Operations & compliance
Cost allocation keys
Cost allocation keys are the metrics used to divide a shared cost pool, such as IT or HR spend, between the group entities that benefit from it, and choosing the right key is central to defending an intercompany charge, since tax authorities expect the key to reflect actual usage rather than convenience. This matters in.
Common keys include headcount, revenue, operating costs, transaction volumes or square footage, and the appropriate choice depends entirely on the nature of the underlying cost, for example headcount suits HR support well while transaction volume better reflects the cost of running a shared billing platform. This matters in.
Applying one blanket key, usually revenue, to every shared cost pool is a frequent audit finding, since it rarely reflects how the underlying costs actually behave and can systematically overcharge larger entities for services they use proportionally less than smaller entities in the same group structure. This matters in.
In practice
What matters when applying cost allocation keys
- Match the key to cost behaviour
- Avoid one key for every cost pool
- Recalculate keys periodically
- Document the rationale for each key
- Test outcomes for reasonableness
Frequently asked
Common questions
What is the most common allocation key?+
Headcount and revenue are the most widely used allocation keys, but the right choice depends on which metric best reflects how each specific shared cost actually behaves across entities.
How often should allocation keys be updated?+
Allocation keys should be reviewed at least annually, and sooner if there is a material change in headcount, revenue mix, or the underlying cost pool being allocated. This matters in practice.
See how the tooling handles this in practice
Our transfer pricing tools calculate intercompany charges, benchmark financing and reconcile the intercompany ledger from your own data. Book a short walkthrough and we will show the workflow on a scenario that matches your group structure, rather than a generic demo dataset.