How to show what IT costs each part of the business

Sooner or later someone asks what IT costs their part of the business. The honest answer is usually “we can tell you what we spend, not what you consume” — and closing that gap is a project with more politics in it than arithmetic.

Start by reconciling to the ledger

Whatever model you build, the total has to tie to the general ledger. A cost model that produces a different total than finance has is not a cost model; it is an opinion, and the first challenge will be about the difference rather than about the allocation.

This constraint shapes everything downstream. Build it with the finance team, in their chart of accounts, and agree the reconciliation before anyone sees a business unit number.

Separate what is consumed from what is shared

Three categories do most of the work:

• Directly attributable — an application only one business unit uses, devices issued to their staff, a team dedicated to them.
• Consumption-based — shared services where usage can be measured: storage, transactions, tickets, seats.
• Genuinely shared — the network, security, the leadership team. These are overhead and should be labelled as such.

Arguments happen in the third category. Naming it as overhead up front, rather than allocating it by a formula that looks precise, removes most of the heat.

Choose a driver people already accept

A perfect driver nobody understands is worse than a rough one everybody does. Headcount, devices, transactions, and named users are all defensible because a business leader can see the relationship between the number and their own decisions.

The test is whether a manager can change their cost by changing their behaviour. If they cannot, the allocation is a tax and will be treated as one.

Publish before you charge

Showback — publishing what each unit consumes, without moving money — does most of the useful work. It starts the conversation about demand, surfaces the applications nobody knew they were paying for, and lets the method be challenged while the stakes are low.

Charging real money should come later, if at all. Plenty of organizations get the entire benefit from showback and never take the next step.

Expect the first version to be wrong

Business units will find errors, and some of them will be real. Treat the first two cycles as calibration: publish, take the challenges, correct, republish. A model that has survived being argued with is trusted; one that arrived complete and unquestioned is not.

What this is actually for

Not recovering cost. The value is that demand conversations change when a business leader can see what their requests consume. Projects get sharper, low-value applications get retired by the people who own them, and the argument about the IT budget becomes an argument about business priorities — which is where it belonged.