An ERP program cost is not a single figure; it is a distribution. The same scope quoted to two comparable companies can differ by a factor of two, because the price turns on legal entities, country count, data quality, and how much the business will standardise, not on revenue alone. The job of a benchmark is not to give you the number. It is to give you the range your number should sit inside, and the arithmetic to rebuild it.
How to read the range
Start from first-year software, because every other line is expressed as a multiple of it. If you do not have a real quote yet, estimate software as named users times an annual per-user rate; the rate is higher for a tier 1 platform than a tier 2 one. Then layer the implementation multiple, the run cost, and an allowance for your own people on top.
The point of expressing everything as a ratio of software is that a discount you negotiate on the licence flows through the whole model. A thirty percent software discount does not just cut the software line; it cuts the implementation and run lines that were sized from it.
An illustrative five-year shape
The table below shows the shape of a five-year program for a mid-market company, expressed as multiples of first-year software so it holds at any discount. Treat the multiples as the benchmark, not the dollar column.
| Line | Low | High |
|---|---|---|
| First-year software | 1.0x | 1.0x |
| Implementation | 1.0x | 3.0x |
| Run cost, years two to five | 0.6x | 1.0x |
| Your own people and change | +15% | +30% |
A program at the low end of every line is a standardised rollout in one country with clean data. A program at the high end is a multi-entity, multi-country program carrying legacy integrations and local processes that were never rationalised. Most real programs sit somewhere between, and the single most useful exercise is deciding, honestly, which end your organisation is closer to.
What moves you up the range
Legal entities with their own statutory reporting, countries with their own tax and payroll rules, the number of integrations counted from an actual interface list, and the share of processes that differ by site for reasons of habit rather than necessity. None of these show up in a revenue figure, which is why two companies of the same size get very different quotes.