① Guide · benchmarking spend
A single blended cost per MSU tells you almost nothing. The products that are overpriced sit hidden behind the ones that are cheap, and the average makes them invisible. A real benchmark breaks spend into product families, expresses each as cost per MSU, and tracks the trend. Here is how to build one that holds up in a negotiation.
48 hour mobilization Audit notice or renewal under 18 months out? We mobilize within 48 hours.
Get expert help →Mainframe software is licensed under mechanisms that do not compare. The IBM Monthly License Charge stack is sub-capacity, billed on a rolling four hour average through SCRT. IBM IPLA tools are a one time charge plus annual support. Broadcom (CA), BMC, Rocket Software, Software AG, and Syncsort (Precisely) each price on their own capacity or consumption basis. Average all of that into one cost per MSU and you have mixed things that were never the same unit, then erased the only signal worth having. A reasonable looking blended figure routinely conceals two or three families priced far above the rest, and those families are exactly where a renewal uplift or a rationalization earns its keep.
The defensible benchmark is granular. It segments spend, expresses each segment as annual cost against the MSU capacity it is licensed for, and tracks that unit cost across renewals so the drift is visible. The point is not a single magic number. It is the comparison: which family is high, which is rising, and which is rising without any new workload to justify it.
Break the estate into these segments and benchmark each on its own basis. The drivers differ, so the levers differ:
| Segment | Licensing basis | What moves the unit cost |
|---|---|---|
| IBM MLC stack | Sub-capacity, R4HA via SCRT | Peak four hour average, soft capping, Tailored Fit Pricing entry terms |
| IBM IPLA tools | One time charge plus annual support | Value Unit tier, support base, shelfware never retired |
| Broadcom (CA) | Capacity, MIPS migrating to MSU | Consumption baseline, idle bundled families, renewal uplift caps |
| BMC | Capacity or zConsumption Licensing | Suite bundling, true up basis, displaced IBM equivalents |
| Rocket Software | Per seat, per core, or capacity by product | Acquired portfolio terms, subscription transition, named user creep |
| Software AG and Precisely | Capacity based by product | Legacy contract scope, audit exposure on long held estates |
How to build it
A benchmark is not a report to file. It is the evidence base for a negotiation. The family priced well above the rest of your estate, and well above the pattern level range, is the one to target first, with a credible alternative or a rationalization plan behind it. The family rising without new workload is the one to challenge at renewal. The benchmark also disciplines the vendor's own framing: when Broadcom or IBM presents an uplift as the market rate, your own history per family is the counter evidence that the rate is drifting, not the workload. This is how our mainframe license negotiation work starts: with the estate measured the way the buyer should see it, not the way the vendor presents it. For benchmarking a single publisher in depth, see benchmarking IBM mainframe spend.
Cost per MSU, segmented by product family, not a single blended number. A blended figure hides the overpriced families behind the cheap ones, which is exactly where the savings live.
It averages away the signal and mixes licensing mechanisms that are not comparable. The benchmark that matters is granular, isolating the families where unit cost is high and rising. See cost per MSU benchmarks and drivers.
From your own history over time and from pattern level ranges, not a published price list nobody honors. Anyone quoting a single firm price per MSU is selling certainty that does not exist.
The family priced above the rest, with a credible alternative behind it, is the first target. Your own per family trend is the counter evidence when a vendor calls an uplift the market rate.
The explainer behind the metric, with a worked cost driver table.
The inventory the benchmark is built on, product by product.