Spend-Based vs Activity-Based Carbon Accounting: Which Is Right?

The spend-based versus activity-based question is the single most common methodological fork in a carbon report, and the way most guides answer it is wrong. The standard framing — "activity-based is more accurate, spend-based is cheaper, use activity when you can" — is technically true but practically useless. It treats the choice as a preference when the GHG Protocol actually specifies a hierarchy, and that hierarchy is what a third-party assurance reviewer will check against. This guide walks through how the decision should actually be made, category by category, and what the numerical difference looks like in a real report.

Short definitions before we go further. Activity-based carbon accounting starts from a physical quantity — kilowatt-hours, gallons of fuel, kilograms of refrigerant, passenger-kilometers flown — and multiplies by a per-unit emission factor. US suppliers typically use the EPA GHG Emission Factors Hub (January 2026 release for current-year reporting) for domestic Scope 1 and 2 activity factors, and DEFRA for international logistics and air travel where EPA coverage is thinner. Spend-based accounting starts from a dollar amount paid to a vendor and multiplies by a sector-average emission factor per dollar. For US-based reports in 2026, EPA USEEIO v2.x (the current US Environmentally-Extended Input-Output model, successor to the older v1.1) is the preferred reference; Exiobase is the European equivalent. Both methods are legitimate under the GHG Protocol. Neither is the universal right answer. The GHG Protocol Scope 3 Standard §7.3.2 tells you which to use for which category, and the answer depends on what data you actually have and how material the category is to your footprint.

The operational question you are actually answering is: for this specific Scope 3 category, in this specific reporting period, given the data I have and the significance of this category to my total footprint, which method is defensible to a reviewer and reasonable for me to collect? Answer that question 15 times — once per Scope 3 category — and you have your methodology.

The GHG Protocol data hierarchy (§7.3.2), in plain terms

The GHG Protocol Corporate Value Chain (Scope 3) Standard establishes a data hierarchy that every reviewer knows by reference. It runs from most-specific to least-specific, and the rule is: use the most-specific data that can reasonably be obtained for the category in question.

Tier 1 — Primary, supplier-specific data: The actual emissions generated by your specific supplier for the specific goods or services they provided to you. Example: your logistics provider sends you their carbon statement showing 0.082 kg CO₂e per tonne-kilometer on the specific routes they served you. This is the gold standard. Rarely available for most SMBs in year 1, reasonable to build toward by year 2 or 3 for top 10 vendors.
Tier 2 — Hybrid: primary activity data + secondary emission factor: You know the physical quantity (14,800 miles driven, 142,800 kWh of electricity, 8 lbs of refrigerant) but use a sector-average emission factor rather than a supplier-specific one. This is the workhorse method for Scope 1, Scope 2, and activity-dominated Scope 3 categories (business travel, employee commuting, fuel-related activities). Most defensible SMB reports sit primarily in Tier 2.
Tier 3 — Secondary, spend-based: You multiply the dollar amount paid to a vendor by a sector-average emission factor per dollar of output in that sector (e.g., EPA EEIO: architectural services = 0.36 kg CO₂e per dollar of producer output). The least specific method, but it is what allows you to produce any Scope 3 estimate when activity data is not available. Appropriate for non-material categories or first-year screening of Category 1.

The hierarchy is directional: you should use higher-tier data when it is reasonably available, and document your data choice for each category. A reviewer who sees your report using spend-based factors for a category where you had activity data in the general ledger (fleet fuel, for example) will flag it as a methodological weakness. A reviewer who sees spend-based factors for a Category 13 (downstream leased assets) disclosure where primary supplier data was not obtainable will accept it — provided you document why.

Which method wins, category by category

The 15 Scope 3 categories divide cleanly into three groups: activity-dominated (use Tier 2), spend-dominated (Tier 3 is the realistic default), and hybrid (Tier 2 for top vendors, Tier 3 for the long tail).

Activity-dominated categories (Tier 2 expected)

Category 3 — Fuel- and Energy-Related Activities: Upstream emissions from the fuel you burn in Scope 1 and the electricity you buy in Scope 2. Since you already have fuel volume and kWh from Scope 1/2 calculations, activity data is reasonable here. Use EPA Factors Hub (most recent release) upstream factors, not EEIO.
Category 4 — Upstream Transportation and Distribution: Third-party logistics you paid for. If freight bills include shipment weight and distance, activity-based (tonne-km × DEFRA factors) is preferred. If they do not, spend-based is the fallback.
Category 6 — Business Travel: Always activity-based if you have a Concur or Navan export. Flight segments and distances are in the data; apply DEFRA short/medium/long-haul factors (DEFRA remains the global industry reference for air travel factors and is accepted by US auditors for this category). Going spend-based on business travel when Concur data exists is a classic first-draft mistake.
Category 7 — Employee Commuting: Activity-based via survey. Spend-based does not apply to this category (there is no typical spend line item for commuting).
Category 11 — Use of Sold Products: Activity-based for product companies. Unit sales × lifetime energy consumption × grid factor. Often the single largest Scope 3 category for manufacturers of electrified products.

Spend-dominated categories (Tier 3 is the realistic default)

Category 2 — Capital Goods: FY additions to PP&E. Activity-based would require getting emissions data from the manufacturer of every capital asset, which is rarely practical. Spend-based with EEIO is the accepted method for almost all SMBs.
Category 5 — Waste Generated in Operations: Spend on third-party waste disposal × waste-sector EEIO factors. Activity-based would require knowing tonnage by waste stream, which most SMBs do not track. Tier 3 is defensible here.
Category 8 — Upstream Leased Assets: Typically small for office-based SMBs, and activity data (utility consumption in leased space) is often not available when the landlord bills flat rent including utilities. Spend-based or use of CBECS sector averages is acceptable.

Hybrid categories (Tier 2 for top 10 vendors, Tier 3 for tail)

Category 1 — Purchased Goods & Services: Always the largest category for services firms and the most important to get right. Year 1: spend-based with EEIO for all vendors is standard. Year 2+: build activity-based data for your top 10 vendors by spend (typically 60–80% of Category 1 emissions), keep spend-based for the long tail. This is where the data maturity progression below actually matters.
Category 9 — Downstream Transportation: If you have carrier data with weight and distance, activity-based for top carriers. Otherwise spend-based.
Categories 10, 12, 13, 14, 15: Most SMBs will screen these as not material and exclude them with a documented rationale. Exclusion is legitimate — just document it.

Worked example — the $8M firm, Category 1 calculated two ways

Back to the $8M architecture firm from the CPA Guide. FY2024 Category 1 (Purchased Goods & Services): $3.8M in non-utility vendor spend across 2,847 AP line items. Here is how the number changes depending on whether you go pure spend-based or build out activity-based data for the top 10 vendors.

Path A: Pure spend-based (Tier 3)

Each of the 2,847 line items is classified to an EPA EEIO commodity code. The largest single classification is "Architectural, engineering, and related services" (NAICS 5413), which covers $1.2M of subcontracted design work to three external architecture firms. The EEIO emission factor for that commodity is approximately 0.36 kg CO₂e per dollar of producer output. Applied across the full $3.8M spend with appropriate classification to each vendor's NAICS code, the pure spend-based Category 1 total comes out to 1,368 tCO₂e.

Path B: Hybrid — activity data for top 10 vendors, spend-based for the tail

The top 10 vendors account for $2.4M of the $3.8M spend (63%). Reach out to those 10 with a short emissions-data request. Here is what typically comes back in a real engagement:

Top 3 architecture subcontractors ($1.2M total): 2 of 3 provide their own Scope 1 + 2 intensity figure, averaging 0.14 kg CO₂e per dollar of their revenue. The third does not respond; default to EEIO. Activity-based subtotal for responders (0.14 × $780K) = 109 tCO₂e. Non-responder spend ($420K × 0.36 EEIO) = 151 tCO₂e. Subtotal: 260 tCO₂e (compared to $1.2M × 0.36 = 432 tCO₂e pure spend-based).
Top cloud/SaaS vendor ($180K): Publishes a public carbon intensity figure (common for Microsoft, AWS, Google Cloud, Adobe, Autodesk — all major SaaS vendors now do). 0.08 kg CO₂e per dollar. $180K × 0.08 = 14.4 tCO₂e (vs $180K × 0.21 EEIO IT services = 37.8 tCO₂e).
Top logistics/print vendor ($140K): Provides activity data (tonnes of print material shipped). Applied DEFRA factors. Subtotal: 28 tCO₂e (vs $140K × 0.34 EEIO printing = 47.6 tCO₂e).
Remaining 5 top vendors ($880K): No primary data available; keep spend-based. Weighted EEIO average of 0.29 kg CO₂e/$ applied. Subtotal: 255 tCO₂e.
Long tail (2,837 line items, $1.4M): Spend-based EEIO across ~80 distinct commodity codes. Subtotal: 420 tCO₂e.
Hybrid Category 1 total: 977 tCO₂e (29% lower than the pure spend-based 1,368 tCO₂e).

The 29% reduction is not noise — it is the information premium you get from using primary data. It is also what a reasonable reviewer expects to see improve year over year. A company that reports the same spend-based Category 1 number three years running with no vendor engagement is signaling that its data program is not maturing, which is a qualitative finding on its own.

When spend-based is the right answer, and when it is a liability

Spend-based is appropriate when:

First-year reporting for any category. You do not have supplier relationships, you do not have activity data, and you need to produce a defensible screening estimate. Tier 3 is the honest method.
The category is <5% of your total footprint. Spending 40 hours to move Category 5 (Waste) from spend-based to activity-based may shift 2 tCO₂e on a 1,000 tCO₂e total. Not worth it. Leave it spend-based, document the rationale, move on.
Primary data is not reasonably available. Vendor does not respond, activity data does not exist in a trackable form, or collection cost would be disproportionate to the materiality of the category.
You are screening for materiality, not yet reporting. Spend-based is perfect for the initial materiality screen that determines which Scope 3 categories you will fully disclose.

Spend-based becomes a liability when:

The category is >20% of your footprint and you made no attempt at primary data. This is the single most common assurance finding on Scope 3. If Category 1 is 70% of your footprint and you used EEIO for 100% of it in year 2, a reviewer will flag that your data program has not matured.
The sector has high intra-sector variation. EEIO gives you a sector average, but the actual variation between best-in-class and worst-in-class suppliers in that sector can be 5x. Agriculture, heavy manufacturing, data centers, and shipping are all sectors where "sector average" hides huge differences. For categories dominated by spend in these sectors, activity-based (or supplier-specific) data matters much more.
You are going for third-party limited or reasonable assurance. ISO 14064-3 reviewers will test the data hierarchy choice per category. Pure spend-based in year 2 or 3 on material categories creates a finding.
The number needs to drive an actual decision. If your Category 1 number is feeding a reduction target, a supplier engagement program, or a procurement decision, spend-based is too blunt. Sector-average cannot tell you which supplier to prioritize — you need activity-based for that.

The three-year data maturity progression

A defensible carbon program does not go from zero to full activity-based in year one. It progresses. Here is the sequence that most mature SMB programs actually follow, and what a reviewer expects to see.

Year 1 — Screening

Scope 1 and 2 fully calculated with activity data (Tier 2) using the current EPA GHG Emission Factors Hub release. All 15 Scope 3 categories screened with USEEIO v2.x spend-based factors to determine materiality. Material categories (typically Categories 1, 6, 7, and sometimes 3, 4, or 11 depending on industry) included in the disclosed Scope 3 total. Non-material categories excluded with a one-line rationale per exclusion. Year 1 Category 1 is almost always 100% spend-based — this is fine, and expected.

SB 253 nuance: California SB 253 filers have the first Scope 1 & 2 disclosure due August 10, 2026 covering FY2025 emissions. CARB has signaled at its March 2026 workshop that the 2027 cycle (covering FY2026) will begin tightening primary-data expectations for material Scope 3 categories. Year 1 reporters get meaningful flexibility; year 2 reporters should have vendor engagement already underway.

Year 2 — Top-10 supplier engagement

Scope 1 and 2 unchanged. Categories 6 (business travel) and 7 (commuting) refined with better survey methodology. Category 1 starts the hybrid transition: identify your top 10 vendors by spend (usually 60–80% of Category 1 emissions), send each a data request, accept whatever primary data they return, fall back to EEIO for non-responders. Typical year 2 result: 30–50% of Category 1 spend is now covered by activity-based or supplier-specific data. Total footprint often drops noticeably as EEIO overestimates are replaced with real data.

Year 3 — Systematic primary data

Vendor data requests built into procurement onboarding (new vendors over a threshold spend are asked for emissions data as a condition of working with you). Category 1 hybrid coverage rises to 60–80%. Categories 11 (if product company), 12, 13 built out if material. Materiality screening refreshed and documented. Report format stable, assurance-ready, and sufficient for most supplier disclosure requests and SB 253 / CSRD obligations.

One caution: the year-over-year number will often decrease in year 2 not because emissions fell but because EEIO estimates are typically conservative (high). This is the right outcome methodologically but creates a communication challenge with executives — you have to explain that the apparent reduction is a data quality improvement, not an actual emissions reduction. Pre-brief this with the CFO before the year 2 report goes out.

Vendor engagement playbook for getting primary data

The choke point for moving to activity-based Category 1 data is vendor responsiveness. Most first-time requests get a 30–50% response rate. Here is what works, based on 40+ supplier engagement programs I have seen run.

Who to ask

Your top 10 vendors by spend. Not the top 10 by emission intensity, and not every vendor. The 80/20 applies here — 10 vendors will typically cover 60–80% of Category 1. Every hour spent chasing vendor #47 is better spent engaging vendors #1–10.

What to ask for (in this order of preference)

Their Scope 1 and 2 emissions and their revenue, for the same reporting period. Intensity = (Scope 1 + Scope 2) / revenue. Apply this intensity × what you paid them.
Their published sustainability or carbon report. Most mid-sized vendors either have one or are building one.
An industry-specific factor they recommend using for their sector (e.g., the DEFRA factor for printing, or a SBTi-approved intensity figure).
A commitment to provide data next year even if they cannot this year. This builds the baseline for year 2.

The template email that works

Subject: Quick carbon data request — [your company name] FY[year] emissions report. Body: short, direct, tells them exactly what you need and why. Three paragraphs: (1) we are building our company carbon report; your supply to us is one input; (2) here are the three specific data points we need (listed), and the format; (3) the deadline (30 days out) and that non-response defaults to EEIO sector average, which usually overstates your carbon footprint. That last line converts about 15% of non-responders in my experience — they would rather you have their real number than an overestimate.

What to do when they do not respond

Default to EEIO spend-based, document it, and add them to the year 2 outreach list with a note. Do not delay the full report waiting for one vendor. Partial primary-data coverage with documented rationale is infinitely better than no report at all.

Where our platform fits in this progression

Our Carbon Draft product is purpose-built for the year-1 screening use case. Upload your AP file and utility data; it classifies every line item to the appropriate EEIO commodity code (Category 1 default), applies the latest EPA Factors Hub factors to Scope 1 and 2 activity data, produces the materiality screening across all 15 Scope 3 categories, and outputs a GHG Protocol-aligned PDF report with methodology appendix and per-transaction data lineage. The Enhanced tier at $299 includes the data lineage a third-party reviewer would want to see; the output is designed to be your year-1 baseline that you progressively replace with primary data in year 2 and 3.

If you are starting from zero this year, generate a spend-based screening draft from your AP file.

If you are a CPA firm building this as a client service, the detailed workflow and pricing model is in our CPA's Guide to GHG Protocol Scope 1, 2 & 3.

If you are an SMB supplier asking where to start, ask your CPA first — they already have most of the data.