① Journal · Cost
If the business units that drive the mainframe peak never see the cost of it, the peak never comes down, and neither does the license bill. Chargeback is not an accounting exercise. It is a lever on consumption. Here are four models, what each rewards, and where each backfires.
Chargeback is a price signal. Point it at the peak.
Most mainframe software bills on capacity, and for the IBM Monthly License Charges stack that capacity is commonly the rolling four hour average peak, the R4HA that SCRT reports each month. Whatever drives that peak drives the bill. The problem in most large shops is that the people who shape the peak, the application teams running month end batch into the online window, the report jobs scheduled without thought to timing, never see the cost their choices create. The mainframe is a shared platform paid for centrally, so to the business unit it looks free. Free capacity gets used freely, the peak stays high, and the license cost with it.
A good chargeback model fixes the missing signal. It puts the cost of the peak in front of the units that drive it, in a currency they understand, so that moving a batch window or smoothing a workload becomes visibly worth doing. The model does not need to be perfect. It needs to be accurate enough to be fair and simple enough to be acted on, and it needs to point at the thing the license actually charges for. Get that right and chargeback becomes one of the cheapest cost optimization levers on the frame. Read this with our explainers on Workload License Charges and the cost optimization service.
Mainframe chargeback models · what each rewards and where it backfires
| Model | What it rewards | Where it backfires |
|---|---|---|
| R4HA peak allocation | Moving work out of the peak window; aligns to how MLC bills | Harder to explain; needs clean SCRT attribution by workload |
| MSU consumed | Overall efficiency; simple and fair for steady workloads | Understates the peak signal; a unit can be cheap yet drive the peak |
| Transaction or job count | Business relevance; units recognize their own volume | Weak link to capacity; high count can be low cost and vice versa |
| Tiered or flat split | Simplicity and predictability for budgeting | No behavior signal; the peak driver pays the same as the saver |
These are general patterns; the right model depends on how your estate is licensed and how much behavior change you need. Where IBM MLC dominates, an allocation that tracks the R4HA peak aligns the signal to the bill. Validate any attribution against independently checked SCRT data before charging on it.
If your largest cost is the R4HA peak, allocate against it. A model that charges on something the license does not bill for sends the wrong signal, however fair it feels. The units that drive the peak should feel the peak in their budget, because that is the only thing that makes moving a batch window worth their effort.
Charge for what the license charges for.
A chargeback nobody trusts is a chargeback nobody acts on. Attribution must reconcile to validated SCRT and R4HA data, and the method must be explainable to a non technical owner in a sentence. Over engineered models that only the capacity team can reconcile get quietly ignored, and an ignored model changes nothing.
A number nobody trusts changes no behavior.
Chargeback lowers consumption; the renewal locks the saving in. A peak reduced through better behavior should translate into a lower negotiated baseline, not just a quieter internal ledger. Time the optimization ahead of the renewal so the improved profile is reflected in the contract, not absorbed by the vendor as headroom.
Turn the lower peak into a lower baseline.
④ Where allocation becomes savings
A peak nobody pays for is a peak nobody lowers. Make the cost visible to the cause. Allocate to the peak, then bank the saving at renewal.
Typical reduction negotiated on renewal spend
Mainframe spend negotiated on the buyer side
Engagements delivered since 2019
Most mainframe software bills on capacity, commonly the rolling four hour peak measured by SCRT. Whatever drives the peak drives the bill. A model that ignores the peak gives business units no reason to move work off it, so the peak and the license cost stay high. A model aligned to the peak makes the cost visible to its cause, which changes behavior.
The ones that expose the cost of the peak to the units driving it. An R4HA peak allocation aligns directly to how IBM MLC software bills and rewards moving work out of the peak window. An MSU consumed model is simpler and fair for steady workloads but understates the peak signal. Fit the choice to how your estate is licensed.
When it is too coarse to influence behavior, a flat split charges the saver and the peak driver alike, or so complex nobody trusts the number. The goal is a model accurate enough to be fair and simple enough to act on. See MSU optimization quick wins.
Lower the peak through behavior, then carry the improved profile into the renewal so it becomes a lower negotiated baseline rather than vendor headroom. Time the optimization ahead of the contract. Our cost optimization service links the chargeback work to the renewal so the saving lands in the deal.
Related: Workload License Charges · MSU optimization quick wins · soft capping wins · cost optimization · MSU optimization
Audit notice or renewal under 18 months out? We mobilize within 48 hours.
Get expert help →