Scope 3 data challenges: why it's so hard to get right

Why is Scope 3 data so hard to get right?

Because almost none of it is yours. Scope 1 and Scope 2 come from meters, invoices and contracts inside your own systems. Scope 3 covers the emissions of a value chain, so the underlying figures sit with the organisations you buy from, invest in and sell to, each measuring on its own timetable to its own standard, or not measuring at all.

The GHG Protocol Scope 3 Standard defines fifteen categories, from purchased goods and services through to investments. A team is not solving one data problem fifteen times either, because the categories behave differently: Category 1 depends on supplier disclosure, Category 11 depends on assumptions about how a product gets used for years after it is sold, and Category 15 depends on what portfolio companies publish.

The harder version of the problem is not producing a number once. It is producing one that can be repeated next year, compared with last year, and explained to an auditor in between. A one-off calculation is an afternoon with a spreadsheet. A system that gives the same answer twice is the actual project.

Why is the data spread across so many systems?

Because a value chain is hundreds or thousands of separate organisations, and each one holds its own emissions data in its own format. There is no central register to query, so a Scope 3 inventory is assembled from whatever each of them has, in whatever state it is in.

The spread of capability across a single vendor or portfolio list is wide. Some organisations have an assured inventory, a science-based target and a published report. Some have a partial figure for their own operations. Many have never measured anything, particularly private companies, which have no disclosure obligation and therefore no reason to have started.

Internally the data is scattered too. Spend sits in the ERP, contracts sit with procurement, logistics volumes sit with a freight forwarder, and emissions factors sit in a spreadsheet somebody maintains by hand. Assembling a category total means joining datasets that were never designed to be joined, which is where most of the effort actually goes.

Why don't suppliers answer emissions surveys?

Because your questionnaire is one of many, and it arrives without the thing that would make answering worthwhile. Understanding the incentives on the other side explains most of the shortfall.

• The same request arrives repeatedly. A supplier of any size is asked by several customers, and increasingly by investors too, each with a different form, a different boundary and a different deadline.
• Nothing is reusable. An answer given to one customer cannot be handed to the next, so the work repeats every time rather than compounding.
• It reaches the wrong desk. Emissions requests land with account managers or sales contacts who do not hold the data and have no route to whoever does.
• There is no reward for replying. A supplier that measures, discloses and improves is usually treated identically to one that ignored the request, so the effort buys nothing.
• Smaller suppliers lack the means. For a company without a sustainability function, the honest answer is that it cannot produce the figure, so it does not reply at all.

The share of a list that replies is what decides how much of an inventory stays modelled. That is why chasing harder rarely fixes it: the constraint is on the other side of the request, so the fix is making the request easier to answer or removing the need for it. A prepopulated request, where the figures are already filled in and the organisation only has to check them, gets a higher response rate than a cold survey.

Why is supplier emissions data inconsistent when it does arrive?

Because each organisation made its own defensible choices, and those choices do not line up across a list. Data arriving in five different shapes is normal rather than a sign anyone did it wrong.

• Boundaries differ. Two organisations can consolidate differently, include or exclude the same subsidiaries, and both comply with their standard.
• Scope 2 comes in two versions. Market-based and location-based figures are not comparable, and a disclosure quoting only one of them is quoting the more flattering one.
• Reporting years differ. Financial years, calendar years and disclosure cycles rarely align with yours, so a current inventory carries prior-year inputs.
• Units and intensities vary. Absolute tonnes, revenue intensity and unit intensity are all reported as "emissions", and converting between them requires figures the disclosure may not include.
• Verification status varies. Some figures are assured against a recognised standard, some are internally reviewed, and some are estimates. Only the first carries weight with an auditor.

Validation and normalisation on arrival are what make the figures aggregable: mapping to a common boundary, recording the source and period, and flagging what was assumed. Skipping that step produces a total nobody can explain six months later.

What is wrong with relying on proxy data?

Nothing, until you try to act on it. Proxy and spend-based figures give complete coverage of a footprint from data you already hold, which is the right way to produce a first full inventory. The problem is what happens when a programme is still running on them three years later.

Proxy factors describe a sector, not a counterparty. Every organisation in a category shares the same figure, so the one running on renewable electricity and the one running on coal look identical per pound spent. Nothing in the footprint distinguishes them, and nothing rewards the better one.

They also do not respond to improvement. When a supplier cuts its emissions, or when volume moves to a cleaner alternative at the same price, a spend-based total holds still. Reported emissions and real progress drift apart, and the drift widens every year.

And they respond to things that are not emissions. Price rises, inflation and currency movements all push a spend-based total up with no change in physical activity, while a negotiated discount pushes it down. A year-on-year movement has to be unpicked before it means anything.

The full comparison of the two approaches, including the four calculation methods the GHG Protocol allows for Category 1, is in the guide to primary vs spend-based emissions data.

Why is it hard to match emissions data to what you actually bought?

Because the record of what you bought and the record of who emitted are keyed differently, and joining them is an identity problem before it is a carbon one.

A vendor list holds names as somebody typed them: abbreviations, trading names, regional entities, duplicates and misspellings. An emissions disclosure is published by a legal entity, usually a parent. Connecting the two means resolving each line to a real organisation, which is why evaluation conversations turn to identifiers early: DUNS, LEI, ISIN, VAT and tax IDs, and whether the match is to the subsidiary you pay or the group that reports.

Parent and subsidiary is where the error gets large. A group inventory covers operations you never bought from, and applying it to a purchase from one small subsidiary can be wrong by an order of magnitude in either direction.

Category mapping is the second half. One supplier can span several spend categories, so its emissions have to be allocated across them by spend, by volume or by assumption. Those assumptions are what stop a Scope 3 total from supporting product-level or category-level analysis, unless they are recorded well enough to be revisited.

Why doesn't the GHG Protocol solve the consistency problem?

Because it sets requirements rather than a single prescribed calculation. The GHG Protocol Scope 3 Standard defines the fifteen categories and what has to be disclosed, and for Category 1 it allows four calculation methods: supplier-specific, hybrid, average-data and spend-based. Two companies in the same sector can follow it properly and produce very different numbers.

ISO 14064-1:2018 pushes in the same direction. It requires an organisation to identify and quantify its significant indirect emissions, explain the criteria used to judge significance, document the quantification methodology and the reason it was chosen, and address uncertainty. That makes a choice traceable, not uniform.

The result is comparability inside one inventory and weak comparability between two. A supplier scorecard built from disclosures is comparing methods as much as performance, unless the figures were normalised first.

Reporting frameworks are narrowing this. CSRD and ESRS E1 ask for methods, assumptions and data quality to be disclosed alongside the numbers, and CDP asks respondents to state which method produced each category. Disclosure of the method is becoming the mechanism, rather than agreement on one method.

Why does Scope 3 work run out of internal resource?

Because most of it is manual, and the manual part scales with the length of the vendor or portfolio list. A team that can handle Scope 1 and Scope 2 in a week can lose months to Category 1 alone.

The work is spread across functions that do not report to each other. Sustainability owns the number, procurement owns the supplier relationships, finance owns the spend data, and IT owns access to the systems holding it. Every refresh needs all four, and none of them has the cycle in their objectives.

Then there are the spreadsheets. Factor tables maintained by hand, mapping files with a single owner, and reconciliation done by eye. It works once. It survives one handover badly, and it makes an auditor's request for a source trail into an archaeology exercise.

The cost of that is opportunity rather than hours. A team spending its cycle assembling the number has nothing left for the part that reduces emissions, which is talking to the organisations behind the largest lines.

Why do the numbers move every year?

Because three things change underneath a Scope 3 inventory between one cycle and the next, and only one of them is emissions.

The counterparty list changes. Suppliers are added and dropped, contracts are renegotiated, portfolios turn over, and acquisitions bring whole populations of new organisations with them. Coverage built last year does not cover this year's list.

The inputs change. Emissions factor sets are revised and reissued, disclosures are restated, and a supplier that reported an estimate last year may report an assured figure this year. Each improvement moves the total.

The methods change. Replacing a spend-based line with a reported figure is progress that shows up as a change in reported emissions, which is why a movement has to be attributed before it is announced.

This is what a recalculation policy is for: a written rule for when the base year is restated, what threshold triggers it, and how method changes are separated from real reductions in the year-on-year commentary. Without one, the honest answer to "did emissions fall" is that nobody can tell.

What does progress on Scope 3 data look like?

t looks like a sequence, and the order matters more than the pace. Teams that get somewhere tend to work through the same four steps.

1. Cover everything roughly. A modelled figure across the full list, so the total is complete and the largest lines are visible. Nothing can be prioritised before this exists.
2. Improve where it counts. Replace modelled figures with reported ones, starting at the top of the ranking. Chasing accuracy on immaterial lines is the most common way to spend a year and move nothing.
3. Engage the organisations behind the largest lines. Reported data and reduction come from the same conversations, so this step does two jobs.
4. Track and restate deliberately. A recalculation policy, a documented method per line, and commentary that separates method changes from reductions.

Two things follow from the sequence. Coverage precedes accuracy, because you cannot target what you cannot see. And the inventory becomes useful before it becomes precise, which is what makes the first cycle worth running rather than waiting for better inputs.

What does it take to get numbers you can defend?

A source for every figure, a documented method behind it, and a record of what changed since last time. That is a data problem with a supply answer: the figures either exist somewhere already or nobody has them, and finding the ones that exist is work that does not have to be done in-house.

DitchCarbon provides verified emissions data for over 2 million organisations, so procurement, sustainability and finance teams can measure and act on supply chain and portfolio emissions from one source. Four things about it speak directly to the problems above:

• Coverage before requests. The data exists before anyone is asked for it, so a complete first inventory does not depend on response rates.
• Matching that holds up. Entity resolution against DUNS, LEI and ISIN identifiers connects a vendor line or a holding to the organisation that reports, including private companies.
• Provenance on every figure. Every figure carries its source and change history, with coverage gaps shown rather than hidden, and continuous refresh with documented sources instead of an annual snapshot.
• Verification of the calculation itself. The DitchCarbon Portal calculator is verified to ISO 14064-3, limited assurance, by UL Solutions, renewed annually, and the emission factor methodology was independently assessed by Globus Thenken in August 2025. Both reports are downloadable from the trust centre.

The end state is not a perfect dataset, because a value chain does not hold still long enough for one. It is numbers you can defend within 2 weeks, with the method and the source attached, and a coverage figure that improves every cycle rather than resetting.