Scope 3 is most of the footprint, the part the board understands least, and the part the assurer is least likely to sign. The failures are rarely in the arithmetic. They are in the boundary, the consistency and the governance around the number, and they are governance problems, not technology problems.
Every board that has signed a sustainability statement has signed a Scope 3 number, and in most organisations it is the largest number on the page. Across 100 of the largest listed companies in the United States, Scope 3 averaged 67 per cent of total reported emissions (The CPA Journal, December 2024, based on 2021 reports). It is also the number with the shortest audit trail, the most assumptions, and the fewest people who could explain, line by line, where it came from.
Scope 1 and Scope 2 are built from meters and invoices the company controls. Scope 3 is built from other people's data: suppliers, employees, customers, and, where none of that is available, from money spent multiplied by an industry average. The GHG Protocol Corporate Value Chain (Scope 3) Standard, published in 2011 and still the operative text, sets out fifteen categories (GHG Protocol, Scope 3 Standard, chapter 5, Table 5.3), and its companion Calculation Guidance allows more than one method for each. That flexibility is where inventories go wrong, because flexibility that is not written down becomes inconsistency the following year. What follows is drawn from building and rebuilding corporate inventories across multi-site businesses in several countries.
The Standard requires a company to account for all Scope 3 emissions and to disclose and justify any exclusions (GHG Protocol, Scope 3 Standard, chapter 6). In practice, the boundary of a spend-based purchased-goods calculation is a list of ledger accounts that lives in a spreadsheet filter rather than a policy. The following year someone rebuilds the filter, accounts are added and dropped, the total moves by ten per cent, and the report calls it a change in emissions. It is a change in what was counted. The Standard's consistency principle says such changes must be documented and may require the base year to be recalculated (chapter 4). Without a written inclusion list, reviewed and signed off each year, there is no way to tell growth from drift, and no way to defend either to an assurer.
Electricity appears twice in most companies' data: as kilowatt-hours on a utility bill, which belong in Scope 2, and as a line of spend in the ledger, which belongs nowhere, because counting it under purchased goods counts the same energy a second time at a different price. Gas, water, waste and travel can all be double counted the same way. The Standard designs the scopes and categories to be mutually exclusive (chapter 5, sections 5.1 and 5.4); it is the company's own table of emission sources that makes them so, and most inventories do not have one.
The mirror-image error comes from the finance system, which organises itself around legal entities and unit codes rather than sites. One location can sit under two codes; aggregate at the wrong level and it is counted twice or missed, and in a portfolio of forty locations both errors hide inside a plausible total. A maintained mapping from every finance code to a named site is unglamorous and is the most common reason a site-level figure cannot be reconciled to the group total.
A Scope 3 figure is activity data multiplied by an emission factor, and the factor is usually the larger source of uncertainty. Several national factor sets, the UK's DESNZ conversion factors among them, are reissued each year, and methodologies change between editions. An inventory that records only the value of each factor, and not its source, year and version, cannot be reproduced, and a figure that cannot be reproduced cannot be assured. The GHG Protocol's Standard Development Plan of 29 July 2026 lists, among the topics its Scope 3 working group will consider, a data quality scoring matrix, a requirement that data quality improve over time, and restricting or phasing out the spend-based method (Consolidated Standard Development Plan, Appendix C, pages 24 to 25). A company with a versioned factor register will be able to show improvement; one that cannot say which factor it used starts from zero.
Data arrives incomplete: a supplier does not respond, a landlord withholds a utility figure, a ledger extract is missing a quarter. The tempting response is to estimate, because an estimate produces a complete report on time. The Standard requires the methods, assumptions and data sources for each category to be described (chapter 11), and an estimate that is not labelled becomes a fact in next year's comparison. Gaps are also information: a site with no capital expenditure in a refurbishment year, or rent recorded for three of sixteen locations, is telling you something about the finance data. Filling the gap silences the signal; flagging it, querying it and recording the answer builds the audit trail the number will need.
Companies that set targets must choose a base year, adopt a recalculation policy, and recalculate when significant structural or methodology changes occur (chapter 9). Most adopt the policy. Fewer apply it. Sites are acquired and sold, a method changes from spend to activity, a factor set is replaced, and the base year stays where it was, so the reported reduction measures the gap between two different inventories. One 2026 climate accounting statement from a large listed company shows what recalculation looks like when it is done properly: a prior-year Scope 3 figure restated for an updated baseline and revised packaging data, with the policy set out in a note. Restatement of that kind is normal. A base year silently left alone while the organisation changed is what an assurer or a target-validation body will not accept.
Behind the base year sits the deeper problem: no yearly cycle. Scope 1 and 2 data flows through finance every month by accident. Scope 3 data comes from procurement, human resources, facilities, travel and suppliers, none of whom owns the inventory, so it is rebuilt from scratch each year by whoever is available, and each year's judgements leave with them. Every comparison with last year then rests on a method, boundary and factors that were the same, or on every difference having been recorded.
The clearest external test of a Scope 3 inventory is whether an independent assurer will put a name to it. For most companies the answer is partly or not at all. Across 1,400 large companies in 22 jurisdictions, 99 per cent of those reporting Scope 1 and Scope 2 obtained assurance on them in 2023; for Scope 3 the figure was 76 per cent, and most assurance was limited rather than reasonable (IFAC, AICPA and CIMA, The State of Play: Sustainability Disclosure and Assurance, May 2025, pages 5 and 20). Among 100 S&P 100 companies, where Scope 3 was assured at all, 93 per cent of engagements were limited and on average only 73 per cent of the disclosed categories were covered (The CPA Journal, December 2024). The United States Securities and Exchange Commission's stated reason for not requiring Scope 3 assurance was that it is harder to verify than Scopes 1 and 2 (same source).
One 2026 climate accounting statement from a large listed company shows the pattern from the inside: an assurance report covering Scope 1 and 2 for the current year only, a few pages from a goal-progress table describing the Scope 3 inventories as subjected to limited assurance, and two Scope 3 figures for the same year, one physical and one market-based.
Assurers stop where the audit trail stops, and each of the five failures above is a place where it stops. The regime is tightening: limited assurance is required under the Corporate Sustainability Reporting Directive as amended by the Omnibus I Directive, in force from 18 March 2026 (Arthur Cox summary of Directive 2026/470; the directive text is authoritative), and the IAASB's ISSA 5000, the first global sustainability assurance standard, applies to periods beginning on or after 15 December 2026, replacing ISAE 3410 from the same date (IAASB).
Three things are moving at once, and each lands on the board rather than the sustainability team. The standard is being rewritten: the GHG Protocol and the International Organization for Standardization announced on 29 July 2026 that they will merge their corporate carbon accounting standards into one, with public consultation planned for the second quarter of 2027 and publication in the fourth quarter of 2028 (GHG Protocol, Standard Development Updates FAQ, July 2026). The target regime is moving: the SBTi's Corporate Net-Zero Standard Version 2.0, published on 11 June 2026, opens for validation in early 2027 and is mandatory for all new submissions from 1 February 2028 (SBTi, June 2026). Assurance is arriving: limited assurance under CSRD, ISSA 5000 from December 2026, and a market in which 99 per cent of large companies that report Scopes 1 and 2 already have them assured (IFAC, May 2025).
None of this requires a board to understand emission factors. It requires the board to treat the Scope 3 number the way it treats any other figure it signs: to know who owns it, what policy governs it, whether it is consistent with last year, and whether it would survive an assurance engagement. On the evidence above, for most companies, not in its present form.
An inventory that survives scrutiny has five properties. They are observable from outside, which means a director can ask for each of them without knowing how they are achieved.
Any published number can be followed, in a documented path, to a specific row in a specific file, with the factor applied and its version stated. If one step lives in someone's head, the property is not met.
It states what is in, what is out, why, and who approved it. A change in the boundary is recorded as such, not reported as a change in emissions.
Every factor carries its source, publication year, geography and date of adoption, and replacing one leaves a record.
A policy that has never been used is a policy nobody has tested.
Each takes weeks to put in place and a fraction of that to maintain, and each is cheaper than the restatement, the qualified opinion, or the target validation that fails for want of a defensible base year. How they are implemented is specific to the organisation and is the substance of the work I do with clients. What they look like when done is not a secret, and a board is entitled to ask for all five.
One question to ask tomorrow: ask the owner of the sustainability report to show you, for one line of the Scope 3 inventory, the path from the published number back to the source data, with the boundary decision, the factor and its version, and the sign-off along the way. If that takes an hour, the inventory is in good shape. If it cannot be done at all, you have found where it falls apart, and the fix is a governance decision, not a technology purchase.
Sourced as at 13 October 2026. Sources listed in the appendix below.
If your inventory is heading towards limited assurance, a target renewal under SBTi Version 2.0, or a first CSRD report, let's talk about where the trail stops and what it would take to close it.
Connect