Proving a maintenance gain: the baseline protocol

There is a pattern in efficiency projects that repeats with uncomfortable predictability. The technical work is sound, the indicator improves, everyone agrees there was a gain — and the measurement stalls. Not through bad faith, but because the ruler was never agreed. Once each side reconstructs the calculation after the result appears, each side arrives at a different number, and both are defensible.

The baseline exists to prevent this. It is the agreed numerical picture of performance before the intervention, and it is the most important document in any contract carrying a variable fee — more important, in fact, than the technical scope description.

Why it must be signed first

A gain is always the difference between what happened and what would have happened. The second half of that sentence is counterfactual: it is not observable. The baseline is the convention that replaces the counterfactual with something verifiable — the performance of a historical period, measured through defined indicators, under declared conditions.

A convention only works if it is agreed in advance. Once the result is visible, every methodological choice acquires an owner and a beneficiary: the more favourable reference period, the indicator that moved most, the exclusion that helps. The discussion stops being technical.

Hence a rule with no useful exception: without a baseline agreed in writing before the work starts, there is no gain measurement — only a negotiation with decimal places.

The ten items

A protocol that survives its first measurement fixes at least ten things:

1. Historical reference period. Which window represents prior performance, and why. Atypical periods are identified in the definition, not afterwards.
2. Indicators and official sources. Which system is the source of truth for each figure. Where two systems disagree — and they do — the protocol says which prevails.
3. Measurement scope. Which sections, asset families and operations are included. Elastic scope is the most common origin of dispute.
4. Excluded events. Force majeure, strikes, extreme weather and client decisions unrelated to the product. Listed beforehand, not selected afterwards.
5. Volume treatment. If volume rises 15%, cost per kilometre maintained moves for reasons unrelated to the intervention. The protocol states how to normalise.
6. Inflation and operational change. Input cost adjustments, a change in the client’s own maintenance policy, a new corridor entering service.
7. Measurement window. Quarterly, half-yearly or annual — and the cut-off date of each.
8. Attribution rule. When several client initiatives contribute to the same indicator, how the gain is apportioned. This is the item that generates the most friction when left open.
9. Joint validation committee. Who approves the figure, with what composition and quorum — plus audit rights for both parties over methodology and data.
10. Floor and cap. Contractual limits protecting both sides from unforeseen extremes, and in practice what makes the model acceptable to a conservative procurement committee.

The eleventh item

One item rarely appears in standard contract templates and settles most measurement disputes: the treatment of recommendations the client chose not to execute.

The situation is common. The system recommends a sequence; the client, for its own reasons — budget, competing priority, crew availability — executes part of it. At the end of the period, the measured gain is lower than the modelled one.

Without a prior rule, each side has a reasonable argument. The supplier will say the shortfall is due to non-execution. The client will say it pays only for what happened. Both are right, and the commercial relationship pays the bill.

The rule that works has two halves, and both must be written: a recommendation not executed generates no charge on unrealised gain — and cannot be used to invalidate the baseline either. The first half protects the client from being charged for a counterfactual. The second protects the supplier from having an entire period’s result contested because of decisions that were not its own.

Not bureaucracy — the thing that makes the result usable

There is a common reading of this work as excessive formalism. It disappears at the first measurement committee.

A number produced by a well-built protocol is usable in three places where an informal number is not: in controlling, which must recognise the gain in results; in internal audit, which must trace every figure to source; and at board level, where someone will ask whether what is being presented is a result or an estimate.

That is why the proof-of-value methodology is treated as a product, versioned and documented, rather than reinvented per contract — and why it travels with the platform’s audit trail: input data, model version, recommendation issued and decision recorded must all be retrievable the moment someone asks why that number is that number.