Evidence and maturity: the weight a number will bear

Most decarbonisation business cases fail on the same defect: a figure is presented without the two things needed to judge it — where it came from, and how developed the work behind it is. Precision is not the same as confidence.

Part
Part I — Why programmes stall
Chapter
2
Status
Summary published
Last reviewed
2026-08-04

A saving of 1,240 MWh per year looks more authoritative than “roughly 1,200”. It is usually less trustworthy, because the precision is inherited from a calculation whose inputs were assumed, and the assumption has been lost by the time the number reaches a decision-maker.

Two properties determine what a figure is good for, and neither is usually stated.

Evidence basis

Where the number came from. In descending order of strength:

  • Measured — from metering, at a resolution and duration that supports the claim.
  • Estimated — derived from measured data through a stated calculation.
  • Assumed — a reasonable value adopted where data does not exist.
  • Claimed — asserted by a third party, typically a vendor, and not independently verified.
  • Missing — no basis. Worth stating explicitly rather than quietly omitting.

The distinction that causes most trouble is between estimated and claimed. Vendor performance data adopted into a business case without verification becomes, on the page, indistinguishable from the client’s own measured data. Everyone downstream treats it as equally solid.

Maturity

How developed the work behind the figure is. A screening estimate and a tender price are both “the cost”, and they are not interchangeable:

  • Screening — order of magnitude. Good enough to decide what deserves further work. Not a basis for capital approval.
  • Concept — an option has been selected and outlined.
  • Feasibility — enough definition to support a decision, with the estimate basis and its exclusions stated.
  • Definition — scope fixed, interfaces understood.
  • Implementation ready — designed and specified to the point of procurement.

The central argument

A number without its evidence basis and its maturity is not information. It is an assertion wearing the costume of information.

The practical consequence is that decisions get made at the wrong level of confidence. A screening estimate carried into a capital paper is treated as a cost. When the delivered cost differs by the margin a screening estimate always carried, the programme is judged to have overrun — when in fact it was approved on a figure never meant to bear that weight.

The reputational damage lands on decarbonisation as a category. It becomes “the energy projects always cost more than they said”, which makes the next business case harder for reasons that have nothing to do with the next project.

Why labelling is resisted

Stating that a figure is assumed and at screening maturity feels like weakening your own case. It is not — it is the difference between a case that survives scrutiny and one that collapses under it. A capital committee that discovers an unlabelled assumption during questioning stops trusting the whole submission, including the parts that were sound.

There is also a straightforward institutional incentive: labelling makes it possible to check afterwards whether you were right. That is precisely why it is worth doing, and precisely why it is avoided.

Practical implications

  • Label every material figure with its basis and maturity, in the deliverable itself rather than in an appendix nobody opens.
  • Never adopt vendor performance data as your own. Label it as claimed and state what verification would take.
  • Match precision to basis. An assumed input does not produce a four-significant-figure output. Rounding is an honesty mechanism.
  • Make maturity a gate. Define which maturity is required for which decision, and refuse decisions taken below it.
  • Record assumptions where they can be tested later, so verification can identify which one was wrong rather than concluding only that the total missed.

A note on tools

Calculators make this worse if they are careless, because they produce confident-looking outputs from whatever is typed in. Any tool worth using shows its assumptions, its exclusions and its limitations alongside the result, and states which version produced it. A result you cannot audit later is not a result.

About this summary

This is a published summary of a chapter from a book in progress. It sets out the argument and its practical implications. The full manuscript is not published here.

It is general material, not engineering advice for a specific site. See the engineering disclaimer.

Discuss whether this fits

Tell us the problem, the site and the decision you need to reach. If this is not the right service we will say so and point you at the one that is.

United Kingdom and United States enquiries are routed to the same team. We will tell you plainly where we do and do not have local delivery capability.