Make or buy decisions: compare whole life costs, then decide on what cost cannot show

A make or buy analysis compares the whole life cost of producing something in house against the whole life cost of acquiring it, and finds the volume or duration at which the cheaper option changes. That break-even is the straightforward half. What settles most of these decisions is capability, capacity, control and risk, none of which the arithmetic prices.

The comparison is between two whole life costs, not two prices

The quickest way to make the analysis useless is to put a supplier’s quoted price next to an internal build estimate and treat the smaller number as the answer. Those two figures do not cover the same ground. One is the cost of a period of supply; the other is the cost of one phase of a thing that then has to be hosted, supported, patched, changed and eventually switched off.

A make or buy analysis is a comparison of total cost of ownership on both sides, over the same horizon, counting the same categories.

Cost	Making it	Buying it
Up front	Design, build, test	Licence and set-up fees
Running	Hosting, support, on call	Subscription and support tier
Change	Your own backlog	Change requests, at their price
People	Skills you must retain	Contract and supplier management
Ending	Decommissioning	Exit, migration, getting data out

UK government guidance treats this as a named piece of work rather than a judgement call. The Sourcing Playbook calls it a delivery model assessment, formerly known as a make versus buy assessment, and defines it as an evidence based approach to recommending whether to deliver a service in house, procure it from the market, or adopt a hybrid solution. It also fixes the timing: the assessment should take place early enough to inform the strategic outline case, which in plainer language means before anyone has said in public what they intend to do.

That puts make or buy downstream of a needs assessment and upstream of the business case. You cannot compare ways of meeting a requirement nobody has written down, and what the analysis produces is evidence for the options appraisal inside the case rather than a decision the case is left to justify.

Where the two cost lines cross, and the arithmetic that finds it

The formula is one division. Take the fixed cost that only the make option carries, and divide it by the amount making saves in each unit or each period. The result is the break-even point: below it the option with no fixed cost wins, above it the option carrying the fixed cost does.

Take a project that needs a document rendering component. Building it is estimated at £48,000 of effort, after which it costs about £600 a month to host and keep patched. A supported product doing the same job licenses at £2,600 a month with nothing to build. Making it therefore saves £2,000 a month once it exists, and £48,000 divided by £2,000 is 24 months.

Two months of horizon either side of that figure changes nothing, and a bad build estimate changes everything. Move the £48,000 twenty percent in either direction, which is well inside normal estimating error on work nobody has started, and the crossing moves to month 19 or month 29. The recurring figures are firmer, because one is a quoted price and the other is a rate the organisation already pays for similar things.

The factors that decide it once the money is level

Because the two totals so often land inside the error of the estimates that produced them, the honest reading of most analyses is that cost does not decide. Something else has to, and the something else is worth naming rather than leaving to the room.

Factor	The question it actually asks
Capability	Can we build this to a standard we would accept from a supplier, and retain the people who can maintain it?
Capacity	What already committed work does building this displace, and who has agreed to that?
Control	When we need a change, do we schedule it or do we request it and wait?
Confidentiality	Whose data leaves the building, and who owns the intellectual property in what gets built?
Risk	Which risks does the contract genuinely move, and which only look moved?
Exit	What does leaving cost in three years, at whatever the price is then?
Strategy	Is this the thing customers choose us for, or plumbing?

Risk is the one people misread most often. A contract can move liability, so that a supplier carries the cost of a failure. It cannot move accountability, because the organisation that chose to buy still has to explain the outcome to whoever cares about it. A supplier paying a service credit for an outage has not made the outage stop happening to your customers, and the difference between those two sentences is where a great deal of sourcing regret lives.

Buying delivery capacity is a different decision from buying a component

The word buy hides two quite different transactions. Buying a component means acquiring a thing that already exists and paying to keep using it. Buying delivery capacity means paying somebody else’s people to do work that has not been done yet, whether that is a whole workstream, a specialist team or an outcome defined in a statement of work.

The second is harder, because what arrives is not a product with a price and a feature list. It is a promise about future effort, priced before anyone knows how hard the work turns out to be, and the quality of what you get depends on people you did not interview.

It also does not remove the management. The APM is direct that contract management is vital for project managers, and of central importance to most projects, and sets out four controls that belong to the buying side: understanding the contract obligations, planned meetings to review progress and identify issues, formalised reporting and escalation routes, and a route to sharing perceptions of risk. None of that is free, and none of it is done by the supplier.

Once you are buying delivery, the contract stops being paperwork and starts being the operating manual. What counts as done, how a breach is handled, and the notice provisions with their deadlines all become part of running the work rather than part of signing for it. The legal concepts worth recognising in delivery covers that ground, and it is worth reading before the first thing goes wrong rather than after.

In a supply chain the decision repeats, and that changes it

Applied to a supply chain, the same comparison behaves differently, because the thing being decided is a volume that recurs rather than a build that happens once.

The break-even stops being a duration and becomes a quantity. The fixed cost of tooling up spreads across every unit produced, so the answer depends on a forecast of demand rather than on how long a component stays in service, and demand forecasts are the least reliable input anybody brings to the table.

Two further questions arrive that a project version never faces. The first is opportunity cost: making it consumes capacity that could have produced something else, so the relevant comparison is not making against buying but making against whatever the same capacity would otherwise have earned. The second is concentration. Buying moves a share of the supply to somebody outside your control, and the risk is not the price but what happens when that supplier has a fire, a strike or a better customer.

Where a make or buy analysis misleads you

Four failure modes, and the first is structural rather than careless.

The two sides of the comparison are not estimated to the same standard. The buy figure is a price somebody has committed to, checked by a sales process and usually written down. The make figure is your own estimate of work nobody has started, produced by people who would enjoy doing it. Those numbers carry very different confidence and are then subtracted from one another as though they did not.

The buy price is also not fixed for the horizon of the sum. A break-even at 24 months assumes a licence at today’s price for two years, which is a contractual question rather than an arithmetical one, and the answer is normally in a clause about annual uplift that nobody read before the spreadsheet was built.

Money already spent belongs in neither column. A half finished internal build is not an argument for finishing it; the only question is what each option costs from today. That is easy to write and genuinely hard to say out loud in a room where somebody has spent a year on the half finished thing.

And the decision has a shelf life. Capability moves, markets move, and the product that was not worth buying in 2024 may be the obvious answer now, or the supplier you chose may have been acquired by somebody with different priorities. The organisations that get this right are not the ones with the better spreadsheet. They are the ones that wrote down which assumptions the answer rested on, and went back when one of them stopped being true.