Emission Factor Vintage Checker
Most staleness checks look for a number that moved. The expensive ones are the factors where nothing moved at all — the value is byte-identical to the day you copied it, and the thing it is allowed to say has changed underneath it. This checker tests the factor editions you already hold against our own transition register and tells you which of the three failures you have. (If you came here about carbon credits, the year a reduction occurred is a credit vintage and is a different subject.)
An emission factor goes stale three ways: the value moves, the key is withdrawn, or the key is reclassified. The third is the dangerous one — the number is identical and your use of it may be wrong.
Seven questions about the factors you already hold, about a minute. Nothing is sent anywhere — the check runs in your browser against the edition and supersession registers this page loads, and the link it produces encodes only your answers.
- Re-record the factor under the key it is actually published at — the lifecycle-intensity key — so the name in your inventory matches the instrument behind the number.
- Check whether that figure sits inside a Scope 2 total. A lifecycle intensity asserts no GHG Protocol scope; used as a location-based factor it overstates Scope 2.
- Where a national direct factor exists for the market, that is the correct Scope 2 source. The UK and the US always had one, which is why neither was ever affected.
- The successor value is recorded as unchanged, which is exactly why this is easy to miss: a comparison of the two numbers finds nothing.
- A figure built on this is already published, so the correction is a restatement question rather than a quiet edit in next year's file.
- A reclassification is not a value change, so a materiality test that looks at how much the number moved will not catch it. The question is whether the instrument was right, not whether the figure shifted.
What would change this verdict
- Re-pulling the affected set and recording the edition you pulled clears the edition limb.
- Recording the same value under the lifecycle-intensity key clears this: the figure is unchanged and the name then matches the instrument.
- A market with its own national direct factor was never affected — Ember never served a location-based row for the UK or the US.
Editions checked — every limb, its transition and the register it came from
| Limb | What you hold | Recorded in |
|---|---|---|
| Factor set | Ember Yearly Electricity DataRecognised publisher | meta.sources — the edition register |
| Edition held | The current editionNo edition transition | meta.sources — edition comparison |
| Grid row recorded | Ember country factor, location-basedRecorded transitionreclassified | meta.supersessions — the transition register |
| Market | Another marketWithin the reclassified setreclassified | meta.supersessions — relation by market |
| GWP basis | AR5, 100-yearBasis stated | Declared by you |
| Last re-pull | Previous reporting cycleWithin one cycle | Declared by you |
| Published figure | Yes, already publishedRestatement question | GHG Protocol Corporate Standard ch.5 |
Why this verdict: The declared grid row is an Ember country factor recorded as location-based. That key is recorded in the supersession register as reclassified, with its successor at the lifecycle-intensity key and its value unchanged. Location-based is a GHG Protocol Scope 2 term of art asserting a direct grid-average instrument, while the figure Ember publishes is a full-lifecycle, all-gas, net-generation intensity. So the number stands and the assertion attached to it does not. Because a figure built on it has already been published, the correction is a restatement question rather than a next-cycle edit.
Values, where any are shown above, are labelled as published in a named edition. They are not marked old or correct: the register records what each publisher published, and does not rule that one of them is right. A pair is shown only where both editions come from sources we are licensed to redistribute; otherwise the transition is named and the figures are not.
- meta.supersessions — the transition register, read for every declared limb
- meta.sources — the edition and cadence of the declared publisher
Recorded at the point each lookup returned, not declared in advance. A key this check could not resolve is listed as unresolved rather than left out.
This is the one check that goes out of date on its own. Publishers reissue, keys get withdrawn, and a reclassification changes nothing you can see. We will email you when an edition named in your answers changes — not otherwise.
This checker reads our own edition and supersession registers. Absence of a recorded transition means no transition has been recorded for that key at this corpus edition — it is not a guarantee that a publisher has not moved. It does not tell you which factor to use, and it does not decide whether a change is material enough to restate.
What this verdict means
Five outcomes, ordered by how much trouble they are rather than by how loud they look. The checker reports the worst limb you declared, never an average — a current DEFRA edition does not offset a reclassified grid row, and averaging them would let it.
Reclassified
The value stands and the meaning does not. The key you hold was found to have misdescribed what it carried, so the figure survives under a different name with a different instrument attached. Nothing in your spreadsheet looks wrong, which is why this ranks first.
Withdrawn
Nothing is published under that key any more, and no successor was recorded. It is a real problem and a loud one: the key resolves to nothing and you find out the first time you look.
Superseded
A newer edition exists. Your figures are not wrong — they are what the publisher published — but they are not the current ones, and if the instrument changed as well as the number, the gap is larger than a diff suggests.
Not established
A limb could not be tested because it is not written down anywhere: an edition, a grid instrument or a GWP basis. This is reported as its own verdict rather than folded into a pass, because an untested limb is not a passing limb.
Current
No transition is recorded against anything you declared. Read that precisely: it is a statement about what our register holds at this corpus edition, not a guarantee that a publisher has not moved since.
The three ways a factor goes out of date
Everyone checks the first one. Almost nobody checks the third, because there is nothing to check against unless someone recorded what happened to the key — and recording it is a deliberate act, not something that falls out of a database.
The value moved — a new edition of the same instrument
The publisher reissues, the number changes, the key and the meaning stay put. This is the failure everyone has a process for, and it is the least interesting of the three because a diff finds it. The UK grid factor moved from 0.177 kg CO₂e per kWh to the 0.13096 [GreenCalculus grid.gbr.electricity.location_based · DEFRA 2026 'UK electricity'!E25 · v2026.284] we serve today, on a genuine 2024-data decarbonisation recorded in DEFRA 2026 — same instrument, same key. If your model still carries the older figure, a re-pull fixes it and nothing conceptual is at stake.
The key was withdrawn — nothing is published there now
The publisher stops issuing the thing entirely, or we stop serving it because no publisher issues it. A withdrawn key is the honest failure mode: it resolves to nothing, your lookup breaks, and you go and find out why. Annoying, visible, fixable.
The key was reclassified — same number, different meaning
A reclassification means the value was right and the name was wrong. The figure carries over byte-identical, so every check built on comparing numbers reports that nothing has happened. What changed is the claim the key was making about what the number measures — and that claim is what your disclosure inherits.
This is the failure that a comparison cannot catch, an AI summary cannot know, and a vendor’s “factor freshness” dashboard cannot see, because all three are looking at values. It needs a register that records what happened to a key, with enough vocabulary to distinguish outcomes that mean opposite things.
| Relation | What happened | Does your figure stand? | Does your use stand? |
|---|---|---|---|
| renamed | Same value, same meaning, better key | Yes | Yes |
| reclassified | Still published, but the old key misdescribed it | Yes | Maybe not |
| instrument changed | The key survives; a different measurement now answers it | No — superseded | Re-check the basis |
| dropped | Nothing published since | No | Re-source it |
The obvious design is a single “see instead” pointer from the old key to the new one. It would have been dangerous. A pointer says your figure stands, carry on — which is true for a rename and false for a reclassification, and the reclassifications outnumber the renames by every entry we hold.
What this checker reads
Two registers, both of which we maintain as data rather than derive at runtime. Neither is a factor value: this tool interrogates factors, it does not serve them. If you want the number itself, that is what the factor browser is for.
The source register — every publisher with its edition and cadence
Each of the source documents behind our 21,133 factors carries its publisher, its edition year, the date we retrieved it, its reissue cadence and its licence position. That is what lets the checker say “DEFRA reissues each June” rather than “check whether there is a newer one”, and it is why the edition question in the tool names the edition we currently serve instead of asking you to know it.
The supersession register — recorded transitions, four relations
Every key transition is recorded once, keyed by the superseded key — which is the whole point, because a reader arrives holding the old one. At corpus edition 2026.182, in August 2026, the register held 212 entries under the four relations above, including 208 keys that no longer exist in the corpus at all. That count is stated as of a named edition rather than served live, because “no transitions in this release” is a real and healthy state, and a counter that can legitimately read zero has no business sitting mid-sentence.
Absence from the register means no transition has been recorded for that key at this corpus edition. It is not a guarantee that a publisher has not moved. The register records transitions; it does not assert their absence, and the checker says so rather than rounding it up to “you’re fine”.
The 2026 grid reclassification, and who it caught
The clearest example of the third failure is one of ours, and it is the reason this page can be specific where everyone else has to be general.
Why an identical number can be the wrong number
Ember publishes a country electricity intensity that is full-lifecycle, all-gas and measured on net generation. We had been serving it at a key ending location_based. That phrase is not a description — it is a GHG Protocol Scope 2 term of art, and it asserts a direct grid-average instrument. The number was Ember’s and correct; the assertion attached to it was ours and wrong.
So the rows moved to a lifecycle-intensity key, and the values went with them unchanged. Anyone who had put one of those figures into a location-based Scope 2 total had overstated it — not because the number was bad, but because a lifecycle intensity includes things a Scope 2 figure is not supposed to include. Our own release note put it plainly: it would not survive audit.
The two markets that were never affected are the instructive ones. The UK and the US both have a genuine national direct factor — DEFRA and eGRID — so neither ever sat on an Ember row, and neither moved. If you hold an “Ember location-based factor” for either country, you are holding something that has never existed, and the checker says so rather than guessing which of the two things you actually have.
The four countries where the instrument changed, not just the value
Four markets kept the same key and changed publisher, basis and number together. Reading that as a value change under-describes it: a consumption-basis national figure includes interprovincial trade and transmission losses that a generation-basis figure does not, so the two are not competing estimates of one quantity. They measure different things.
| Market | As published in the earlier edition | As published in the current edition | What actually changed |
|---|---|---|---|
| Canada | Ember, lifecycle intensity | ECCC National Inventory Report | Publisher, and the basis is consumption |
| Australia | Ember, lifecycle intensity | DCCEEW National Greenhouse Accounts | Publisher, and it is an NGER Scope 2 figure |
| France | Ember, lifecycle intensity | ADEME Base Carbone | Publisher and basis |
| Netherlands | Ember, lifecycle intensity | CBS national statistics | Publisher and basis |
The checker shows the figures on either side of a transition like these inside its audit trail, labelled as published in each edition — never as old and correct. It is not our place to rule that one publisher’s number is right; the register records what each of them published. Where a source’s licence does not permit us to republish its values, the transition is named and the figures are not.
Edge cases that change the answer
A source that does not state its GWP basis
Some databases publish CO₂e without naming which GWP set produced it. That is not an error on your side and it is not one you can resolve by reading harder — the basis genuinely is not stated. We record it as a property of the row rather than a note in prose, so it can be filtered rather than discovered. If you are on such a source, the honest verdict here is not established, with the reason named, and the thing to do is record the limitation rather than assume a basis.
A supplier-specific factor we cannot see
If your factor came from a supplier rather than a publisher, no register can test its edition and this checker will say so instead of pretending. What still applies is the discipline: a supplier factor needs its own version, its own date and its own retained copy, for exactly the reasons a published one does. The Assurance Readiness Checker is the tool that tests whether you hold that evidence.
A frozen worked example is not a stale factor
You will find the superseded 0.177 UK grid figure in more than a hundred places on this site, and it is there deliberately. A worked example is an audit record: it shows a stated input producing a stated result, and the arithmetic has to remain checkable years later. If we made the factor self-updating while the printed result stayed fixed, the example would quietly contradict itself — a live number multiplied into a frozen total. So worked examples are frozen and labelled as the vintage they were computed on, and every other number on the site is served live from the corpus.
A frozen figure that says which edition it came from is a record. The same figure sitting unlabelled in a live inventory is a liability. An unlabelled 0.177 in your spreadsheet is a different thing from a labelled one in ours — and telling the two apart is the entire job of this checker.
What this checker does not decide
It does not tell you which factor to use. Choosing between two live factors is a question about fit, not currency, and the factor browser is where you look one up.
It does not adjudicate your GWP basis. Whether AR5 or AR6 is right for you depends on the regime you report under rather than on any edition — Australia’s standard pins a set, most others do not — so this page asks which basis you are on and stops there. The comparison itself is covered in AR5 vs AR6.
It does not recompute your inventory. If the finding is that a whole set has moved underneath you, the Scope 1, 2 and 3 inventory aggregator is where the numbers get rebuilt with an audit trail attached.
And it does not decide whether a change is material enough to restate. That turns on your recalculation policy, and whether you have one is what the Carbon Accounting Maturity Assessment tests. Note the trap in applying an ordinary materiality threshold here: a reclassification does not move the number, so a test that asks how much the figure changed will report nothing. The question is whether the instrument was right.
How we keep this current
Every transition is recorded once, at the release that caused it, and verified when it is written rather than computed on demand. That is deliberate: 208 of the entries map mechanically and a renderer could derive them in three lines, but deriving them would encode one incident’s shape as though it were the format. The next withdrawal will not look like this one — two of the four relations currently have no instances at all, and their absence is a fact about this incident rather than about the vocabulary.
This page is checked against corpus edition 2026.284, covering 21,133 factors across 151 countries. A verdict from this tool is perishable by design, which is the one thing every other check on this site is not: the answer can change without you touching anything. That is what the re-check in the result panel is for — it watches the specific editions you named, and nothing else.
Frequently Asked Questions
Because a reclassification does not change the number. It changes what the number is entitled to be used for. The figure you hold is the same figure we serve; what moved is the key it is published under, and with it the instrument the key asserts. If the old key claimed a direct grid average and the value is a lifecycle intensity, a Scope 2 total built on it is overstated even though nothing about the arithmetic changed.
Yes, and it is the most common way this tool returns a finding. Edition currency and instrument correctness are separate axes. You can be on the newest release of a publisher’s data and still have recorded one of its rows under a key that no longer describes it. The worked example the tool loads with is exactly that case.
This checker does not make that call, because it turns on your own recalculation policy and on the significance threshold you set. What it does do is tell you that the question exists, and warn you that the usual test will not fire: a materiality rule expressed as a percentage change in the number finds nothing when the number did not change. If the figure was assured, the practitioner who signed it is part of the conversation too.
That no transition is recorded against the limbs you declared, at the corpus edition stamped on your result. It is a statement about the register, not a warranty about the world. A publisher can reissue tomorrow and the register will record it at the next release — which is precisely why the tool offers to watch the editions you named rather than inviting you to re-run it on a hunch.
Not fatal, but it is a finding in its own right rather than a gap in the tool. An edition you cannot name is an edition you cannot show an assurer, and it is the first artefact an engagement asks for. It is also the cheapest of these problems to close: write the publisher, the edition and the retrieval date into the inventory file itself, and keep your own copy rather than a link to theirs.
Because those four behaved differently from everywhere else. They kept the same key and changed publisher, basis and value together — an instrument change rather than a reclassification. Grouping them is not a simplification; it is the shape of what actually happened, and treating them like the other markets would tell you your figure stands when it does not.
No, and the tool says so rather than returning a confident silence. The registers cover the publishers and keys in our corpus. A supplier-specific factor, or a set we do not carry, cannot be tested for edition currency here — what applies to it is the same discipline, not the same check.
Not at all — it is the right instrument for a product footprint or an LCA, and we publish it as its own key for that reason. It is wrong only inside a GHG Protocol Scope 2 total, which is a narrower thing that deliberately excludes upstream lifecycle burden. The problem was never the figure; it was a lifecycle figure wearing a Scope 2 name.