explains why that feels productive and is not sufficient. Capacity programmes and portfolio focus are instruments for the complicated domain, where cause and effect are knowable and the answer can be planned. The question of what business model is permitted to serve a lifelong job sits in the complex domain, where the answer has to be discovered. Complicated-domain tools do not fail loudly in complex situations. They succeed at stabilising a structure that no longer fits, which is more dangerous, because it produces the sensation of progress.

Revenue model: most of the value falls outside it

Here is the arithmetic that should worry a strategist more than the share price.

The value created by keeping a population at a healthy weight consists overwhelmingly of costs that never arise. Those savings appear in insurers’ claims, in employers’ absence figures, in public health budgets, and in a person’s own later life. They appear years later. They appear in budgets Novo cannot invoice.

What Novo can invoice is a box. So it competes on the price of the box, and it competes hard: self-pay prices for the Wegovy pill run from $149 to $299 a month depending on dose, a Medicare arrangement took effect on July 1, 2026, and the list price halves in January 2027. Every one of those moves is rational. Every one of them prices the means.

Commoditisation is not the diagnosis here. It is the receipt. A company that is only able to sell the copyable part of the job will, in the end, compete on the price of the copyable part. The 87% volume growth alongside a 17% revenue decline is what that looks like once it reaches the accounts.

Which brings the subscription back into view, and it deserves a fair hearing rather than applause. A subscription is not automatically a new unit of account. It can also be a volume discount with a calendar attached. The test is simple and portable to any industry: is the company now paid for the maintained state, or still paid per box, merely in a more predictable rhythm? On current disclosure, the rhythm has changed and the unit has not. That is a real improvement in retention and not yet a change of business model. Sixteen years ago I described German publishers looking at the Kindle and seeing only a new distribution channel. Direct-to-consumer without a new unit of account is the same mistake wearing a different coat.

Corporate spirit: who is allowed to end the old Novo Nordisk business model

The fourth element is the one most analyses skip, and here it is the most concrete of the four.

Novo Nordisk is controlled by a foundation. At the end of 2025, Novo Holdings held about 28.1% of the share capital and about 77.3% of the votes through unlisted A shares that are never traded, and the Foundation’s articles oblige it to keep a majority of the votes. The structure did what it was built to do. In May 2025 the board removed the chief executive of eight years. In October the chairman and the independent directors stepped down, and at an extraordinary general meeting on November 14 shareholders approved the new board with 93% of votes. Lars Rebien Sørensen, chairman of the Foundation and chief executive of Novo Nordisk from 2000 to 2016, took the chair of the company as well. Some minority shareholders abstained in protest.

Two readings, both true. Foundation control makes decade-long commitments possible that a quarterly owner would block, and it is the reason the company could stay in metabolic disease for a century. And foundation control concentrates the authority to abandon a successful model in a small number of people who grew up inside it.

Ownership structure is not the context of a business model. It is part of it. Who may declare the old model finished is answered by the share register long before it is answered by a strategy document. I have argued that culture and values are constitutive rather than decorative, and that the people doing the work often see the shift earlier than the management formed by the old model. Replacing the leadership answers who decides. It does not answer what they are able to imagine.

The counterargument: maybe the classical model still works

The objection deserves its own section, because it is strong and the last eight months have strengthened it.

Perhaps this is simply a product race that Novo lost and is now winning back. In the head-to-head SURMOUNT-5 trial, mean weight change at week 72 was -20.2% with tirzepatide against -13.7% with semaglutide (Aronne et al., NEJM 2025). A real gap, not a marketing artefact. Then Novo did the classical pharmaceutical thing, and it worked. It launched the Wegovy pill in America on January 5, 2026, ahead of Lilly’s competing oral product. IQVIA reports that within eight weeks the pill had taken around a third of new-to-brand prescriptions, mostly through cash-pay channels, with roughly two-thirds of volume coming from patients new to any GLP-1 therapy. By mid-July, cumulative prescriptions for the pill had passed five million. Guidance went up twice.

The deeper version of the objection is more interesting than the numbers. Perhaps behaviour, food, coaching and daily life are simply not a drugmaker’s business, and should not be. Perhaps the right division of labour is exactly the one we have: the manufacturer makes the most effective molecule it can, others hold the relationship, and the falling price is the ordinary maturation of a product category rather than a sign of anything wrong. On that reading, Novo is not failing at a business model. It is being an excellent supplier in a market that is finally getting competitive.

I take that seriously, and part of it is right. Here is where it stops being right. A product move buys share. It does not buy price. In the same half-year in which the pill succeeded, American injectable revenue fell 17% at constant rates and adjusted gross margin fell nearly four percentage points, both attributed by the company to realised prices. Share returned; margin did not. Meanwhile the classical instrument is misfiring elsewhere: CagriSema matched tirzepatide on weight but not on blood-sugar control in the REIMAGINE 4 head-to-head, and the cardiovascular candidate ziltivekimab missed its primary endpoint in ZEUS. Sandoz has asked the Unified Patent Court to revoke a semaglutide patent covering eighteen European countries, Cipla has filed for a generic in the United States, exclusivity is already lost in Canada, and Novo is suing Lilly over comparative advertising. When the contest moves into courtrooms over patents and adjectives, the substance itself has stopped doing the differentiating.

A superior molecule is a single component, and single components are rentable rather than defensible. Durable advantage comes from the interplay of several components you own. The pen advantage of 1985 combined a molecule, a device, a manufacturing capability and a daily presence in a person’s life. The pill advantage of 2026 rests on being three months early.

Three business models a drugmaker is actually allowed to run

If the relationship is off limits, the honest question is what remains. Three positions, each with a real cost.

Relationship through the product. The 1985 move, repeated. Devices, formulation, dosing, packaging, anything that legitimately occupies a person’s day and is hard to copy. The oral formulation is a version of this and it worked commercially, which is not nothing. The cost is that formulation advantages are increasingly short-lived.

Orchestrator behind partners. Build the widest portfolio for the most patient types and let others hold the account, which is the current course. It is viable and it may well be the most profitable option for years. The cost is stated plainly: whoever holds the continuing contact can add products to it, and whoever holds only the product cannot retrofit the contact. That asymmetry does not resolve in favour of the supplier.

Revenue tied to the outcome rather than the transaction. If you cannot invoice the patient for the maintained state, invoice the party in whose budget the avoided cost lands. Employers and insurers can be contracted with, they can measure, and they have decades of horizon. This is the only one of the three that reconnects the revenue model to where the value actually falls, and it is the hardest, because it requires the company to be paid for something that does not happen.

None of these is a plan I can hand over. In a complex domain nobody can. What the four elements provide is not the answer but the right question: which element is the bottleneck, and what would we have to believe for this configuration to work? That is the difference between a diagnosis and a slogan, and it is why bad strategy travels so easily in situations like this one.

6 business model lessons every incumbent can learn

1. Ask whether the job your customers hire you for is an event or a continuing state. Most industries were shaped by the problem that existed at their founding, and that shape is invisible from inside. Pharma was built for episodes because infections and fractures are episodes. If the job has quietly become a state and your model still assumes an event, nothing in your operating metrics will tell you.
2. When the job outlasts your revenue window, you are monetising a fraction of the value you create. Patents, contracts, warranty periods, project scopes: each defines how long you get paid, and none of them is the customer’s timeline. Put the two clocks side by side and look at the gap.
3. If the value you create lands in someone else’s budget, you will end up competing on the price of the means. Prevention, durability, safety and maintenance all create value that shows up as costs not incurred, elsewhere, later. Where there is no mechanism to invoice the party that benefits, price erosion on the product is not a market failure. It is arithmetic.
4. Check what you are allowed to be, not only what you would like to be. Regulation, licensing and channel rules decide which elements you may own at all. A strategy that requires an element you are structurally barred from holding is not a bold strategy; it is not a strategy. Design inside the constraint, as the pen did, or change the element you can change.
5. Whoever holds the continuing contact can add products. Whoever holds only the product cannot retrofit the contact. This asymmetry is the single most underrated force in business model competition, and it applies far beyond healthcare. If a partner sits between you and the person using your product, they are not distributing for you. They are accumulating the asset.
6. Ownership structure is part of the business model, not its context. Who is permitted to declare the old model finished, and how quickly, is decided by the share register. Before designing a transformation, establish who holds the authority to end the thing that currently pays for everything, and whether their own history makes that thinkable.

One observation that is not quite a lesson. Novo Nordisk answered a version of this question once already, in 1985, with a pen, and it answered it inside the same regulatory constraint that binds it today. The cheapest source of imagination an incumbent has is usually its own history, and it is the source most reliably ignored, because the past gets filed as heritage rather than as method. Unlearning is the hard part; relearning is comparatively easy.

Understand. Imagine bigger. Act.

Commented bibliography

Aronne, L. J., et al. (2025). Tirzepatide as Compared with Semaglutide for the Treatment of Obesity. New England Journal of Medicine, 393(1), 26-36. doi:10.1056/NEJMoa2416394. The first published head-to-head trial of the two leading obesity drugs, and the source of the efficacy gap that makes the product-race counterargument serious rather than convenient.

The Economist (2025, December 15). The plan to rescue Novo Nordisk. Link. The reporting basis for the Doustdar interview, the supply and compounding sequence, the leadership changes and the December pricing position.

European Union (2001). Directive 2001/83/EC on the Community code relating to medicinal products for human use, consolidated version. Link. Article 88 requires member states to prohibit advertising of prescription-only medicines to the general public; Article 86 defines advertising broadly. The legal reason a European drugmaker cannot simply decide to own the customer relationship.

Rodriguez, P. J., et al. (2025). Discontinuation and reinitiation of dual-labeled GLP-1 receptor agonists among US adults with overweight or obesity. Link. Real-world persistence in a cohort of 125,474 adults, and the finding that discontinuation is far higher without a diabetes diagnosis, where coverage is weaker.

IQVIA (2026). The outlook for obesity from 2026 to 2030. Link. Uptake data for the oral products, including the finding that most oral volume comes from patients new to GLP-1 therapy, which is why the pill expanded the market rather than merely shifting it.

Novo Nordisk (2026). Financial report for the period 1 January to 30 June 2026, Form 6-K, August 4. Link. Primary source for volumes, realised prices, gross margin, guidance, the subscription programme, the distribution partners and the litigation. Read it rather than the headlines: the price-volume divergence is stated in the company’s own words.

Novo Nordisk Foundation (2025). Ownership. Link. Capital and voting percentages, and the articles of association that make the controlling stake permanent.

Snowden, D. J., & Boone, M. E. (2007). A Leader’s Framework for Decision Making. Harvard Business Review, 85(11), 68-76. Link. The Cynefin distinction between complicated and complex problems, and why the tools of the former are misleading in the latter.

Stähler, P. (2021). Das Richtige gründen. Werkzeugkasten für Unternehmer. 5. aktualisierte Auflage. Hamburg: Murmann Publishers. The four-element model used throughout: value proposition, value architecture, revenue model, corporate spirit.