Companies that have committed to disclosing their carbon footprints are confronting a measurement problem that financial accounting does not face. The numbers that regulators, investors, and customers increasingly demand depend on data held not by the reporting company but by its suppliers, its customers, and a network of partners whose own measurement practices vary widely. The result is a system in which the accuracy of disclosure rests on a chain of estimates whose weakest link defines the credibility of the whole.

Carbon accounting is conventionally divided into three categories. Direct emissions from a company’s own operations are the most straightforward to quantify, drawing on fuel purchases, equipment registries, and similar internal records. Emissions associated with purchased electricity require slightly more effort but are bounded by what the company itself uses. The third category, covering emissions across the value chain — from the production of purchased materials to the use and disposal of sold products — is conceptually the largest for most companies and the hardest to measure. It is also the category where regulation and investor pressure are focusing most intensely.

The data problem in the third category is fundamental. To know the emissions associated with a component a company purchases, the company needs information from the supplier that produced it, which depends in turn on data from that supplier’s suppliers, and so on through a chain that may run many tiers deep. The company at the end of the chain is asked to report a number that summarizes activities it does not control and processes it does not observe. The number is necessary, in some form, for any meaningful disclosure to exist; the difficulty of obtaining it accurately is what makes much of the existing disclosure uneven.

In practice, companies fill the data gap with industry-average emissions factors, derived from databases that estimate the typical carbon intensity of producing a kilogram of a given material, a unit of a given service, or a dollar of a given input. These factors allow a number to be calculated, but they substitute averages for actuals and tend to obscure differences between suppliers that may differ substantially in their underlying practices. A company that has actually invested in lower-emissions production receives no credit in the calculation of a downstream buyer who uses an industry average; a company that has not invested faces no penalty. The averaging blurs distinctions that the system is meant to reveal.

The push to replace averages with actual supplier data has produced an emerging architecture of digital reporting. Standards organizations have developed common formats. Software providers have built platforms to collect and exchange the data. Larger companies have begun to require disclosure from their suppliers as a condition of doing business, propagating measurement requirements down through the chain. The architecture is slowly taking shape, but progress is uneven and many gaps remain, particularly for smaller suppliers without dedicated resources for the task.

The accounting framework also struggles with overlaps. The emissions associated with a tonne of steel show up in the value-chain accounts of every company that buys steel, directly or indirectly, throughout the chain. Aggregating across companies double- or triple-counts emissions in ways that obscure the actual total. Frameworks attempt to address this through guidance on attribution and allocation, but the question of whose emissions a given molecule of carbon represents has no purely technical answer. The choice is a convention, and different conventions produce different numbers.

The credibility of carbon disclosure depends increasingly on assurance, the equivalent of auditing for financial accounts. Independent reviewers are asked to verify that the numbers a company reports are calculated according to stated methods and supported by evidence. The assurance industry for sustainability reporting is growing rapidly, but it confronts the same data challenges as the companies it reviews. An assurer cannot verify supplier emissions any more accurately than the company can, and the resulting opinion often carries qualifications that reflect the limitations of the underlying data.

The stakes of getting the system right are substantial. Carbon disclosures feed into investor decisions, into the terms of sustainability-linked financing, and into the design of policies that price or regulate emissions. Inconsistent or unreliable numbers undermine each of these applications. The risk that the system produces false comfort, with companies and investors reassured by figures whose underlying data is too thin to support them, is a recurring concern of those who follow the field. The risk in the other direction, that the system produces such unreliable figures that it is widely dismissed, is also real.

The path forward, to the extent there is one, involves the slow improvement of supplier data, the convergence of accounting conventions, the maturing of assurance practices, and the integration of carbon reporting into the routine financial systems of companies rather than its treatment as a separate exercise by sustainability teams. None of these is a quick fix. The system that emerges over time will likely be one in which the numbers are better than they are today but never as precise as financial accounting, with all the implications that imprecision carries for the decisions built upon them.