The margin you believe you gain by making it yourself is not a benefit of making. It is a benefit of scale — and if you do not already have the scale, bringing production in-house will cost you the margin you were trying to capture.
A manufacturing executive brings a paper to the board. The argument is clean: the contract manufacturer takes a visible slice of every unit sold, the volumes are now substantial, and the capital required to bring the line in-house has a payback the finance function can live with. Recapture the slice, own the quality, control the schedule. The numbers work.
They work because of an assumption nobody in the room states out loud, and which the paper does not test: that the plant will run.
This is the most common structural error in make-or-buy analysis, and it is not an arithmetic mistake. It is a category mistake. The paper presents a margin recapture as a benefit of manufacturing, when it is in fact a benefit of high, continuous utilisation that manufacturing merely allows you to collect. Those are different propositions with different risk profiles, and only one of them is a decision the board is actually being asked to approve.
The Strategic Context
Every enterprise that converts inputs into outputs faces a version of this question, and it is rarely confined to factories. A professional services firm deciding whether to build a delivery team or subcontract; a bank deciding whether to run its own data platform or licence one — all are making the same structural choice about where the boundary of the firm sits.
What makes the question strategic rather than operational is that the boundary is expensive to move and slow to move back. Outsourcing an activity is generally reversible at the cost of a notice period. Insourcing it is reversible at the cost of writing off plant, releasing people, and rebuilding a supplier relationship you have just spent two years telling a counterparty you no longer need. The asymmetry means the two directions deserve different evidentiary standards, and they almost never receive them.
The conventional framing — cheaper to make or cheaper to buy — is insufficient for three reasons. It treats a fixed-cost commitment as though it were a variable cost. It ignores what the decision does to the organisation's capacity to learn. And it says nothing about what happens to the firm's position when demand moves, which is the only condition under which the decision is genuinely tested.
What Leaders Commonly Misread
The margin is read as earned rather than conditional. A contract manufacturer's margin is visible on every invoice, which makes it feel like leakage. It is not leakage; it is the price of someone else carrying the fixed cost of capacity, the risk of idle plant, and the burden of technology currency. When you bring the activity in-house you do not delete that cost — you assume it. Whether you are better off depends entirely on whether your utilisation is high enough to spread it more cheaply than theirs.
Idle capacity is treated as a timing problem rather than a cost. Plant that runs two months in twelve does not cost one-sixth as much as plant that runs continuously. It costs the same and produces a sixth as much, which is to say it inflates unit cost rather than reducing it. Seasonal, campaign-driven or project-lumpy demand is precisely the pattern that makes in-house production expensive, and precisely the pattern most likely to be described in a board paper as "growing".
Quality control is assumed to follow ownership. Owning a line gives you the authority to control quality. It does not give you the capability. If the technology is not mastered internally — genuinely mastered, in the sense that your own people can diagnose a process excursion without calling the vendor — insourcing transfers a quality problem from an organisation that has solved it to one that has not.
The decision is taken once and never revisited. Make-or-buy is treated as a structural fact of the business rather than a position that should be re-tested when volume, technology or the supply market moves. Firms discover they have been making something for a decade that the market now supplies better and cheaper, and discover it through a competitor's pricing rather than through their own review cycle.
Reframing the Issue
The useful reframe is this: make-or-buy is not a question about cost. It is a question about which capabilities the enterprise intends to own, and what it is willing to pay in fixed commitment and lost flexibility to own them.
That reframe changes what evidence the decision needs. A cost comparison asks: which is cheaper today? A capability decision asks a harder set of questions. Is this activity a source of durable advantage, or is it a commodity the market supplies competently? Does controlling it let us learn something we could not otherwise learn? And what happens to our cost position if demand falls by a third?
The distinction matters because the two questions can point in opposite directions and frequently do. An activity can be cheaper to make and still be wrong to make, because it converts flexible cost into fixed cost at a moment when the demand signal is weak. An activity can be more expensive to make and still be right to make, because it is where the product's differentiation actually lives and outsourcing it would hollow out the thing customers pay for.
The Five Conditions, and Why Four of Them Are Usually Assumed
A defensible in-house decision rests on five conditions holding simultaneously. Most papers demonstrate one and assume the rest.
Sufficient volume. Scale economies must be genuinely achievable at your volumes, not at the volumes in the growth case.
Regular, continuous demand. The plant must be able to run at high utilisation across the year. Lumpy demand is the condition under which in-house production destroys value most reliably, and it is the condition least likely to be examined, because annual volume totals conceal it.
Mastery of the production technology. Not access to it — mastery. The organisation must be able to hold quality without external dependence, including when something goes wrong.
Capital available at a cost that reflects the risk. The commitment is plant, working capital, and a labour force. Each is slow to unwind.
Alternative use for the capacity. If the line can only make this product for this customer, the enterprise has concentrated risk, not diversified it.
Where several of these fail, the honest conclusion is not "outsource because it is cheaper". It is "outsource because we do not have the demand profile that makes ownership pay, and the moment we do, this decision should be re-opened".
What the Boundary Does to Learning
Firms that outsource a production process progressively lose the ability to specify it well. The first generation of engineers who wrote the specification understood what each tolerance was for; the second generation inherits the document. Over time the enterprise retains the design of the product and loses the manufacturability of it, and discovers this when it attempts a step change and finds nobody internally can say whether the change is feasible.
The mirror risk is real and less discussed. Firms that insource everything accumulate obligations to keep technologies current that they have no scale to justify, and slowly become worse than the market at activities they insisted on owning. The supplier is investing across dozens of customers; you are investing across one.
Neither failure is a cost problem, and neither shows up in a payback calculation. Both are capability problems, and they are the reason the decision belongs in a strategy discussion rather than a procurement one. [Related article: We Bought the System. Did We Buy the Outcome?] examines the same failure in the context of enterprise systems; here the reverse error is at least as common — building a capability without asking what it must be good at.
What You Give Away When You Outsource
Outsourcing transfers more than a manufacturing step. It transfers process knowledge, tolerances, supplier lists, failure histories and, over time, the ability to make the thing at all. The commercial protections that are supposed to contain that transfer — confidentiality obligations, ownership clauses, registered rights — are worth having and are routinely relied on more heavily than they deserve to be.
[Related article: What Did You Hand Your Contract Manufacturer That You Cannot Get Back?] deals with what actually survives the end of the relationship and which protections stop at a border. The relevant point for the make-or-buy decision is narrower. The strength of your legal position should be an input to the boundary decision, not a consolation after it. If the activity carries knowledge you could not tolerate a counterparty retaining, that is an argument for making it, and it should appear in the paper alongside the unit cost.
Decision Framework
Run the decision in three passes. Most organisations run only the first.
Pass one — the cost comparison, corrected. Recompute the in-house case at realistic utilisation rather than nameplate capacity, and separately at seventy per cent of forecast demand. Include the cost of quality assurance, technology currency, and the management attention the activity will consume. If the advantage disappears at seventy per cent demand, you are not making a cost decision; you are making a bet on the forecast.
Pass two — the capability test. For each activity under consideration:
| Question | If yes | If no |
|---|---|---|
| Is this where our differentiation lives? | Bias to make | Bias to buy |
| Do we master the technology today? | Make is feasible | Make is a capability project first |
| Would losing this knowledge weaken our ability to specify? | Bias to make, or invest in retained expertise | Buy is low-risk |
| Does the market supply this competently and competitively? | Buy is efficient | Make may be necessary |
| Can the capacity serve more than one product or customer? | Make is less concentrated | Make concentrates risk |
Pass three — the reversibility test. State plainly what it would cost to undo the decision in three years, and who would carry that cost. If insourcing, that is written-down plant and released people. If outsourcing, that is a rebuilt capability and a supplier who now knows your volumes. A decision whose reversal cost nobody has estimated has not been evaluated.
Set an explicit review trigger at the point of decision: a volume threshold, a technology change, or a defined date. Make-or-buy decisions that are never revisited become organisational facts, and organisational facts are difficult to argue with.
From Strategy to Execution
Immediately. Re-run the utilisation assumption in any live make-or-buy paper and require the seventy-per-cent-demand case to be shown alongside the base case. Separate the margin recapture from the scale condition in the paper's own language, so the board is approving the proposition it thinks it is approving.
Over the medium term. Build a standing view of which activities the enterprise intends to own and why, reviewed annually rather than assembled when a proposal appears. Where an activity is outsourced but carries specification-critical knowledge, fund retained expertise deliberately — a small internal capability whose purpose is to remain able to challenge the supplier, not to duplicate them.
Over the longer term. Treat the boundary of the firm as a strategic position that moves with the market rather than a settled fact. The activities worth owning in a mature market are rarely the ones worth owning in an emerging one.
Signals to Monitor
- Utilisation drifting below the level the business case assumed — the single most reliable early indicator that an insourcing decision is failing.
- Quality escapes rising after an insourcing transition, particularly those requiring external diagnosis. This indicates the technology was accessed rather than mastered.
- Growing reliance on a supplier for specification input. When your own engineers cannot say whether a change is feasible without asking, the knowledge boundary has already moved.
- Consolidation in the supply market for an activity you have outsourced. Fewer credible suppliers changes the negotiation and may change the boundary decision.
- Capacity built for one product remaining single-purpose eighteen months after commissioning.
- A make-or-buy decision that has not been reviewed in three years on an activity where technology or volume has moved materially.
Questions for the Leadership Team
- In our most recent insourcing case, what utilisation did the business case assume — and what has actual utilisation been since?
- Which of the activities we outsource carry knowledge we could not tolerate a counterparty retaining, and what have we done about that beyond contracting?
- For each activity we perform in-house, can we say what it would cost to exit — and has anyone estimated it?
- Where we have outsourced, do we still have the internal expertise to challenge the supplier's technical advice, or only to accept it?
- Which of our boundary decisions were made more than three years ago and have never been re-tested against current volumes and current supply markets?
- If demand fell by a third next year, which of our fixed production commitments would become the constraint on our recovery?
Closing Perspective
The make-or-buy decision looks like a cost question because it arrives with a cost case attached. It is better understood as a decision about what the enterprise intends to be good at, taken under a fixed-cost commitment that is far easier to make than to unwind.
The margin that motivates most insourcing proposals is real, but it is not a reward for owning the plant. It is a reward for running it — and the difference between those two things is the difference between a decision that improves the business and one that quietly converts a flexible cost base into a fixed one at the moment the market turns. What that fixed base costs when volume falls is a question the capital paper rarely answers: [Related article: What Is Your Quality Failure Costing You — and Can Finance Produce the Number?] takes up the part of this that the finance function usually cannot see.
The organisations that get this right are not the ones with a preference. They are the ones that can state, for every activity above a materiality threshold, which side of the boundary it sits on, why, and what would have to change for that to be wrong.
About the author
Kevin Jogin is Founder & Principal Advisor at EraNorth. Meet the Founder.