Spend-Based vs Activity-Based Scope 3: Which One Survives an Audit

Ask two companies in the same sector for their Scope 3 Category 5 emissions and you can get numbers an order of magnitude apart without either of them having done anything wrong. The gap is usually not a data error. It is that one of them multiplied waste spend by an economic emission factor and the other multiplied weighed tonnage by treatment-specific factors, and those two operations answer different questions.

This post is about which method the GHG Protocol actually sanctions for waste, when an estimate is defensible, what you are obliged to disclose about your choice, and what an assurance provider will ask to see.

What is the difference between spend-based and activity-based accounting?

A spend-based calculation takes the money you spent on a purchased good or service and multiplies it by an environmentally-extended input-output factor expressed in kilograms of CO₂e per unit of currency. The GHG Protocol defines it in its technical guidance for Category 1, purchased goods and services: the activity data is the "amount spent on purchased goods or services, by product type, using market values," and the factor is "cradle-to-gate emission factors of the purchased goods or services per unit of economic value (e.g., kg CO2e/$)." The same guidance is explicit that spend-based and average-data methods "use secondary data (i.e. industry average data)," and that spend-based is what you apply "if the supplier-specific method, hybrid method, and average-data method are not feasible (e.g., due to data limitations)."

An activity-based calculation takes a physical quantity — tonnes of a specific material, sent to a specific treatment operation — and multiplies it by a factor specific to that material and that operation. The unit of the activity data is mass, not money.

The consequence is structural rather than a matter of precision. A spend-derived figure moves when your rates move. A tipping fee increase raises your reported emissions. A successful procurement renegotiation lowers them. Shifting ten tonnes of food waste from landfill to anaerobic digestion at unchanged cost changes nothing at all. An activity-based figure moves when material flows move, which is the only kind of movement a diversion programme can produce.

Does the GHG Protocol allow a spend-based method for waste?

Not as a named method for Category 5. This surprises people, and it is worth being precise about it.

The technical guidance for Category 5, waste generated in operations, lists exactly three methods: the supplier-specific method, "which involves collecting waste-specific scope 1 and scope 2 emissions data directly from waste treatment companies"; the waste-type-specific method, "which involves using emission factors for specific waste types and waste treatment methods"; and the average-data method, "which involves estimating emissions based on total waste going to each disposal method (e.g., landfill) and average emission factors for each disposal method." All three take a physical quantity as their activity data. None of them takes a dollar figure.

The guidance also supplies a decision tree. If waste contributes significantly to Scope 3 or engagement with treatment providers is otherwise relevant to business goals, and the treatment company can provide waste-specific Scope 1 and 2 data, use the supplier-specific method. If it cannot, but the reporting company "can differentiate its waste streams," use the waste-type-specific method. Only if the company cannot differentiate its streams does it fall to the average-data method — and note that the gating question is about your own data, not your hauler’s.

What does a waste-type-specific calculation actually look like?

The guidance gives a worked example, and its own arithmetic is more instructive than any vendor illustration. In the example, a plastics manufacturer collects tonnage by waste type and treatment, sources a factor for each combination, and sums the products. The source labels the activity data and factors as being "for illustrative purposes only," and we reproduce them here on the same basis.

Two things fall out of that table. First, the same material carries a factor that differs by a factor of twenty depending on treatment, which is why treatment attribution is not a rounding detail. Second, every row requires a tonnage and a destination. There is no arithmetic path from an invoice total to this table.

The magnitudes are larger still where avoided emissions from material recovery are in play. EPA’s Waste Reduction Model documentation works through recycling one short ton of aluminium cans instead of landfilling them as (1 × −9.13 MTCO2E/short ton) − (1 × 0.02 MTCO2E/short ton) = −9.15 MTCO2E. A method that cannot see the difference between an aluminium stream and a mixed stream cannot see that result either.

When is a spend-based or average-data estimate actually defensible?

Genuinely often, and it is worth saying so plainly rather than pretending otherwise.

An estimate is defensible when it is proportionate, disclosed, and bounded. A first-year inventory being built to identify which Scope 3 categories are material at all does not need weighed tonnage from every site; the screening exercise exists precisely to tell you where to invest in primary data. A newly acquired portfolio whose hauler contracts you have not yet seen will be estimated, and saying so is better than delaying the inventory a year. A long tail of small sites contributing a fraction of a percent of total tonnage may reasonably stay on an average-data basis indefinitely.

What is not defensible is presenting uniform confidence across sources of wildly different quality. The Corporate Value Chain (Scope 3) Accounting and Reporting Standard supplies five data quality indicators for exactly this purpose, and they are a better self-audit than any maturity model.

What does the GHG Protocol require you to disclose about your method?

More than most published inventories actually include. The Scope 3 Standard’s reporting chapter states that companies shall publicly report, for each Scope 3 category, "a description of the types and sources of data, including activity data, emission factors and GWP values, used to calculate emissions, and a description of the data quality of reported emissions data"; "a description of the methodologies, allocation methods, and assumptions used to calculate scope 3 emissions"; and "the percentage of emissions calculated using data obtained from suppliers or other value chain partners."

That last requirement is the quiet one. It is a single number that discloses how much of your inventory rests on primary data, and it is not optional. A Category 5 figure derived entirely from an EEIO factor applied to procurement spend reports zero percent against it. Nobody has to interrogate your methodology if you have disclosed that correctly — which is the argument for disclosing it correctly.

What do regulators and auditors ask for?

California’s programme is the clearest currently-legislated example because it names the standard. Health and Safety Code section 38532, added by SB 253, requires a covered entity to "measure and report its emissions of greenhouse gases in conformance with the Greenhouse Gas Protocol standards and guidance." The statutory text also sets an escalating assurance path, and CARB’s programme page confirms the population: entities "with total annual revenues in excess of one billion dollars ($1,000,000,000) that do business in California," disclosing Scope 1, 2 and 3.

The federal picture is less settled. The SEC adopted climate-related disclosure rules in March 2024, published in the Federal Register; those rules have been the subject of litigation and subsequent Commission action, so confirm their current applicability to you with counsel rather than with a vendor. What has not changed either way is what an assurance provider does when it arrives: pick a figure, ask for the calculation, ask for the factor and its source, and ask for the document the activity data came from.

What does it take to move from spend-based to activity-based?

The carbon mathematics is arithmetic. The work is data engineering, and it has three parts.

Getting tonnage, which means either extracting weights from hauler invoices where they exist, obtaining hauler-reported tonnage where the vendor will provide it, or weighing at the point of generation where neither is available. Getting stream attribution, which means a per-site taxonomy applied at the bin rather than reconstructed later from a service description. And getting destination attribution, which means the receiving facility and the operation it performed — the piece haulers are least likely to volunteer and auditors are most likely to probe.

Then all of it has to be normalized: billing periods reconciled across haulers, units converted to a common basis, duplicate invoices stripped, facility IDs mapped onto your own site hierarchy. This is the part that quietly consumes a sustainability analyst’s year and the reason so many inventories fall back to a spend proxy in the last two weeks before a filing deadline.

Dyrt built Impact Engine for that pipeline specifically — ingesting invoices, utility bills, facility scale data, RFID readings and field measurements via APIs, EDI, email parsers and PDF OCR, normalizing them, and calculating against actual material flows using the factors the target framework requires. Where the missing input is a weight at the point of generation, a DWIT kiosk and industrial scale produces the weighed, timestamped, stream-tagged record; we describe that hardware in Inside DWIT. The reason the ingest and normalization layer is shared with Spend Intelligence is that the invoice telling finance what a haul cost is the same document telling sustainability what moved.

So which method should we use?

Use the most specific method your data supports, per site, and disclose the mix rather than the average. A portfolio where forty percent of tonnage is weighed and stream-tagged, forty percent comes from cooperative hauler reporting, and twenty percent is estimated is a perfectly respectable inventory if it says so. The same portfolio reported as a single confident figure is not, because the reader has no way to tell which forty percent to trust.

And be honest about the direction of travel. The estimated share should shrink every year on a deliberate plan, starting with the sites carrying the most tonnage. An auditor is untroubled by an estimate that is labelled, justified and bounded. What troubles them is an inventory whose provenance nobody tracked — because if nobody tracked it, there is no basis for believing any part of it. For the full walkthrough of the measured version, see our guide to Scope 3 Category 5 waste reporting.