How do I find out why a factor says what it says?

Part of How a factor is decided. The last page of the section, and the one to read when a number looks wrong.

Every consensus factor answers the question in one word, its derivation, and the word says where to look next.

`derivation`	it means	look at
`sole`	only the method's own publisher stated it	the publisher's own row, on the same flow under their implementation
`agreed`	two deciding implementations stated one number within 2 %; the more precise printing is published and the other is in `also_stated`	both rows on the flow's page
`ruled`	a curator chose between numbers, or published in spite of the model	`data/lcia-factor-rulings.json`, and the queue row — now marked ruled — with the comment
`restated`	only a transcription stated it, and this list already published the same number for the same substance in another context of the category, within 2 %	the same substance's other contexts
`adopted`	a person signed for the substance taking another substance's number — an ion its element's, a pesticide a catch-all's, a land class its family's	`data/lcia-factor-adoptions.json`: the entry names the donor and the relationship
`carried`	nobody characterised the flow; the number was carried from a neighbouring context by the convention	`source_flow_uuid` is the flow it came from; the rule is `data/context-carry-rules.json`'s row for this flow's context
`moved`	published on the wrong flow and moved to the one it is about	`data/lcia-misattributed-factors.json`
`refined`	a rounded printing replaced by the precise value its own source states	`data/lcia-rounded-printings.json`
(no factor)	a queue is asking, a wall stands, or the blank has no rule	below

Where the numbers are

On the flow's page in the review application, every implementation's number for every category, side by side, with the consensus row's derivation. This is the first place to look: most surprises are one implementation disagreeing with another, and the page shows it without a query.

In the database: lcia_characterization_factors, one row per (category, flow, place), joined to lcia_impact_categories for the method, the category and implemented_by. Only the consensus implementation's rows carry a derivation; a transcription arrives at nothing, it says what its publisher said.

As a file: lcia-factors.json.gz, every published factor of every implementation, following the build's own redirects so that a factor about a withdrawn flow is on its survivor (outputs).

Where the disagreements are

/factors/differences, or lcia-differences.json: every (flow, category, place) where two implementations state different numbers, banded by how far apart — within tolerance, 2×, 10×, 100×, more — and every one where one implementation characterises a substance in a context and skips the context beside it. The consensus row's derivation is on each difference, so a reader can see that a 397× gap was ruled and a 10× gap is still open.

Where the questions are

Three queues, each a work list of rows a person can answer, grouped by substance and category because that is the unit a decision is made in:

queue	asks	today
`contested-factor`	two implementations differ; which?	32 open, 6 ruled
`contradicted-factor`	the model underneath disagrees with both; publish anyway?	8 open
`proposed-factor`	only a transcription speaks and nothing vouches for it; publish?	16 open, 48 answered

A ruled row stays on its page marked ruled, with the verdict and the comment; the queue is a record as well as a work list.

Where the blanks are

/factors/coverage counts, per context, the flows this list has that no deciding implementation characterises while the same substance is characterised next door — 4,859 for EF 3.1 on the build of 1 September 2026, 3,093 of them in the unconfined aquifer. tools/count_context_blanks.py prints the same census per pair of contexts with what each publisher did; data/context-carry-rules.json says which of those pairs carry. The same page counts the blanks beside a substance — an ion with no factor where its element has one in the same compartment, 634 pairs over 25 ions for Stepwise 2006 on the build of 2 September 2026 before the ions were signed, 24 after — which no convention fills: those are the adoptions file's.

Where the decisions are

Every curated answer is a file under data/, registered in domain/rulings.py with a sentence saying what it decides, and every entry carries a mandatory comment. None of them is a value in a database. The files a factor can point to:

lcia-factor-rulings.json — a choice between stated numbers, pinned to them
lcia-factor-adoptions.json — a substance taking another's number, adopted or declined
lcia-factor-adoptions-stepwise-2006.json — the same for Stepwise 2006, where the recipient is stated by nobody and takes the donor's published number
lcia-misattributed-factors.json — a number moved to the flow it is about
lcia-rounded-printings.json — a precise printing preferred
context-carry-rules.json — the convention for the blanks, and the walls
lcia-underlying-model-factors.json — what USEtox says
lcia-substance-decisions.json — which flow of ours a GreenDelta row is about

And plans/lcia-consensus-decisions.md is where the shape of all of it is argued, with the measurements it rests on.

Report a problem with this page

Previous ← When nobody characterised the flow Next Overview →