What a number owes you

There is a moment in every sustainability reporting cycle that nobody puts in the process documentation. The deadline is close, the consolidation is nearly done, and there are cells that are still empty. Somebody makes a decision about those cells. Often that decision is sensible and well-reasoned. Almost always, the reasoning does not survive into the report.

A year later, an auditor asks about one of those figures. The person who made the decision has moved teams, or moved on. The spreadsheet has been through four versions. The email that explained the assumption is in an archive nobody can search. The number is still there, sitting in a published report, and nobody can fully reconstruct how it got there.

This is not a story about carelessness. The people who do this work are usually careful and well-intentioned. It is a story about tools that do not preserve what careful people know.

Zero is a claim

The specific failure that causes the most damage is small enough to be invisible. A value is missing, and the system stores it as nought.

Those are entirely different assertions. “We do not have this data” is a statement about our knowledge. “This value is zero” is a statement about the world. A spreadsheet cannot tell them apart, so it renders them identically, and once they are rendered identically the distinction is gone forever – not hidden, gone. No amount of later diligence recovers it.

The consequences compound quietly. A total that includes a silent zero is understated, and nobody can tell by how much. An intensity ratio calculated from partial coverage looks like a real performance figure. A year-on-year comparison between two periods with different coverage reads as improvement. None of these are lies. All of them are wrong in a way that will not be found until somebody outside the organisation goes looking.

Four states, not two

The fix is not complicated to describe. It is that a value needs more possible states than present and absent.

Recorded. A figure with a known source and a known method. It came from somewhere, and you can go there.

Approved estimate. Somebody with the authority to do so decided to estimate, using a method they stated, and put their name to the decision. This is a legitimate and often necessary thing to do. It stops being legitimate the moment it becomes indistinguishable from a measurement.

Unavailable. The value exists in the world. You cannot obtain it. That is a fact worth recording, along with why.

Not applicable. The question does not apply to this entity, this activity or this period. This is a real answer and it should not look like a gap. Reporting “not applicable” as missing understates completeness just as badly as reporting missing as zero overstates it.

Four states. Most systems have two, and the collapse from four to two is where defensibility is lost.

What a report looks like when it is honest

A report built this way looks worse than one that is not. This is the thing that makes the discipline hard to sell.

It shows coverage that is less than total. It marks estimates as estimates in the tables, not in a footnote at the back. Its completeness figures are lower than a competitor’s because a competitor’s completeness figure is counting silent zeroes as data.

But it has a property the other report does not: every one of its weaknesses is one you found first. There is a specific kind of institutional damage that happens when an external party discovers a gap in a published sustainability report that the publishing organisation did not know about. It is not the gap that does the damage. It is the discovery that the organisation did not know.

Where a figure came from is part of the figure

The other half of this is where a figure came from, and the mistake is to treat that as metadata – something attached to the value, useful for audit, not part of the value itself.

Where a number came from is part of what it means. Two figures with the same value and different origins are different figures. One is a metered reading from a device with a calibration record; one is a supplier’s own estimate from a template they filled in at speed. Treating them as the same number, carrying the same weight in the same total, is a mistake — and it is the mistake most sustainability data rests on.

Which is why nothing in the record is quietly replaced. If a figure can be overwritten, what you know about it lasts only as long as the last person who edited it. If corrections are added alongside what they correct, the history is a fact rather than a recollection.

The obligation

A number that goes into a published report has been asked to do something serious. An investor may allocate capital on it. A regulator may act on it. A board has certainly signed something because of it.

What it owes the people relying on it is straightforward: to say where it came from, to say how confident it is, and to say plainly where it does not know. Nothing about that is technically difficult. It is difficult because it requires a system built around the assumption that gaps are normal, that estimates are legitimate when they are declared, and that the honest answer is more valuable than the complete-looking one.

That is the system we are building.