What should be challenged in an infrastructure O&M cost model?

The decision

An operations and maintenance cost model usually reaches a decision-maker looking finished. It balances. Its totals reconcile to its inputs. Its structure is familiar. Someone is then asked to accept it, approve it, rely on it, or challenge it — and the instinct, reasonably, is to look for the error.

Arithmetic correctness is necessary. It is not sufficient. A model can calculate precisely and still be wrong about what operation is required, what asset has actually been inherited, why maintenance demand arose, what resources the work will take, whether its renewal assumptions are credible, and which expenditure properly belongs inside it at all.

It is worth being clear at the outset about what challenge is for. A cost model that is too low is not a success. The operating organisation has to fund what the infrastructure actually requires, under real access constraints, with real crews and real spares, and a model that under-provides transfers a problem rather than solving one. The purpose of challenge is not reduction. It is to establish whether the model’s assumptions reconcile with one another and with the evidence on which the decision depends.

Start with what the operation requires

Most O&M models are built by activity, with each activity driven by something. Published public-sector methodology sets this out plainly. The Federal Railroad Administration’s NEC FUTURE work groups operating and maintenance costs by functional area and assigns each a cost driver — track maintenance driven by track miles, station services by ridership, crew costs by labour hours — and organises the model’s inputs into physical characteristics, operating statistics, crew labour hours, fleet requirements, and ridership and revenue.

That structure has a consequence which is easy to miss. If the operating assumptions are wrong, every driver-based line is wrong in proportion, and none of it shows up as an arithmetic error. Service pattern, operating hours, seasonal profile, the availability the operation is expected to sustain, and the access and possession regime under which work can physically be done are all cost assumptions before they are operating ones. They belong at the front of a review, not the end.

Reconcile the asset and maintenance assumptions

Two questions follow, and they are different questions.

The first is what is actually being maintained. A model prices a described asset. An operating organisation inherits a built one — with its actual population, configuration, modifications, age profile and condition. Where the described and the inherited asset diverge, that difference can alter the cost basis — and it may not be visible in a model that continues to price the described asset.

The second is why maintenance demand exists. Planned demand, corrective demand and demand arising from a specific unresolved condition behave differently over time and should not be forecast the same way. This matters because of how unit costs are usually built. The FRA methodology derives them by dividing existing costs by existing cost-driver values, then applying the resulting rates to projected driver values. That is a sound method. It also means that whatever was in the historical cost is now in the rate, and whatever is in the rate is now in the forecast.

Challenge the cause, not merely the cost

This is where a senior review earns its place. A cost line is the last visible point in a longer sequence:

underlying condition → operational or inspection response → mitigation or corrective intervention → recorded O&M expenditure → historical actual → future model assumption

A model normally sees the right-hand end of that sequence. Assurance has to work back along it. The question is not whether the expenditure was incurred. It is why it arose.

Consider the pattern in which historical expenditure includes recurring intervention associated with an abnormal underlying condition. The expenditure is real. It was properly recorded. Extrapolating it forward is nonetheless a modelling decision, not an accounting one — because it assumes that the condition, and the response to it, are a legitimate steady state.

That assumption may be entirely correct. Some conditions will not be resolved, and the cost of living with them is a genuine cost of operating the asset. The point is narrower and harder to argue with: the model should say which it is. Historical maintenance expenditure is evidence of what has been spent. It is not necessarily evidence of what should continue to be spent.

Separate maintenance, lifecycle and abnormal expenditure

A second point makes the same argument from the classification side. Work performed by the maintenance organisation and recorded in the maintenance ledger is not, by that fact, maintenance demand — activity can arise from causes that sit outside the maintenance regime altogether. Its causal classification is a separate matter, and it is material to a forecast.

Three further bases sit alongside this and are commonly assumed rather than tested.

Resources. Crew hours, competence, access windows, spares holding and logistics are the physical constraints under which the money is actually spent. A model that assumes labour is available in the quantity and at the moment the plan requires is making an operational claim, not a financial one.

Lifecycle. Renewal and replacement timing usually rests on an assumed component life. What that assumption is founded on — manufacturer statement, population evidence, observed condition, or a figure inherited from an earlier model — is a reasonable thing for a reviewer to be told.

Commercial. Not everything recorded in an O&M ledger is an O&M cost. Scope boundaries, the treatment of work arising from causes outside the operator’s responsibility, and the risk allocation the model implicitly assumes all determine whether a line belongs in this model at all. Where a contract form applies, its allocation is a fact about the model’s inputs; what that allocation means in law is a separate question and depends on the contract and applicable law.

Six questions that should survive the review

A review has not finished its work until the organisation can answer these:

1. What operation is this model actually pricing, and does it match the service the organisation intends to run?
2. Which asset does it price — as specified, or as built and currently held?
3. Of the maintenance demand in the base data, how much is planned, how much corrective, and how much attributable to a specific unresolved condition?
4. Can the resource plan be delivered under the real access constraints, with the people and spares assumed?
5. What is each assumed component life founded on?
6. Which lines are in this model because they belong to it, and which because that is where the work happened to be recorded?

These are not a checklist to be completed. They are the positions a decision-maker should be able to hold under challenge from someone else.

The decision that follows

The output of a serious review is rarely a verdict. When the Union of International Railways reviewed operating and maintenance cost estimates for the California High-Speed Rail Authority’s 2012 business plan, it examined the model’s cost categories for consistency and completeness, then separately tested the assumptions behind them against operating practice — and concurred with most of them while identifying specific ones that needed refinement. That is the normal and useful shape of an answer.

So the decision that follows a challenge is not binary. It may be to accept the model, to re-price part of it, to re-scope what it covers, to investigate a cause before committing, or to accept it with a stated qualification. Which of those is right is the organisation’s decision. The review’s job is to make sure it is made on assumptions that have been seen.

Westheath reviews O&M cost models where the question is whether the model will hold.