In-House or Outsourced Procurement: How to Decide

The in-house-versus-outsourced framing is the problem before it is the answer. Procurement is not one activity; it is a dozen with almost nothing in common. Negotiating a three-year agreement with a strategic supplier and keying a goods receipt are both procurement in the same sense that surgery and hospital catering are both healthcare.

Ask the question at the level of the function and you get one answer applied to all twelve, which is wrong for most of them. Ask it activity by activity and it usually answers itself.

#The test

For each activity, three questions. An activity that answers no to all three is a candidate; one that answers yes to any is not.

1. Does it require judgement that only your context supplies?

Not expertise — expertise is purchasable. Context: knowing that this supplier is also a customer, that the plant manager will not accept a change mid-quarter, that the board has a position on this category. Judgement that depends on things nobody wrote down cannot be delegated.

2. Does it carry an accountability you cannot transfer?

A regulated approval, a signature on a contract, an airworthiness or clinical determination. You can outsource the work of preparing the decision; you cannot outsource being the person who made it.

3. Is it a source of durable competitive advantage?

Rarely, but genuinely, yes — a supplier relationship that is itself a moat, a category where your buying is a strategic weapon. If it is, keep it and invest in it.

Applied honestly, the split in most mid-market organisations is lopsided. Strategy, policy, supplier relationships and final approvals sit on the retain side. The transactional and administrative layer — a large majority of the hours, and a small minority of the value-adding decisions — sits on the other.

#Running the test

#What the cost comparison misses

The usual business case compares fully-loaded internal cost per FTE against an outsourced rate and reports a difference. It is not wrong, but it is the least interesting part of the decision, and it leads people to outsource for the wrong reason and then be disappointed.

Three effects matter more, and none appears in that comparison.

1. Coverage. Most mid-market teams do not have someone doing catalogue maintenance or master data hygiene badly — they have nobody doing it at all, because it has never justified a hire. The comparison is not cheaper-versus-dearer; it is done-versus-not-done.
2. Capacity release. If your buyers spend 60% of their time on transactional work, moving it does not save 60% of their cost. It returns 60% of your scarcest people to work only they can do.
3. Continuity. An internal team of three loses a third of its capability when one person leaves. A contracted team with named cover does not, and does not stop while you recruit.

#The four objections worth taking seriously

"We will lose control of our suppliers"

You will if the arrangement is structured that way. You will not if the work runs inside your own ERP, under your own approval workflow, with your own thresholds — which is the only structure worth agreeing to. The test is simple: after go-live, does a purchase order still follow your rules and your sign-off? If the answer requires explanation, the design is wrong.

"Nobody will understand our business"

This is a real risk and it is a transition-design problem, not an argument against. It shows up as the same clarification being asked three times, exceptions handled inconsistently, and a slow erosion of trust in months two and three. What prevents it is a documented decision framework, a named team rather than a rotating pool, and an escalation path that puts genuinely ambiguous cases in front of someone in your organisation rather than guessing.

"Our data will be somewhere we cannot see"

Ask for specifics rather than assurances: named sub-processors, data residency options, breach notification terms, the security framework and its current status, and a DPA you can put in front of your own counsel. A provider that cannot produce those quickly is telling you something useful.

"It will be hard to unwind"

Which is why the exit terms belong in the first conversation, not the last. Notice period, data return format, documented process handback, and a defined transition-assistance period. A provider confident in the work will agree these readily; reluctance here is the single most useful signal available before signing.

#A sequence that de-risks it

The failure mode of procurement outsourcing is scope: a large first phase, a long transition, and a year before anyone knows whether it worked.

1. Baseline first. Volumes, cycle times, exception rates and cost, measured before anything moves. Without this, no one can later tell whether it worked.
2. Pick one high-volume, rules-based process — invoice processing or PO administration is the usual choice — and one entity or site.
3. Run it for a quarter against explicit service levels, with the baseline as the comparison.
4. Review honestly, including what went wrong. A transition with no problems in month one was not measured properly.
5. Extend by process, not by everything at once.

The first step is the one most often skipped and the one that matters most. An engagement with no baseline cannot be evaluated, only defended — which suits the provider considerably better than it suits you.