Emissions data that holds up in a CSRD report
A freight emissions figure is only as defensible as the lane data beneath it; the tender is where that data is either captured clean or lost.
Computing a freight emissions figure is not the hard part of a CSRD report. The formulae are well established and the arithmetic is trivial. The hard part is standing behind the number when someone asks where each input came from — because an emissions figure inherits every weakness in the lane data beneath it, and freight lane data is where the weakness usually lives.
The figure is the easy part
Emissions for a lane come from a small set of inputs: how far the freight moved, by what mode, carrying what load, with which carrier. Multiply, sum, report. If those inputs are clean and sourced, the figure is defensible almost automatically. If they are approximated, averaged, or reconstructed from memory, the figure is a confident output resting on soft ground — and a reviewer who pulls one thread finds the whole number moves.
So the real question behind a CSRD emissions figure is not "is the calculation right" but "can each input be traced to something true." That question is answered at the tender, not at reporting time.
This reframes where the work belongs. Teams preparing a CSRD report often invest heavily in the emissions methodology — the factors, the boundaries, the calculation — and comparatively little in the lane data feeding it. The proportion is backwards. A defensible methodology applied to soft inputs produces a soft number wearing a rigorous coat; the same methodology applied to sourced inputs produces a number that holds. The scarce, valuable effort is in the inputs, and the inputs are set at the award, long before the report is ever opened.
The inputs that have to hold
- Distance — a real lane distance tied to a stable lane identifier, not a rounded guess between two countries.
- Mode — recorded per lane, because a blended assumption across a portfolio distorts every figure it touches.
- Load basis — the weight or unit basis the emissions are computed against, consistent across lanes.
- Carrier identity — the awarded carrier for the lane, so the figure attaches to who actually moved the freight.
- Period — the volume and window the figure covers, so the report states what it measured rather than implying it.
Every one of these is a field the award already handles. Captured at award time and kept, they turn the emissions report into a read of existing data. Left uncaptured, they force a reconstruction from invoices and contracts, which is both slow and — because it happens after the fact — far harder to defend.
Portfolio figures inherit portfolio gaps
A report does not present one lane; it presents a total across hundreds. That aggregation is where weak inputs do their quiet damage, because a total hides the state of its parts. A portfolio emissions figure computed over four hundred lanes looks equally confident whether every lane was sourced or forty were guessed. The number gives no sign of which. So the defensibility of the total is really the defensibility of its worst-sourced lane, and a reviewer who samples the report will find that lane, not the four hundred that were clean.
This is why coverage matters as much as accuracy. A figure that covers 92% of lanes on sourced data and flags the remaining 8% as estimated is more defensible than a figure that silently covers 100% by guessing the gaps. The first states what it knows and what it does not; the second asserts a completeness it has not earned. Reporting the coverage alongside the figure is not a weakness to hide — it is the thing that lets a reviewer trust the part you are confident about.
Estimates, honestly labelled
Some inputs will genuinely be unknown. A distance may be missing, a mode may be assumed for a minor lane. The disciplined response is not to fill the gap with a plausible number that later reads as fact. It is to label the estimate as an estimate and, where a figure cannot be sourced at all, to show it as an em-dash rather than a zero. A report that is honest about its gaps is more defensible than one that hides them, because the first reviewer to find a fabricated certainty stops trusting every other figure on the page.
Where software assists — flagging an outlier lane, suggesting a likely mode — treat the assistance as advisory. It can point you at the input worth checking, but the figure you report is one you have sourced and can defend, not one a model asserted on your behalf. The computation is deterministic; the judgement about what to trust stays yours.
It also means the emissions report and the freight award are not two separate exercises to be reconciled later. They are the same lane data, read for two purposes. A team that keeps them together — the awarded carrier, the distance, the mode, all captured once — produces both the award and the emissions figure from one source, and neither can quietly contradict the other. Kept apart, they drift, and the drift between what you awarded and what you reported is precisely the kind of inconsistency a reviewer is trained to find.
A CSRD emissions figure that holds up is not a triumph of calculation. It is the visible tip of a clean tender: lane data captured once, sourced, and honestly labelled, so the number on the report is one you can still explain a year after you filed it.