PCAF data quality: how the score is set, asset class by asset class

Reference guide. Last reviewed September 2026, against PCAF Part A Third Edition (December 2025).

Your financed emissions number is only as good as the method behind each line of it, and PCAF scores that method rather than the number. Most explanations of the score stop at a single five-tier table. There is no single five-tier table.

What does a PCAF data quality score actually measure?

A PCAF data quality score records how close the underlying data sits to the counterparty, on a scale of 1 to 5, where 1 is the highest quality. It scores method, not accuracy. A score of 5 does not mean the tonnage is wrong; it means the tonnage was derived from a sector emission factor and an outstanding amount, so it will move when the exposure changes and not when the counterparty cuts emissions.

Three properties follow, and they explain most of the confusion around the score.

• The same score can be reached by different routes. For corporate exposures, a score of 2 comes either from the counterparty's own unverified emissions figure or from its metered energy consumption. A reader of your disclosure cannot tell which.
• Better data does not always mean a better score. An unverified figure the counterparty published itself scores 2. Primary production data you collected from that counterparty scores 3.
• The score belongs to the exposure, not to the company. It is assigned per loan or investment, then normalised across the portfolio by outstanding amount.

Which version of the PCAF standard applies?

PCAF published the Third Edition of Part A, Financed Emissions, on 2 December 2025. It covers ten asset classes, up from seven in the December 2022 Second Edition and six in the first: listed equity and corporate bonds, business loans and unlisted equity, project finance, commercial real estate, mortgages, motor vehicle loans, use of proceeds structures, securitisations and structured products, sovereign debt, and sub-sovereign debt.

For anyone maintaining an existing model, the useful detail is what did not change. The data quality option tables for the six original asset classes are unchanged from the Second Edition, and so is the weighted score requirement. What arrived is three new tables, a rule that use of proceeds structures and securitisations take a weighted average of their underlying assets' scores rather than having tables of their own, and two new reporting recommendations: a fluctuation analysis explaining what drove the change in absolute financed emissions between periods, and an optional inflation adjustment that PCAF scopes specifically to score 4 and score 5 emission factors.

PCAF publishes no effective date for the Third Edition and no transition period. Its own signatory count reached 719 across more than 85 countries as of December 2025, managing close to $100 trillion in assets, per the PCAF Impact Report 2025.

One conformance point worth carrying into an audit conversation. The Built on GHG Protocol mark covers the six asset classes reviewed under the 2020 first edition. The GHG Protocol has since closed that review service, so the additions in the 2022 and 2025 editions carry no GHG Protocol review. PCAF states this in the Third Edition itself.

Why is there no single PCAF data quality table?

Because scoring depends on what data can exist for the asset. PCAF puts it directly: "Data quality scoring is specific to each asset class." The general scorecard in Chapter 4 is a gradient diagram, not an operative table. The tables that decide a score sit in the asset class chapters and in Annex 10.1.

A generic table that says "score 3 equals company-specific physical activity data" is therefore right for a business loan and wrong for a mortgage, where score 3 is an estimate from an energy label and a floor area, and wrong for a motor vehicle loan, where score 3 is regional statistical distance data. Applying one table across a mixed book produces scores your assurance provider will not be able to trace back to the standard.

How is data quality scored for listed equity, corporate bonds, business loans and unlisted equity?

These asset classes share one structure. Three options, seven sub-options, five scores. Listed equity and corporate bonds attribute against enterprise value including cash, or total equity plus debt for bonds to private companies. Business loans and unlisted equity attribute against total equity plus debt, or EVIC for loans to listed companies. The scoring is otherwise identical.

Two constraints sit in the footnotes rather than the table. Option 2a can only score scope 1 and scope 2, because scope 3 cannot be estimated from energy consumption, so an energy data route leaves your scope 3 lines to be scored some other way. And substituting a different financial indicator for revenue under option 3a is permitted without affecting the score.

How is data quality scored for commercial real estate and mortgages?

Real estate uses a different logic, and the word "actual" carries the weight. Both asset classes use the same table, and both attribute against the property value at origination.

Here the route to each score is single. There is no duplication to explain in a footnote, which makes the real estate portion of a disclosure easier to evidence than the corporate portion.

How is data quality scored for motor vehicle loans?

Score 1 has two routes, which is unique among the asset classes. Actual fuel consumption earns a 1, and so does actual distance travelled combined with fuel efficiency from a known make and model. After that the ladder is about how specific the statistics are: local statistical distance data with a known make and model scores 2, regional statistical distance data with a known make and model scores 3, local or regional statistics with only the vehicle type scores 4, and local or regional statistics against an average vehicle scores 5.

How does the attribution factor relate to the data quality score?

The attribution factor decides how much of a counterparty's emissions land in your inventory. The data quality score describes how those emissions were arrived at. They are independent: a precisely attributed exposure can carry a score of 5, and a rough attribution can carry a score of 1.

PCAF requires a fixed point in time for lending and investment positions, and the GHG accounting period has to align to the financial accounting period. Where a vehicle's value at origination is unknown, PCAF asks for a conservative assumption of 100% attribution.

What separates DQ 2 from DQ 3?

For corporate exposures, the step from 2 to 3 is the step from energy input data to production output data. Option 2a takes the counterparty's metered energy consumption by source and applies emission factors specific to that data, with process emissions added before the attribution factor. Option 2b takes physical units produced and applies a production emission factor. Energy data scores better because the emissions follow more directly from measured consumption than from inferring them from output volumes.

The practical consequence is the one most programmes get backwards. A score of 2 is also available through option 1b, the counterparty's own unverified published figure, which costs the counterparty nothing it has not already done. Collecting primary production data from that same counterparty is real work on both sides and lands at 3. So the fastest route up the hierarchy for most of a book is not a data collection exercise. It is finding what has already been published, and evidencing it.

How is the weighted data quality score calculated?

By outstanding amount, not by financed emissions. PCAF's formula multiplies each exposure's outstanding amount by its data quality score, sums that across the portfolio, and divides by total outstanding amount:

Weighted DQ score = Σ (outstanding amount × DQ score) / Σ outstanding amount

PCAF's own worked example produces 3.03 for a business loan portfolio and 3.53 for an oil and gas sector view. Two rules govern how it is reported. Institutions should publish the weighted score by outstanding amount or explain why they cannot, and PCAF's accompanying guidance describes normalisation to outstanding amount as a requirement of the standard. Where scope 3 emissions are reported, the weighted score for those emissions shall be reported separately from the score for scopes 1 and 2.

The weighting choice matters more than it looks. Because the score is weighted by exposure rather than by emissions, a large exposure in a sector that emits little can make the portfolio score look better while the emissions that actually drive your inventory still rest on a sector average. Reporting the score by asset class and by sector, as PCAF's disclosure template lays out, is what makes that visible.

What does PCAF require you to disclose about data quality?

The requirements sit in Chapter 6 of Part A, and PCAF separates what you shall do from what you should do.

• Shall: use the most recent or otherwise appropriate data available, accepting that emissions data and financial data may represent different years. Report the weighted data quality score for scope 3 emissions separately from scopes 1 and 2. Disclose the methodology if an inflation adjustment is applied. Disclose absolute financed emissions for all relevant asset classes and justify exclusions, and disclose the percentage of total loans and investments covered.
• Should: describe the types and sources of data used, including activity data, assumptions, emission factors and publication dates, written to create transparency. Publish the weighted score by outstanding amount or explain why you cannot. Use the data hierarchy tables in Chapter 5 as the guide for disclosing data quality, and explain how data quality is assessed. Verify data to at least limited assurance where possible over time, and disclose whether data is verified and to what level.

PCAF's own disclosure template puts the weighted data quality score in the same table as the emissions, one row per asset class, alongside the outstanding amount, scope 1 and 2, scope 3 and emissions intensity. The score is presented as part of the number, which is why an unexplained score is treated as an incomplete disclosure rather than a footnote.

What actually moves a portfolio up the hierarchy?

Four moves, and they cost very different amounts. Being honest about which is which is what makes a DQ roadmap survive its first review.

The first three rungs turn on information that already exists somewhere. The fourth is a genuine data collection project, and it earns its place on the exposures carrying the emissions rather than across a whole tail. A roadmap that puts collection first spends its budget on the hardest rung while the estimated tail stays where it is.

One trap in the assurance step. A counterparty whose scope 1 and 2 were third-party assured but whose scope 3 was not is the common case, and treating that as one assured figure misstates the scope 3 score. Because PCAF asks for the scope 3 weighted score separately, assurance status has to be held per scope to be reportable at all.

How does DitchCarbon fit into a PCAF model you already run?

DitchCarbon provides verified emissions data for over 2 million organisations, so procurement, sustainability and finance teams can measure and act on supply chain and portfolio emissions from one source. For a PCAF programme, that is an input to the model you already have rather than a replacement for it. Your asset class methods, your attribution factors, your factor set and your published baseline stay where they are. What changes is how much of the estimated tail can move up the hierarchy inside them.

Three things are needed before a score can move, and each is a mechanism you can inspect.

• The counterparty has to be identified correctly. Organisations are matched on DUNS, LEI, ISIN or VAT number, with domain, email, HQ region and industry as supporting signals, and parent and subsidiary relationships resolved, so a figure attaches to the entity you hold rather than the group above it. A DQ score describes how close the data sits to the counterparty, so it only holds if the counterparty was the right one.
• The method behind each figure has to be visible. Every organisation carries its own method marker: its own reported emissions, product data from a published carbon footprint, activity data from what was actually consumed, or an industry factor where nothing better exists. The column is filterable, so the rows still resting on a sector average are a list you can pull rather than a proportion you have to estimate. A headline toggle switches the figure between organisation emission factors and industry factors only, and the difference between the two numbers is how much of the total rests on what organisations actually published.
• The document has to be there when someone samples it. Every source document is listed with its year and three assurance columns, one per scope, so a third-party assured scope 1 and 2 with an unassured scope 3 reads as exactly that. Annual reports, CDP questionnaires and product footprints sit together. Click any emissions figure and it opens as its own calculation, one row per year, naming the methodology, the description it was calculated against, the quantity, the emission factor with its units and the source that factor came from. The same detail exports as an Excel audit workbook: totals by methodology, a summary per organisation, the line by line calculations and the factors applied.

When your assurance provider asks how the figures were produced, the answer is the workbook and the source documents. The calculator that produced them is verified to ISO 14064-3, limited assurance, by UL Solutions, renewed annually, and DitchCarbon data has been used in emissions reports subsequently assured by ten different third-party assurance providers, including Big Four firms. Both reports are on the trust centre.

What sits underneath is three layers, and a PCAF model draws on all three: company emissions data for over 2 million organisations, including corporate-level GHG inventories; supplier-specific emissions at spend, activity and product level; and a large library of published emission factors, including ecoinvent, CEDA, EPA, DEFRA and EXIOBASE. The generic layer is what closes the lines where nothing has been disclosed, and it is fourth of four in the order rather than the first answer.

Emissions history runs fifteen years with restatements handled as a single series, and coverage gaps are shown rather than hidden. For a fuller answer on why most portfolios sit at DQ 4 and DQ 5 in the first place, see PCAF data quality scores explained.

Sources

• PCAF (2025). The Global GHG Accounting and Reporting Standard Part A: Financed Emissions, Third Edition, December 2025. Asset classes Table 5-1; data quality tables 5.1-2, 5.2-1, 5.4-1, 5.5-1 and Annex 10.1; attribution factors Chapter 4 and Chapter 5; weighted score and disclosure requirements Chapter 6.
• PCAF (2022). Part A: Financed Emissions, Second Edition, December 2022.
• PCAF. Impact Report 2025, signatory and asset figures as of December 2025.
• PCAF. PCAF launches updated GHG accounting standard, 2 December 2025.