Every value enters the model with its evidence attached.
A figure is only ever as good as what it arrived with. Every value is anchored to a purchase order, attached to the document that produced it, and graded on the strength of that support before it is allowed anywhere near a calculation.
Sample values, to show the shape of the record.
“How do you know what a supplier sends is right?”
By what it arrives with. Most values come from documents the facility already keeps for its own accounting, collected inside your supply chain by analysts who know which record each process produces. A supplier’s unsupported word is the weakest grade we accept, and the record labels it that way.
Five grades, and every value carries one.
The grade is a property of the value, not of the supplier. The same mill can give you a metered reading for energy and an attestation for waste, and the record says so input by input.
- Grade 1 · Documented
Metered, invoiced or on a production record
Energy, water, yield and output taken from the paperwork the facility already produces for its own accounting and its own tax position. Nobody’s memory is involved, and the document stays attached to the value it produced.
- Grade 2 · Certified
Issued by somebody other than the supplier
Mill certificates, test reports and transaction certificates, held against the product and the order they belong to rather than in a shared drive. A third party put their name on these before we ever saw them.
- Grade 3 · Attested
Entered by the facility, named and dated
A value typed by the supplier in its own portal, with the person and the date on it. Weaker than a document and treated that way: never left as the only support behind a critical input.
- Grade 4 · Recorded on their behalf
For the partners who will never log in
Plenty of tier-3 and tier-4 facilities will not use software, ever. Our team records what they send by email, phone or paper, and the entry says that is what happened. An honest label beats a missing link, and beats a fabricated login even more.
- Grade 5 · Absent
Named and graded, never filled in
A missing input stays missing. It is carried as a gap graded critical, important or optional, and our analysts work with the facility to close it. Until then the record shows it as a gap.
Every step starts with a document.
Collection is designed around the two places data quality is actually lost: translation and transcription. Both are engineered out before anyone talks about coverage.
- 01
Anchor it to an order
Collection starts from a purchase order, so every value belongs to a product, a facility and a date. A number without an anchor cannot be checked later, whoever supplied it.
- 02
Ask in their language
Suppliers answer in their own language, and the platform ingests it. A mill guessing at an English form field is a data-quality problem long before it is a convenience one.
- 03
Read the document, don’t retype it
Uploaded invoices, certificates and reports are read into records by the AI. Manual transcription is where digits move; the file stays attached either way.
- 04
Approve before it saves
Nothing the AI reads becomes a record until a person agrees to it. That is the point where an extraction error is caught, and it is not optional.
- 05
Grade it, and name what is missing
Each value inherits its grade from what supports it, and the gaps are graded critical, important or optional. Our analysts then work through them with the facility.
The parts that keep the intake honest.
Each one is tagged with which half of the work it is: the software, our team, or the place the two meet.
- Document of recordPlatformThe file that produced a value stays attached to it, so the evidence is one click from the number
- Multilingual ingestionPlatformSuppliers answer in their own language rather than guessing at ours
- AI with human approvalBothDocuments read and turned into records, nothing saved until a person agrees
- Offline suppliersBothPartners who will never log in are recorded on their behalf, and the entry says so
- Evidence & gap trackingBothEvery missing input named and graded critical, important or optional
- Anonymous worker surveysPeopleSocial data collected without names and reported only in group form, so answering it is safe
- A dedicated analystPeopleOne named analyst embedded on your account, working with your suppliers on the evidence
A method published in the peer-reviewed literature.
The hardest inputs are never collected by software. They are collected by an analyst who knows which document a dye house actually keeps, and who has asked for it in 35+ manufacturing countries before.
We can only collect what a facility records, and some facilities do not meter. Where that is true, the result says so on its face. Collection also runs over seasons: the first season with a new supplier is usually the thinnest, and evidence grades rise in the seasons after it as the record is built inside the facility.
The other parts of the method.
Where the number came from, what it arrived with, who signed for it, and what keeps it true afterwards. Each has a page of its own.