Spend-Based vs Activity-Based Emission Factors: The Scope 3 Choice That Decides Your Footprint

I once watched a Scope 3 number drop by 38% overnight. No supply chain changes, no efficiency program, no fancy decarbonisation deal. Someone had simply replaced a spend-based estimate for purchased steel with the supplier's actual EPD-backed factor. Same steel. Same year. Different method. Different reality.

That's the choice this article is about. If you do any kind of value-chain footprint — Category 1, Category 2, Category 4, you name it — you're picking between two ways of measuring. One uses what you spent. The other uses what you actually consumed. They almost never agree, and the gap between them is where most Scope 3 arguments happen.

If you're new to the underlying mechanics, skim our emission factors explainer first. This piece assumes you already know what an emission factor is — we're getting into how you build one for things you didn't burn yourself.

The fork: dollars or things

Pull out your trial balance. You bought $4.2 million of "industrial chemicals" last year. How much CO₂-e is that?

You have two roads from here.

The spend road. You multiply the $4.2m by some factor like "0.42 kgCO₂e per USD of industrial chemicals". You get a number in about three minutes. The factor came from an environmentally-extended input-output model — EEIO — which is a big economic table that connects industry sectors to their emissions. You haven't touched a kilogram of anything. You're working with money.

The activity road. You email procurement and ask what you actually bought. Two tonnes of caustic soda. Three tonnes of sulphuric acid. Half a tonne of a specialty solvent. You match each one to a physical-unit factor — kg CO₂e per kg of caustic soda, etc. Maybe a week of work, maybe a month, depending on how much your AP team likes you. The number is grounded in something physical.

Both are legitimate. Both are GHG Protocol-compliant. They will give you different answers, sometimes by an order of magnitude, and which one you picked usually matters more than any single supplier conversation you have all year.

What spend-based actually is

People throw "EEIO" around like everyone knows what it means. Quick demystification.

Statistical agencies build input-output tables that show how every sector of the economy buys from every other sector. Auto manufacturing buys steel from metals. Metals buys electricity from utilities. Utilities buys coal from mining. The whole web. Now bolt environmental data onto the same web — how much CO₂-e each sector emits per dollar of output — and you've got an environmentally-extended input-output model. EEIO. That's it.

When you use a spend-based factor for "industrial chemicals," you're not estimating your specific chemicals. You're using the national average emissions intensity of the entire chemicals sector, scaled to your dollars. Every cleaning fluid, every fertiliser, every pharmaceutical precursor, every plastic feedstock — averaged together.

That's the trade. You get coverage everywhere instantly, at the cost of resolution. A Tesla and a 1998 Hilux look identical to a spend-based factor for "motor vehicles." Different planet, same number.

Spend-based factors are a map of the economy, not a map of your supply chain. They tell you what the average dollar in a sector emits — which is useful, until you remember your dollar isn't average.

USEEIO, EXIOBASE, CEDA — which one and why

If you're using spend-based methods, you're using one of these three. Each has a personality.

The interesting thing is that the same imported good — say, $1m of imported textiles into a US footprint — can come out 40-50% apart depending on which model you use. A recent EPA-affiliated comparison found CEDA import factors were on average 44% higher than US-domestic factors, EXIOBASE was 16% higher, and GLORIA was 8% lower. Same physical world. Three different stories.

That's not a flaw, it's just modelling. Different aggregation assumptions, different trade matrices, different reference years. But it tells you why year-on-year consistency matters more than picking the "correct" model. There isn't a correct one.

A second thing worth knowing: the major EEIO models got refreshed in 2024 (USEEIO v2.5, CEDA 2024, EXIOBASE v3.8.2). The median industry-average factor came down roughly 10-20% in both USEEIO and CEDA. If your spend-based footprint mysteriously dropped between the 2024 and 2025 reporting cycles and nobody can explain why, this is your suspect.

Where spend-based quietly breaks

Spend-based methods have four failure modes I see again and again. None are theoretical — all of them have been argued out in real assurance meetings.

1. You buy something unusual

The sector average is fine for ordinary things. The moment you buy something that isn't sector-typical, it collapses. A pharmaceutical company purchasing high-purity solvents will look identical to one buying bulk industrial degreasers in a spend-based view. They're orders of magnitude apart in reality.

2. Your supplier is much better (or much worse) than average

If you're sourcing aluminium from a smelter running on hydroelectric power in Iceland or Quebec, you're 5-10x cleaner than the global aluminium average. The spend-based factor doesn't know that. You'll over-report — sometimes by huge amounts. The reverse is also true: a steel supplier running on captive coal looks like an average-grid producer in spend, when they're actually well above the sector mean.

3. Prices and emissions move independently

This is the one that catches people. Suppose semiconductors decarbonise — fabs switch to renewables, etch chemistries get cleaner. Per-unit emissions fall. But chip prices spike at the same time because of demand. Spend-based says your emissions went up. Reality says they went down. Spend-based factors implicitly assume the relationship between price and emissions intensity is stable. It usually isn't.

4. You decarbonise, the number doesn't move

This is the painful one. You switch to a lower-carbon cement supplier. You keep buying the same dollar amount of cement. Spend-based factor doesn't budge. You did real work and your inventory is silent. After about year two of this, executives start asking why anyone bothers.

The inflation trap (and the deflator nobody applies)

This deserves its own section because it's the single most common error I see in spend-based inventories, and it's getting worse with every year of post-2021 inflation.

Every EEIO factor is denominated in a specific currency, in a specific price year. USEEIO v2.5 factors are in 2017 producer-price USD. EXIOBASE v3.8.2 base year is 2022. CEDA 2024 is calibrated to 2022. When you multiply 2025 spend by a 2017-priced factor, you're saying "$1 in 2025 buys the same amount of stuff as $1 in 2017." It doesn't. Not even close.

US PPI for chemicals rose roughly 30% from 2017 to 2025. If you apply a 2017-priced USEEIO factor to 2025 spend without deflating, you'll overstate emissions by about 30%, because you're paying more dollars per unit of stuff than the factor assumes. Conversely, you can understate if the sector's price moved less than general inflation.

The correct sequence:

Most carbon accounting platforms do this for you, but check. I've audited two enterprise platforms that didn't — they were applying nominal current-year spend to old-vintage factors, and customers were unknowingly over-reporting by 20-30%. Ask your vendor for their deflator methodology in writing.

When activity-based earns its keep

The pitch for activity-based is simple: you measure what was actually consumed, in physical units, and multiply by an emissions intensity for that physical thing. Tonnes of cement × kgCO₂e per tonne. Kilometres flown × kgCO₂e per passenger-km. It's the same logic as Scope 1 and 2, just pointed at someone else's combustion.

The upside is obvious. You catch real change. If you switch to lower-carbon cement, the number drops. If your freight forwarder swaps a diesel fleet for biodiesel, the number drops. The inventory rewards actual decarbonisation, which is the whole point.

The downside is also obvious — it's a lot more work, and for some categories the data simply isn't there. Try getting physical-unit quantities for "office consumables" out of an ERP. You'll find yourself negotiating with your AP team about whether 47 line items of "supplies — misc" can be broken out, and the honest answer is no.

Where activity-based wins decisively:

• Anything you can weigh or count. Concrete, steel, aluminium, plastic, paper, packaging. Procurement usually knows the tonnage even when finance only knows the dollars.
• Energy products. Fuel, electricity, gas. The activity unit is unambiguous and factors are everywhere.
• Freight. Tonne-km is the gold standard. Most logistics providers can give it to you for the asking.
• Travel. Distance, mode, class. Travel platforms hand you this directly.
• Use-of-sold-products (Category 11). If you sell anything that consumes energy in use, you need activity-based — there's no spend equivalent that works.

Where it's a fight and rarely worth it:

• Professional services (legal, consulting, marketing). What's the physical unit? Hours? Pages? Don't bother — leave it on spend.
• Software-as-a-service. There's no satisfying answer to "how many kilograms is one Salesforce seat?" Use spend, or use vendor-disclosed emissions per seat where it exists.
• Real estate leases and financial services. Spend or specialised methodologies (PCAF for financial) are the realistic options.

The hybrid method, in plain English

Nobody serious does pure spend-based or pure activity-based. The hybrid method — which the GHG Protocol has effectively endorsed for years and the 2026 revisions reinforce — is the only sensible answer in practice.

The recipe is straightforward:

1. Do a spend-based screen first. Get the whole footprint covered in EEIO factors. This is your hotspot map.
2. Find the top categories that account for ~80% of your spend-based estimate. Usually three to seven line items.
3. Move those to activity-based or supplier-specific data. This is where the work goes.
4. Leave the long tail on spend. There's no value in chasing a $4,000 line item for "office supplies" when your steel category is 200x bigger.

The Pareto thing genuinely holds. In every Scope 3 inventory I've worked on, five to ten categories drive the answer. The rest is noise that you still have to cover, and EEIO is exactly the tool for that.

How this maps to the 2026 GHG Protocol tiers

The GHG Protocol published its Scope 3 Phase 1 Progress Update on 31 March 2026 — the first concrete preview of the revised standard due in 2027. Two changes matter here.

Revision A1: Tier disclosure. You'll need to disaggregate every Scope 3 category by three data-quality tiers:

• Specific — supplier-specific or product-specific primary data.
• Non-specific (average) — physical-unit factors that are sector or process averages (think DEFRA, ecoinvent, NGA).
• EEIO / spend — the spend-based bucket.

And a fourth implicit category for "unknown" where you can't even classify. The headline number doesn't go away, but suddenly readers can see how much of it is real and how much is sector-average dollar-math. Companies that have been hiding behind spend-based aggregates will have nowhere to hide.

Revision B1: 95% coverage floor. The current Standard says you should "disclose and justify any exclusions." The revised one will require you to cover at least 95% of required Scope 3, with anything excluded justified by data. Spend-based is explicitly permitted to fill that 95% — the Standard isn't anti-spend — but it'll be visibly classified as the bottom tier.

The practical effect is that nobody loses the right to use spend-based, but the optics shift. A report that's 90% spend-based and 10% activity-based will look very different from one that's the other way around. Boards, investors, and procurement teams reading the tier disclosure will draw conclusions.

A migration plan that won't burn out your team

If you're starting from a mostly-spend-based footprint and you want to be in defensible shape before the 2027 standard lands, here's how I'd phase it. Three reporting cycles, not three months.

One warning. Don't try to move everything to activity-based in a single year because someone showed you a slide deck. I've seen teams chase that and end up with worse numbers — half a category in activity, half in spend, double-counted, with no methodology note. Slow and consistent beats fast and messy. Especially because the 2026 tier disclosure rewards consistency as much as it does upgrades.

Frequently asked questions

Is spend-based ever the "right" method?

Yes, for two specific jobs. First, screening — getting a quick look at where the emissions are so you know where to focus. Second, covering long-tail categories where activity data costs more to collect than the category contributes to your footprint. Anything beyond that, you're paying a precision tax.

What about DEFRA's spend-based factors?

DEFRA publishes its own set of UK-specific spend factors. They're useful for UK-only footprints but they're a thin layer compared to a full EEIO model — limited sectors, no multi-regional trade flow. Fine for screening; thin for a primary inventory method. See our databases compared guide for where DEFRA sits among the bigger options.

How do I document my method for assurance?

Three things assurers want, in order: the source model and version (e.g. "USEEIO v2.5, 2017 USD basis"), your inflation/currency adjustment method, and your category mapping (which GL accounts hit which EEIO sectors). If you can hand all three over on request, you're already ahead of most reporters.

Does CSRD or CBAM force me to activity-based?

Effectively, yes — for material categories. CSRD's double-materiality requirement and CBAM's product-level reporting both push toward physical-unit data. You can still use spend-based for non-material long-tail items, but if a category is strategic, expect to be asked for primary or activity data.

Can I mix methods within a single category?

You can and you should — that's the hybrid approach. The 2026 GHG Protocol tier disclosure assumes you will, which is why it asks you to break out each category by tier. Just document the split and apply the same split consistently year-on-year.

What happens when the EEIO model gets updated mid-cycle?

If your model vintage changes (e.g. USEEIO 2.4 → 2.5), restate prior years using the new model for comparability, or hold prior years on the old model and clearly flag the methodological discontinuity. The GHG Protocol's general rule is to recalculate if the change is material — typically more than 5% of total Scope 3. The 2024 EEIO refreshes triggered exactly this for many companies; if you haven't restated, your trend line probably has an artificial step in it.