PCAF data quality scores explained: scores 1 to 5 and how to improve

The PCAF data quality score rates every financed emissions figure from 1 (most reliable) to 5 (least reliable), depending on whether it comes from a borrower's reported emissions, from physical activity data or from economic proxies. Institutions disclose a portfolio score weighted by exposure, so readers can see how much of the total rests on estimates. Improving it means replacing proxies with reported or activity data, starting with the exposures that carry the most emissions.

What the PCAF data quality score is

The Partnership for Carbon Accounting Financials (PCAF) publishes the Global GHG Accounting and Reporting Standard for the financial industry. Part A covers financed emissions from loans and investments, and its third edition was released in December 2025. The five-level data quality scale was kept in that edition; what changed was mainly new asset classes and clearer attribution and reporting expectations.

The scale exists because few banks or investors have reported emissions for every counterparty. PCAF does not ask you to wait for perfect data. It asks you to use the best data available for each exposure, estimate the rest, and label each figure honestly. A score of 1 means the number is close to certain; a score of 5 means it is a sector-average estimate that may be far from the truth for that particular company.

For the basics of attribution and asset classes, read our PCAF guide to financed emissions.

PCAF score 1 to 5: the general ladder

• Score 1: reported emissions that have been verified by a third party.
• Score 2: reported but unverified emissions, or (for some asset classes) emissions calculated from primary energy data.
• Score 3: emissions calculated from physical activity data, such as production volumes, with activity-specific emission factors.
• Score 4: emissions estimated from economic activity, typically the counterparty's revenue multiplied by a sector emission factor per unit of revenue.
• Score 5: coarser economic estimates, such as sector factors per unit of assets, or revenue inferred from asset turnover ratios.

The exact mapping differs by asset class. A physical-activity estimate may score 2 in one asset class and 3 in another, so always score against the scorecard for the asset class in question.

Data quality scores by option

For listed equity, corporate bonds, business loans and unlisted equity, the scorecard is organised into three options. The table summarises it in plain terms.

Option	Data used	Score
1a — reported emissions	Company's emissions, verified by a third party	1
1b — reported emissions	Company's emissions, not verified	2
2a — physical activity	Primary energy consumption data with emission factors specific to that energy (plus any process emissions)	2
2b — physical activity	Production data with activity-specific emission factors	3
3a — economic activity	Company revenue with a sector emission factor per unit of revenue	4
3b — economic activity	Company assets with a sector emission factor per unit of assets	5
3c — economic activity	Sector revenue factor combined with a sector asset turnover ratio	5

Real-economy asset classes use different ladders. For mortgages and commercial real estate, metered energy with supplier-specific factors scores 1, metered energy with average factors scores 2, energy labels with floor area score 3, statistical energy use per floor area scores 4, and statistics per building score 5. Motor vehicle loans run from actual fuel use (score 1) down to statistics for an average vehicle (score 5). Check the third edition's scorecards for project finance, sovereign debt and the newer asset classes.

How the weighted data quality score works

A single portfolio figure is produced by weighting each exposure's score by its outstanding amount:

Weighted data quality score = Σ (outstanding amount × score) ÷ Σ outstanding amount

Illustrative example. A loan book of 200 (in millions) contains 50 scored 1, 30 scored 3, 100 scored 4 and 20 scored 5. The weighted score is (50×1 + 30×3 + 100×4 + 20×5) ÷ 200 = 640 ÷ 200 = 3.2.

Report the score by asset class as well as in total. A blended figure can hide a mortgage book scored 4.8 behind a well-covered listed equity book. Many institutions also calculate an emissions-weighted score for internal use. It shows how much of the total financed emissions figure itself rests on proxies, which is often more revealing than the exposure-weighted figure because a few heavy emitters dominate the total.

How to improve your financed emissions data quality

Moving the score from around 4 towards 2 is a multi-year programme. A sequence that works:

1. Rank exposures by impact. Sort counterparties by estimated financed emissions and by exposure. In most books a short list of large, carbon-intensive borrowers drives both the total and the score.
2. Harvest public data first. Listed companies often publish emissions in annual or sustainability reports, CDP responses or, in India, BRSR filings. Each one found moves an exposure from 4 or 5 to 1 or 2 at little cost.
3. Request data from the rest. Send structured requests to material unlisted borrowers, asking for Scope 1 and 2 (and Scope 3 for relevant sectors), the reporting year, and whether figures are verified.
4. Use physical data where emissions are not reported. Energy bills, production volumes, floor area and energy labels support scores 2 and 3. Build their collection into credit onboarding and annual reviews.
5. Upgrade the proxies. Even for score 4 estimates, use sector factors that match the counterparty's region and activity rather than a global average.
6. Track the score over time. Set an internal target, for example a fall in the weighted score each year in the highest-emitting sectors, and report progress.

Common mistakes with PCAF scores

• Scoring the company, not the figure. One counterparty can have verified Scope 1 and 2 but estimated Scope 3. Score each scope separately.
• Treating unverified as verified. Only third-party verified reported emissions earn a 1. Note who verified and to what level.
• Ignoring the attribution data. Scores describe emissions quality, but outstanding amounts and company values (EVIC, equity plus debt, property value at origination) must also be correct and current.
• Mixing reporting years silently. Record the year of each emissions figure; PCAF expects the most recent available data and an explanation where years differ.
• Dropping estimates to improve the score. Excluding hard-to-measure exposures flatters the score and weakens coverage. Disclose coverage alongside quality.

How EKOFIN helps

EKOFIN calculates financed emissions across PCAF asset classes and assigns a data quality score to each exposure based on the data source behind it. Dashboards show the weighted score by asset class and sector, and highlight the counterparties where better data would move the score and the total most. Counterparty data requests can be sent and tracked from the platform, and the scores feed directly into IFRS S2 and TCFD-ready outputs.

See how the method works on our financed emissions page, or read how the score fits with portfolio metrics in WACI and financed emissions metrics.

Frequently asked questions

What is a good PCAF data quality score?

Lower is better, with 1 the best possible. First inventories often sit towards the weaker end of the scale because many exposures rely on revenue-based estimates. What matters to readers is a credible improvement trend and transparency about which asset classes are weakest.

Is the PCAF data quality score weighted by exposure or emissions?

The disclosed portfolio score is weighted by outstanding amount. An emissions-weighted score is a useful internal view because it shows how much of the emissions total rests on estimates, but label clearly which weighting you use.

Does PCAF require verified emissions?

No. PCAF accepts unverified and estimated data, provided each figure carries the right score. Verification simply earns the best score for listed equity, bonds and loans.

Did the 2025 PCAF update change the data quality scale?

The December 2025 third edition of Part A kept the five-level scale and added new methods (use of proceeds, securitisations and structured products, sub-sovereign debt) and new reporting recommendations. Check the asset-class scorecards in the current edition when you set up scoring rules.