Scope 3 emissions definition: what counts, what does not, and where people get it wrong

The Scope 3 emissions definition, straight from the GHG Protocol: all indirect greenhouse gas emissions that occur in the reporting company's value chain, excluding the indirect emissions from purchased energy that are already counted in Scope 2. One sentence. Everything a company causes but does not directly emit or buy as energy: the goods it purchases, the freight it pays for, the flights its people take, the products it sells being used and thrown away.

The sentence is easy. Applying it to ten thousand accounts-payable lines is the actual job, and it is where first inventories stall. This post unpacks the definition into its working parts: the boundary logic, the 15 categories, what is in, what is out, and the classification mistakes that show up in real ledgers over and over.

The logic underneath the definition

The three scopes exist to make company inventories add up across an economy without double counting inside any one report. Scope 1 is direct: your combustion, your leaks. Scope 2 is purchased energy, split out because it is the one indirect source you control by contract. Scope 3 is everything else, and the key design feature is that it is deliberately someone else's Scope 1 and 2. Your supplier's factory gas is their Scope 1 and your category 1. Your carrier's diesel is their Scope 1 and your category 4. When your customer reports the electricity your product consumes, that is their Scope 2 or category 11 view of the same tonnes you report in category 11 yourself.

This is why "Scope 3 double counts" is a misreading. Within one company's inventory nothing is counted twice. Between companies, overlap is intentional: it puts the same tonnes in front of every party that can influence them. The buyer can specify recycled aluminum; the smelter can switch power contracts. Both see the tonnes; either can act.

The 15 categories, compressed

The Protocol's Corporate Value Chain Standard divides Scope 3 into eight upstream and seven downstream categories. Upstream: purchased goods and services (1), capital goods (2), fuel- and energy-related activities not in Scope 1 or 2 (3), upstream transportation and distribution (4), waste generated in operations (5), business travel (6), employee commuting (7), and upstream leased assets (8). Downstream: downstream transportation (9), processing of sold products (10), use of sold products (11), end-of-life treatment of sold products (12), downstream leased assets (13), franchises (14), and investments (15). Full definitions with examples live on our scope 3 categories reference; the compressed version here is enough to follow the boundary cases below.

What is in: the tests that decide

Three questions classify nearly any line into or out of Scope 3:

• Did money or goods flow because of your business? If you paid for it, bought it, leased it, or sold it, it is in your value chain. The AP export is therefore a nearly complete upstream census, which is why spend-based screening works as a first pass.
• Is it already Scope 1 or 2? Fuel you burn and electricity you buy are out of Scope 3, with one subtlety: the upstream production of that fuel and electricity (extraction, refining, transmission losses) is category 3. Your diesel has a Scope 1 combustion component and a category 3 well-to-tank component, and mature inventories report both.
• Who controls the asset? Control decides between Scope 1 and category 4 or 8: your truck versus your carrier's truck, your meter versus your landlord's meter. Our Scope 1 and 2 guide treats these boundary cases in detail.

What is out

Some things genuinely sit outside the definition, and knowing them saves debate. Emissions with no transactional or operational link to you: your employee's holiday flight, the coffee shop next door. Carbon credits you purchase are not negative Scope 3; offsets are reported separately from the inventory, never netted against it. Sold products' emissions do not disappear when you add a disclaimer; if your product consumes energy in use, category 11 is yours whether or not you like the number. And your own Scope 1 and 2, obviously, are not re-counted in Scope 3, however tempting a "total impact" number looks in a deck.

The classification calls that go wrong most often

• Everything dumped in category 1. Purchased goods and services is the default bucket, and lazy classification makes it 95% of Scope 3, which destroys its usefulness. Capital purchases belong in 2, freight in 4, waste contracts in 5, travel agencies in 6. The point of categories is to route data collection; a ledger where category 1 is everything routes nothing.
• Freight misrouted. Inbound freight you pay for: category 4. Outbound freight you pay for: also category 4, despite the direction. Category 9 is only transport you do not pay for, like a customer-arranged pickup. Who pays is the test, not which way the truck drives.
• Commuting versus travel. The same employee, the same car: to the office is category 7, to a client is category 6. Transit benefits are 7; the sales team's flights are 6. Mixing them muddies both the number and the reduction lever, because you influence travel policy and commuting policy differently.
• The leased office counted twice or not at all. If the landlord holds the meter, the office energy is category 8. Teams either forget it entirely (no utility invoice arrives, so no line exists) or count it in Scope 2 anyway "because it is our office". Category 8, once, from the service-charge data or an intensity estimate.
• Category 3 skipped. Fuel- and energy-related activities is the least intuitive category and the most systematically omitted. It follows mechanically from Scope 1 and 2 quantities, so once meters are collected it is nearly free to compute, and auditors increasingly look for it.
• SaaS and cloud forgotten. Software companies routinely classify their cloud bill as "not physical" and drop it. Data centers burn real electricity; the cloud invoice is category 1 with a respectable factor, and for a SaaS business it is often the biggest single line.

Measurement: the definition does not require perfection

Nothing in the Scope 3 definition demands primary data on day one. The standard explicitly permits estimation, and the accepted maturity path is: spend-based screening across everything, activity-based measurement where the tonnes concentrate, supplier-specific data where influence lives. What the definition does demand is honesty about which method sits behind which figure, category by category, and a boundary statement for anything excluded. A screening estimate labeled as a screening estimate is compliant methodology; the same estimate presented as measured fact is not. That labeling discipline is exactly what our carbon accounting software enforces per line, and you can see it in the confidence and basis columns of the live demo.

Why the definition is suddenly enforceable

For two decades Scope 3 was a voluntary framework with voluntary vagueness. The enforcement layer arrived recently: CSRD's ESRS E1 requires material Scope 3 categories with stated methods for companies in scope, ISSB's IFRS S2 carries Scope 3 into jurisdictions adopting it, California's SB 253 phases it in for large companies, and, faster than any regulator, big customers now push supplier questionnaires down the chain because your emissions are their category 1. The requirements and timelines are summarised on scope 3 emissions reporting.

Minimum boundaries: what the standard actually requires per category

The Corporate Value Chain Standard does more than list the 15 categories. For each one it defines a minimum boundary: the set of activities that must be included for the category to count as complete, plus optional elements a company may add if it wants a fuller picture. This is the part of the definition most first-time readers skip, and it is the part an assurance provider reads first, because it converts "we report business travel" from a claim into a checkable statement.

Three examples show how the mechanism works:

• Category 1, purchased goods and services. The minimum boundary is cradle-to-gate: all upstream emissions of the things you buy, from raw material extraction through the supplier's factory gate. Not just the supplier's own Scope 1 and 2, but their upstream chain as well. A category 1 figure built only from tier-one supplier disclosures is below the minimum boundary and should say so.
• Category 6, business travel. The minimum boundary is the Scope 1 and 2 emissions of the carriers transporting your employees: the airline's fuel, the rail operator's power. Hotel stays are an optional element. A company may include them, and many do, but leaving them out is not an omission against the standard, while leaving out flights is.
• Category 11, use of sold products. The minimum boundary is the direct use-phase emissions of products sold in the reporting year, over their expected lifetime: fuel burned by engines you sold, electricity drawn by appliances you sold. Indirect use-phase emissions, such as the energy used to wash a garment, are optional. That distinction decides whether an apparel company owes a category 11 number at all.

The practical consequence: when you compare two companies' category figures, or your own figures year over year, the minimum boundary is what makes the comparison meaningful. A category label without its boundary is just a heading. The per-category definitions, with these boundaries alongside, are on our scope 3 categories reference.

Materiality and exclusions done properly

The definition says all indirect value-chain emissions, and the completeness principle in the standard backs it up. Yet no real inventory covers everything at equal depth, and the standard knows it. What it permits is documented exclusion: a category or activity may be left out if the exclusion is disclosed and justified. What it does not permit is silence, and the difference between the two is the difference between a methodology and a mistake.

Defensible justifications come in a few shapes:

• Not applicable. A company with no franchises has nothing to report in category 14. Stating that costs one sentence and closes the question permanently.
• Immaterial, with evidence. "Category 13 screened at under 1% of total Scope 3 using a spend-based estimate" is a legitimate exclusion. The screening estimate is the evidence: you cannot claim a category is small without having sized it, however roughly.
• Data not yet obtainable. Acceptable in early inventory years if paired with a plan to close the gap, and increasingly challenged when it reappears unchanged year after year.

The asymmetry to internalise: a documented exclusion is a boundary decision that a reviewer can accept or contest, while a silent omission is indistinguishable, from the outside, from an error. Under CSRD this reasoning folds into the materiality assessment, which reviewers expect to see in writing, not reconstructed from memory when the question arrives. The working rule is short: exclude with a sentence, never with silence. Common boundary questions of this kind are collected on our FAQ.

How the definition interacts with targets

Boundary decisions look like bookkeeping until you set a reduction target, at which point they decide what you have committed to. The Science Based Targets initiative applies thresholds directly to the inventory: if Scope 3 is 40% or more of total emissions, which is the case for most companies outside heavy industry, a Scope 3 target is required rather than optional. Near-term targets are expected to cover roughly two thirds of Scope 3 emissions, and net-zero targets raise that coverage expectation to around 90%.

Run the logic backwards and the definition's weight becomes obvious. Whether category 11 sits in your inventory can determine whether Scope 3 crosses the 40% line at all. Which categories you include determines which two thirds you are able to select for a near-term target, and a target built on an understated inventory is a commitment to reduce the wrong number. It is common for a first complete screening to reveal that the categories a company had ignored are exactly the ones any credible target would have to cover.

The sequencing that follows is worth stating plainly: settle the boundary before setting the target, not after. A complete screening, even a coarse one, tells you what a target must cover; a target set before the screening becomes an argument for keeping the inventory incomplete. How targets, offsets and residual emissions then fit together in the accounts is the subject of our post on net zero accounting.

The practical takeaway

Treat the definition as a routing rule, not a philosophy question. Every line in your AP export either is already Scope 1 or 2, or it lands in one of 15 categories, or it fails the value-chain test and is out. Run that routing over a complete export and you have a defensible first Scope 3 screening: complete by construction, coarse by admission, and improvable by design. That routing pass is what the Scope Classifier demonstrates on sample or pasted data, in about a minute, with the basis and confidence of every call stated on the row.