Outgrace, home
StrategyFramework

Build, buy or partner: how to choose before the proposals arrive

A capability gap is not settled until the company has said what time, control, cost and reversibility are worth. A five-pillar test, with minimums set in advance, makes that choice explicit.

Most capability gaps arrive as a line in a budget. A finance lead sees a request for an analytics platform, a product director asks for four engineers, and someone points out that a supplier can do the work by spring. Three answers are now on the table to a question nobody has written down. Build, buy and partner are not philosophies. They are three ways of paying for a capability, and each one spends something different: time, control, cash or the freedom to change your mind. The test below makes those costs visible before anyone signs. It has five pillars: time, control, cost, reversibility and where the knowledge lives. Score each option against each pillar as yes, no or not yet, and let no option win on a total that hides a failure on one pillar.

Chapter one: 1. Time to a working capability

Start with the date the business needs the capability, not the date a supplier promises it. Building often looks slowest on paper, yet it is sometimes the only route to a capability that does not yet exist inside the company. Buying can be quick to sign and slow to become useful, because the vendor's product still has to be fitted to your data, your approval chain and your staff. A partner can arrive fastest and then stall over who owns decisions. In our experience, the first date a supplier offers is the signature date, not the date the thing works.

The test is simple. Write down the date by which the capability must run in production. Then ask each option for the date it would be running in production, and ask for that date in writing, with the conditions attached. Discount any answer that describes only a pilot, a demonstration or a steering committee. If an option cannot commit to a production date, it has not yet cleared this pillar, however attractive its price.

Chapter two: 2. Control over what differentiates you

Separate the capability your customers pay for from the capability they never see. Keep the first in-house or build it with care. Buy the second without apology. A payroll engine is rarely a source of advantage. A pricing model your sales team trusts may be. The question to ask is blunt: if the supplier changed its product, its prices or its staff next year, would your competitive position change? If the answer is yes, control matters more than speed, and the option that gives you control has to clear this pillar before the others are compared.

Control has a second meaning that leaders often miss. It is the ability to change the work without asking permission. A contract that fixes the scope of a service can leave a company unable to alter the thing it has bought. Ask who may change a configuration, a data model or a product roadmap, and how long that change takes under the agreement. If the answer is a change request, a steering group and a three-month wait, the company does not control the capability, however much it pays for it.

Chapter three: 3. Cost, counted all the way to exit

Price each option over its full life, including the cost of leaving it. Build costs are mostly people, and the work those people stop doing. Buy costs are licences, integration and the internal staff who manage the vendor. Partner costs are often the least visible: shared meetings, joint governance and the decision rights handed to someone outside the company. A proposal that shows only year-one cost is incomplete by design. Ask every option for a five-year figure, and ask what the figure assumes about price rises, staff turnover and scope changes.

Treat any stated return with suspicion until you know how it is measured. The World Bank states that each $1 from its shareholders translates into $10 in development finance over 10 years. That is the institution's own claim about its model, and the statement does not show how the figure is built. The same discipline applies to a supplier's claimed savings or a partner's projected gains. Before you accept a multiplier, ask for its definition, its base year and the person who checked it. If the answers are vague, the multiplier is a sales figure.

Chapter four: 4. Reversibility

This is the pillar most boards leave out, and it decides how much a wrong call will cost. Ask what it would take to stop: notice periods, data return, ownership of code and documentation, staff transfer, and the cost of running the work in-house for a year while a replacement is found. An option you cannot exit inside a quarter should be marked as a board-level decision rather than a procurement decision. The board should see the exit cost before it sees the entry price.

Knowing when to stop is hard to do well. NBER Working Paper 34755, by Bailey, Fehder, Floyd, Hochberg and Lee, reports a randomised field experiment with innovation-driven entrepreneurs, testing whether training helps founders recognise when to stop. An NBER bulletin on the same work says entrepreneurship training may cause some startups to shut down sooner while improving outcomes for others. The lesson for a leader is not that stopping is good. It is that the stop rule has to be written before the project starts, because the person who runs a project is usually the last to see the case for ending it. The rule should name the measure, the threshold and the person who may call the stop.

Chapter five: 5. Who keeps the knowledge

Ask where the knowledge that makes the capability work will sit in three years. With a build, it sits with your staff, which is the point and also the risk. With a buy, it often sits with the vendor's implementation team, and it leaves when the contract ends. A partner can split it across two organisations, so that neither owns the whole. The test is to name the three people who could run the capability without the supplier on the day after the contract ends. If you cannot name them, you are renting the capability, not acquiring it. That may still be the right answer, but the board should choose it knowingly.

Borrowing from outside is normal. An NBER paper, 'Foreign Ideas and US Innovation', dated 17 November 2026, reports that 54% of US technologies build on at least one foreign technology. The same paper reports that counties with greater ancestral ties to a country absorb more of its technologies and invent more from them. The paper concerns technology and places, not firms, and it does not tell a leader which option to choose. It is a reminder that building on other people's work is the usual route, and that the useful question is who inside the company knows how to use it.

Before any option is shortlisted

Answer each question for every option. A no on any question means the option needs a written reason to proceed.

  • Can the option commit in writing to a date when the capability runs in production?
  • Does the option leave control over the capability's scope and roadmap with us?
  • Have we priced the option over five years, including the cost of exit?
  • Can the option be exited within one quarter, with our data and code returned?
  • Have we named three people who could run the capability without the supplier?
  • Were the criteria and their minimums written down before the proposals arrived?

Outgrace checklist

Where it breaks

The framework fails in three predictable places. The first is criteria written after the proposals arrive. Once a favoured supplier is in the room, the criteria start to fit the supplier. Write the five pillars and their minimums before the first meeting. Keep the document where the board can see every change, and record who made each change and why. A minimum that moves during a negotiation is no longer a minimum.

The second is measuring a partner by its outputs. Counts are easy to report and easy to mislead with. NBER Working Paper 35551, 'Chinese Sputnik Moments?', records a dispute over China's patent filings. One view holds that the filings show rapid technological advance. The other holds that they overstate real innovation, and some argue that incentive schemes drive incremental filings. If a count can be inflated, a supplier's deliverables can be too. Measure the outcome the business needs instead, such as time to a live decision, error rates in production or the share of work the internal team can take over.

The third is guidance that moves under you. The UK government's Green Book, which sets out how central government appraises spending, was updated after a 2025 review. According to gov.uk, the guidance was made much shorter and simpler, its annexes were merged into the core text, and business case guidance moved to a separate website. Guidance on agile digital and IT projects was removed from the Green Book page, though it remains on an accompanying webpage. The same update reframed value for money and benefit-cost ratios, but gov.uk does not set out the method. A company that adopts a published method should keep its own written copy, with the version date, so that the board can still reconstruct a decision after the source has changed.

We work with local partners and global partners. Because with more resources, we can accomplish more.

World Bank, public statement on its approach, World Bank

Partnerships are often sold on scale, and the sentence above is a good example of the pitch. It states an intention, not a result, and it does not say how the resources are shared or who decides. A partner's proposal will sound like it. On Monday, do three things. Write the five pillars on one page with a minimum for each. Ask every option for its production date, its exit terms and the three named people who would run the capability without the supplier. Send the page to the chair before the proposals are opened, and keep the answers on file for the review that will come when the contract is up for renewal.

Newsletter

The Briefing

One email a week: the leaders we profiled, the ideas worth stealing, and nothing else.

Weekly, free, one click to leave. Read our privacy policy.