Buy, build or partner is the decision about where a capability comes from once a strategy has named one the organisation does not hold. Ronald Coase gave the question its first formal answer in The Nature of the Firm, published in Economica in 1937. A firm expands, he argued, until the cost of organising one more transaction inside it matches the cost of buying that same transaction on the open market. Oliver Williamson turned the argument into a working theory in Markets and Hierarchies in 1975. He shared the 2009 Nobel prize in economics for his analysis of economic governance, and of the boundaries of the firm in particular. A promise, a revenue model, a price and a route to the customer each assume a capability that may not exist yet. The product manager is the person who has to say where each one of them will come from. When nobody asks the question at all, the answer that arrives by default is building.
The sections below set out where the boundary of an organisation comes from and the three ways a capability arrives. A test then separates the capability worth owning from the rest, and one worked comparison puts three year costs against each option. The last two sections cover what changes the answer later and where each syllabus puts the material.
Where the boundary of an organisation comes from
Coase was asking why firms exist at all. If markets set prices efficiently, every task could be bought from somebody outside, and a company would be one person signing contracts. His answer was that using the market costs something before any price is paid. Finding the right supplier, agreeing terms and enforcing the agreement all take time and attention, so a firm exists wherever doing the work internally is cheaper than paying those costs.
Williamson added the property that decides which way a particular task falls, and he called it asset specificity. An input any buyer could use is cheap to buy on the open market, because several suppliers compete for the work and moving between them costs little. An input shaped around one relationship leaves both sides exposed the moment the contract is signed, because neither side can leave without losing what it has put in. Organisations pull work of that kind inside. Asset specificity is the economist's version of a question a product team already asks, which is how much of this capability exists only because of what this product is for.
The three ways a capability arrives
Asset specificity explains where a capability should sit. Three arrangements carry that decision out, and the cost of each one behaves differently over time.
Build. The organisation makes the capability itself, which buys full control of the roadmap, the data and the experience. Building is the slowest of the three and the cost continues long after the first release. Robert Glass collected the figure in Facts and Fallacies of Software Engineering in 2002. His forty first fact records that maintenance takes between 40 and 80 per cent of the lifetime cost of a piece of software, and that the average is 60 per cent. A component that cost £100,000 to write is, on that average, a commitment of roughly £150,000 more before anybody retires it.
Buy. The organisation acquires the capability, which for a product team usually means a licence or a data set and occasionally means a company. A licence is the fast route to something that already exists, and the integration work behind it is underestimated far more often than the fee is. Acquiring a company carries a different order of risk. Roger Martin wrote in Harvard Business Review in June 2016 that between 70 and 90 per cent of acquisitions fail. His explanation is that acquirers concentrate on what the target will give them, when the question that predicts success is what they will give the target.
Partner. Somebody else supplies the capability under an agreement, which costs little at the start and arrives quickly. The partner's roadmap, pricing and survival all become inputs to the product on the day the agreement is signed. Reddit repriced access to its data interface in 2023, and Christian Selig, who built the Apollo client for it, calculated on 31 May that the new terms would cost his app about twenty million dollars a year. Apollo closed on 30 June 2023. Every product built on somebody else's terms carries that risk, and the risk grows with each customer who comes to depend on the arrangement.
Most of these discussions stop at the three options and their prices. If a discussion stops there, it produces whichever answer looks cheapest in the current financial year.
Core and context as the test
The test that settles the choice has nothing to do with price. Geoffrey Moore supplied it in Dealing with Darwin in 2005. Core is any activity that makes a buyer prefer this product at the moment they choose what to buy. Context is everything else, however necessary any of it happens to be. Moore crosses that line with a second one, which asks whether the activity is mission critical, meaning whether a failure does immediate and serious harm.
Geoffrey Moore crossed core against mission critical in 2005, and the four cells give four different instructions. Only the top left cell justifies spending the team that knows the product best.
Marking too much of the product as core is the common error. A capability feels core when the team built it, when it is technically demanding or when it is the part the engineers enjoy. Moore's definition admits none of those reasons. The only question it asks is whether a buyer choosing between this product and the alternative would decide on that capability. For most capabilities the answer is no.
The comparison worked through for one capability
A capability that fails the core test still has to come from somewhere, and the three options price differently over time. The refrigeration maintenance product used in the market sizing is one worked example. It sells to grocery chains in the United Kingdom, and it predicts from sensor data which unit will fail next. The product now needs to route engineers between sites efficiently. Route optimisation is a solved problem with several suppliers, and the failure prediction is what a grocery chain is paying for.
Every figure below is an assumption that can be argued with on its own. Engineers are costed at £110,000 a year once employment costs are included, and the partner takes a fifth of licence revenue, which reaches £917,000 in the third year.
| Option | Year one | Years two and three | Three year total | Weeks until customers have it |
|---|---|---|---|---|
| Build it | £165,000, two engineers for nine months | £33,000 a year in upkeep | £231,000 | 39 |
| Buy a routing licence | £48,000 licence and £30,000 of integration | £48,000 a year | £174,000 | 10 |
| Partner with a field service platform | A fifth of £200,000 of licence revenue | A fifth of £500,000, then a fifth of £917,000 | £323,400 | 6 |
Partnering is the cheapest way to start and the most expensive by the third year. Building does the opposite, because it costs most in the first year and little after that. Neither observation settles anything on its own. Route optimisation is context under Moore's definition, because no grocery chain has ever chosen a maintenance product for the quality of its routing, so £231,000 is the smaller of the two prices the build carries. The larger price is nine months of the only two engineers who understand the failure model, spent on a problem three suppliers have already solved.
Writing the comparison out earns its place, because it puts the opportunity cost in the same units as the licence fee. A finance director will argue for an hour about a £48,000 licence. If the alternative is written on the page beside it, the argument usually ends in a minute.
What changes the answer later
The comparison settles the decision for this year and for no longer than that. A capability that is core today becomes context as competitors match it. That is Moore's own argument, and it is why he treats core and context as positions on a moving line. A component written in 2021, because nothing comparable existed then, is a maintenance obligation by 2026 once four suppliers sell the same thing, and moving to one of them releases the people who have been keeping it alive. An agreement made when a capability was scarce becomes expensive once three vendors offer it, because the price was set against a scarcity that has since gone.
Two questions belong on that review. The first asks whether the capability still decides a purchase. The second asks what reversing the decision would cost now, after a hundred customers have come to depend on it. The cost of reversing rises quietly and never appears in any budget, so the review belongs on the same cycle as the competitive picture. The review itself costs an hour. If nobody holds it, a decision taken once in a planning meeting three years ago goes on governing a share of the engineering budget, long after the condition that justified it stopped holding.
All three options assume the organisation assembles the capability itself, whether by writing it, buying it or contracting for it. A fourth arrangement leaves that assembly to companies outside the organisation altogether. What the fourth arrangement demands in return is where the choices about how to win end.
Where each syllabus puts it
The Pragmatic Framework places Buy Build or Partner in its Business category, beside Business Plan, Pricing, Product Profitability and Innovation. The instruction there is to determine the most effective way to deliver a complete solution to an identified market problem, analysing whether to buy, to build or to partner wherever the current offering has gaps. The Focus course covers the Business and Focus categories together, so a Pragmatic PMC candidate meets this material beside market definition, distribution strategy and product profitability.
The Product Strategy Practitioner course reaches the same decision from the capability side. A promise, a revenue model, a price and a route to the customer have all been chosen by that point, and each of them has named something the organisation cannot yet do. The three options are the same three options. The difference is that the gap now has a strategy behind it, so the test is whether the missing capability carries the advantage the strategy was built on.