Emission Factor Data Quality Assessment
The GHG Protocol does not grade an emission factor in the abstract. It asks whether the data supports the claim you are making on it — which means the same factor can be entirely acceptable for internal tracking and unacceptable in a mandatory disclosure, without anything about the factor changing. Seven questions, about a minute, and this tells you which of those you have and which indicator is the weak one.
Five indicators decide it: geographical, technological, temporal, completeness and reliability. There is no combined score — the weakest indicator, measured against what the figure is used for, is the answer.
Seven questions, about a minute. The GHG Protocol does not grade an emission factor in the abstract — it asks whether the data supports the claim you are making on it. So the same factor can be fine for internal tracking and unacceptable in a mandatory disclosure, and this tells you which of those you have.
- The factor is more than five years old. On a decarbonising grid or a changing fuel blend that is not a small lag — it is a systematic bias in a known direction, because the published figure has been falling while yours has not. Whether a newer vintage exists is a separate and much quicker question than whether this one is defensible.
- A global or unspecified average is carrying a figure held to the highest bar. That is common and often unavoidable, and it is the axis an assurance provider asks about first, because the question has a cheap answer: either a country-specific row exists and was not used, or it does not exist and the proxy is justified. Which of those it is should be written down before somebody asks.
- The quality of this factor has not been recorded next to it. That matters more than it looks, because a factor outlives the person who chose it: the reasoning behind a substitution is the first thing lost on handover and the first thing asked for under assurance. A line per factor — which country, which process, which vintage, which source — is the whole of the fix.
What would change this verdict
- The use can change without the factor changing. A number that was fine for internal tracking is held to a different bar the moment it enters a disclosure or goes through assurance, and nothing about the factor itself moves.
- Temporal fit decays on its own. A factor that was the current vintage when chosen becomes a lagging one without anybody touching it, and on a decarbonising grid the drift runs in one direction.
Coverage position — each test, where you stand, and the rule
| Test | Your position | The rule |
|---|---|---|
| What the figure supports | A mandatory disclosureHeld to the highest barmet | The use decides which weaknesses matter |
| Geographical fit | A global or unspecified averageGlobal averageopen | GHG Protocol geographical representativeness |
| Technological fit | A sector or product averageSector averageopen | GHG Protocol technological representativeness |
| Temporal fit | More than five years oldMaterial lagopen | GHG Protocol temporal representativeness |
| Completeness | All the gases and stages we needCovers what is neededmet | GHG Protocol completeness |
| Reliability of the source | A government body or standards organisationIndependentmet | GHG Protocol reliability |
| Quality rating recorded | Not recorded anywhereNot recordedopen | The Standard expects the assessment to be disclosed, not merely made |
Why this verdict: One indicator is too weak for what this figure is carrying, and the use is what makes it too weak. The GHG Protocol does not grade factors in the abstract; it asks whether the data supports the claim being made on it, so the same factor can be entirely acceptable for internal tracking and unacceptable in a mandatory disclosure without anything about the factor changing. The two weaknesses that most often reach this level are a factor from a different country, where energy figures are not portable between them, and a factor that does not cover all the gases or stages the reported quantity claims. Both are fixable by substitution rather than by argument, and both are considerably cheaper to fix than to defend.
There is no composite score here, deliberately. The GHG Protocol does not define one, and averaging five ordinal scales would invent a number the Standard does not have — so the verdict is the weakest indicator measured against the use, which is how the Standard itself reasons. A PCAF score is a different question again: it depends on your asset class, and the Standard’s tables disagree between them, so it belongs to the PCAF data quality calculator rather than here.
Temporal fit decays on its own — a factor that was current when you chose it becomes a lagging one without anybody touching it. We will email you when a set this page applies to publishes a new vintage — not otherwise.
This assessment applies the GHG Protocol data quality indicators and the representativeness each MasterBrain row records. It reads no inventory of yours and calculates nothing. It tells you whether a factor supports the claim being made on it — not whether it sits in the right scope, which is a boundary question with its own checker.
What this verdict means
Five outcomes. The one at the top is not the worst rating — it is the absence of a rating, which is a different and more awkward position than a poor one.
Cannot be assessed
Three or more indicators are unestablished on a figure that leaves the organisation. Not a low score — no score, and nobody can say whether the factor suits the activity, including you.
Not fit for this use
An indicator is too weak for what the figure carries. The same factor might be perfectly acceptable for internal tracking; the use is what makes it unacceptable here.
Weak on one axis
Usable, with a known soft spot. This is a normal position — the Protocol does not expect every factor to score well everywhere, only that you know where the weakness is.
Quality unrecorded
The indicators support the use and none of it is written down beside the factor. Nothing is wrong; what is missing is the record, and a factor outlives whoever chose it.
Fit for the use
The indicators support what this figure is doing, and the assessment is recorded. A statement about this factor against this use — both halves matter.
The use sets the bar, not the factor
This is the part most quality discussions get backwards. There is no such thing as a good emission factor in general. There is only a factor that does or does not support the claim being made on it, and the claim is what moves.
| What the figure supports | The bar it is held to |
|---|---|
| A mandatory disclosure | Highest. Every weakness needs to be either fixed or disclosed, because somebody is required to look. |
| A figure going through assurance | Highest, and tested. A verifier will ask about the weakest indicator specifically. |
| A voluntary report or CDP | High. The audience is external and comparative, so substitutions need naming. |
| A customer or supplier request | High. You are handing somebody a number they will put in their own inventory. |
| Internal tracking only | Whatever is useful. A rough factor consistently applied still shows direction of travel. |
A factor is not promoted or demoted by its own qualities — it is measured against a use. The most common way to be wrong here is to grade a factor once and reuse the grade, when the thing that changed was where the number went.
The five indicators, and what each one asks
| Indicator | The question | Strong | Weak |
|---|---|---|---|
| Geographical | Is the factor from the same country as the activity? | Country-specific | A different country entirely |
| Technological | Does it describe your process, or an average of many? | Measured for this process | A different but related process |
| Temporal | How old is the vintage? | The current published one | More than five years |
| Completeness | Does it cover every gas and stage the figure claims? | All of them | CO₂ only, where CO₂e is reported |
| Reliability | Whose data is it? | Government or standards body | A supplier or vendor claim |
Two of these behave differently from the others and are worth understanding before you use the tool.
Temporal fit decays on its own. A factor that was the current vintage when you chose it becomes a lagging one without anybody touching it, and on a decarbonising grid the drift runs in a single direction — the published figure falls while yours stays where it was. That is not random error; it is systematic bias with a known sign, which is why it is the indicator most likely to turn a defensible position into an indefensible one through inaction.
Reliability and technological fit can pull against each other. A supplier-specific factor is usually the strongest available answer on the technological axis and the weakest on reliability, because it is self-reported. There is no resolution that makes both strong. What the Standard expects is that you record which trade-off you took and what verification, if any, sits behind the supplier’s number.
Why there is no overall score
This page deliberately does not produce a single quality number, and the omission is not a limitation.
The GHG Protocol does not define a composite score for these indicators. Averaging five ordinal scales — each of which is a set of labels rather than a measurement — would invent a figure the Standard does not have, and it would do the one thing a quality assessment must not: let a strong indicator conceal a fatal one. A factor that is country-specific, current, complete and government-published, but describes an entirely different process, is not eighty per cent acceptable. It is unacceptable, and an average would report it as good.
So the verdict here is the weakest indicator measured against the use, which is how the Standard itself reasons. It is also the form that is actually actionable: knowing your average is 3.2 tells you nothing to do, whereas knowing your temporal fit is the weak one tells you to go and check for a newer vintage.
“Cannot be assessed” is not a low rating. A low rating is a known position that can be disclosed and defended. An absent one means nobody — including you — can say whether the factor suits the activity, so it cannot be defended in either direction. It is also the cheapest of these problems to fix, because the answers come from reading the source documentation rather than collecting any data of your own.
The two substitutions most likely to be challenged
Every inventory contains substitutions. Two of them attract questions far more often than the rest, and both are answerable in advance.
A factor from another country. Energy factors are not portable between countries: a grid factor can differ by an order of magnitude, and a fuel factor differs by refinery slate and blend mandate. Where a country-specific row exists and was not used, the substitution is a choice rather than a constraint — and that distinction is exactly what a reviewer is trying to establish. Knowing which of the two applies, and being able to say so, converts a finding into a disclosure.
A factor for a related but different process. This is the substitution the person who knows your operation can most easily challenge, because they can usually name why your process differs from the one the factor describes. It is often the only option available, and it is never improved by leaving it unstated.
Record the five indicators next to the factor itself — which country, which process, which vintage, how complete, whose source. One line per factor. That record is what turns this from a judgement somebody made once into a judgement the next person can inherit, and it is the first thing requested under assurance.
Edge cases that change the answer
- The use can change without the factor changing. A number that was fine for internal tracking is held to a different bar the moment it enters a disclosure. Nothing about the data moved; re-run the assessment when the destination changes, not on a calendar.
- A global average is not automatically weak. For some activities no country-specific factor exists anywhere. The question is not whether you used an average but whether a better option existed and was passed over.
- Supplier-specific data is a trade, not an upgrade. It strengthens technological fit and weakens reliability. Both are true at once, and the Standard expects the trade to be recorded rather than resolved.
- A CO₂-only factor is not a slightly smaller CO₂e factor. It is a different quantity. The gap is small for most fuels and material for some, and it never shows up as an implausible value.
- Completeness failures hide inside familiar names. A factor missing a lifecycle stage carries the same name as one that includes it. This is where data quality and factor boundary overlap, and the boundary question has its own checker.
- An industry association is an interested party. That does not make a figure wrong. It does mean the reliability axis is not at its strongest, and saying so is cheaper than having it noticed.
What this checker does not decide
It does not produce a PCAF score. PCAF data quality depends on your asset class, and the Standard’s asset-class tables disagree with each other — an average factor supports one score for mortgages and a different one for listed equity. Our corpus deliberately publishes the inputs to that determination rather than a score. The PCAF data quality calculator is where those inputs meet your asset class.
It does not tell you whether the factor is in the right scope. A high-quality factor booked in the wrong scope is still wrong. That is a boundary question, and it has its own test in the Factor Boundary Checker.
It does not tell you whether a newer vintage exists. It asks how old yours is; whether the publisher has moved on is the Emission Factor Vintage Checker.
It does not tell you what to do when no factor exists at all. That is a sourcing gap rather than a quality problem, and the Emission Factor Gap Register handles it — including whether a geographic proxy has been properly disclosed.
It computes nothing. No score, no uncertainty range, no weighting. The indicators are labels, and this page treats them as labels.
How we keep this current
The five indicators are the GHG Protocol’s, not ours. What makes this page a diagnosis rather than a questionnaire is that our factor corpus records three of them against each row — geographical representativeness, technological representativeness and temporal lag — so the vocabulary here is the vocabulary the data already uses.
That coverage is not uniform, and it is worth being straight about: it is effectively complete on fuels, grid, spend-based, food and upstream-fuel families, strong on materials, and thin on transmission and distribution. Where the corpus does not record an indicator, this page asks you rather than assuming.
The checker stamps the date of the check into every result and every export. The re-check control emails you when a factor set this page applies to publishes a new vintage — which matters more here than on most pages, because temporal fit is the one indicator that degrades while you do nothing at all.
Frequently asked questions
None, deliberately. The GHG Protocol does not define a composite score for these indicators, and averaging five ordinal scales would invent a number the Standard does not have — while doing the one thing a quality assessment must not, which is letting a strong indicator conceal a fatal one. A factor that is country-specific, current, complete and government-published but describes an entirely different process is not eighty per cent acceptable; it is unacceptable. The weakest indicator, measured against the use, is the answer.
Two things change without the factor changing. The use may have moved — a number that was fine for internal tracking is held to a different bar the moment it enters a mandatory disclosure or goes through assurance. And temporal fit decays on its own: a factor that was the current vintage when you chose it becomes a lagging one while nobody touches it, and on a decarbonising grid that drift runs in one direction. Both are reasons to re-run this when the destination changes rather than on an annual cycle.
It is better on one axis and worse on another, and there is no resolution that makes both strong. Supplier data is usually the closest available match to your actual process, which is the technological indicator at its best. It is also self-reported, which is the reliability indicator at its weakest. The Standard does not ask you to prefer one; it asks you to record which trade you took and what verification, if any, sits behind the supplier’s number.
Not by itself. For some activities no country-specific factor exists anywhere, and a global average is the only honest option. The question a reviewer is actually asking is different: did a better option exist and get passed over? That has a cheap answer either way, and the answer is what should be written down. A proxy you can justify is a disclosure; a proxy you cannot account for is a finding.
It is the most awkward, though not because the data is bad. A low rating is a known position: it can be disclosed, justified, and defended. An absent rating means nobody can say whether the factor suits the activity at all — including you — so there is nothing to defend in either direction. It is also the cheapest problem here to fix, because the answers come from reading the source documentation rather than collecting any data of your own, and every other judgement on the page depends on them.
There is no universal threshold, and the direction matters more than the number. On a stable process a factor several years old may be perfectly representative. On a decarbonising grid or a fuel with a changing blend mandate, lag is a systematic bias with a known sign — the published figure has been falling while yours has not — so the error compounds rather than averaging out. This page treats more than five years as material and a few years as worth checking, and the check itself is usually a one-minute question with a clear answer.
No, and that is a deliberate limit rather than an omission. A PCAF score depends on your asset class, and the Standard’s asset-class tables disagree with one another — an average factor supports one score for mortgages and a different one for listed equity. Our corpus publishes the inputs to that determination against each factor rather than a score, precisely because the score is not a property of the factor. The PCAF data quality calculator is where those inputs meet your asset class.
The most likely reason is that the quality was never written down. The Standard treats the assessment as something to disclose, not merely to perform, and a verifier cannot take on trust a judgement they cannot see. The second most likely reason is that the query is not about quality at all but about placement — a high-quality factor booked in the wrong scope is still wrong, and that is a boundary question with its own test.
Not to the same depth. The proportionate approach is to assess properly wherever a factor drives a material share of the inventory or supports an external claim, and to record the five indicators for everything else as a matter of routine when the factor is chosen — which costs a line and saves the whole exercise later. The indicators are also usually shared across a factor set rather than differing row by row, so one reading of a source’s methodology often settles a large number of factors at once.