Home / Science & Methodology
Science & Methodology

Data you can trace, methods you can defend.

An AI agent that touches a regulatory filing is only as trustworthy as the standards it's built on and the lineage behind every number it produces. This page is about both.

Standards we build to

Every calculation traces to a named standard — never a proprietary black box.

GHG Protocol — Corporate & Scope 3 ISO 14064-1/2/3 ISO 14040/14044 (LCA) NGFS transition scenarios IEA scenario pathways SSP scenario families SBTi target-setting methodology PCAF (financed emissions)

Where a methodology allows for judgement calls — a consolidation approach, a materiality threshold, an emission-factor substitution — the platform surfaces the choice and its rationale rather than making it silently.


How lineage works

Source record → calculation → disclosure line.

Every number that reaches a filing carries its lineage with it. Click through a figure in a report and you land on the calculation behind it; click through the calculation and you land on the source record it came from.

This isn't a feature added after the fact for compliance optics — it's the reason the unglamorous data-collection layer was built before anything that looks impressive in a demo.

SAMPLE LINEAGE
1Utility invoice, Site 04, MarchSOURCE
2Scope 2, location-based factorCALC
3IFRS S2, Scope 2 line itemFILED
Illustrative output — three clicks from filing to source

How risk modelling holds up to scrutiny

Built to clear internal model review, not just to look right in a demo.

Financial institutions and large enterprises increasingly put any model that feeds a lending, underwriting or capital decision through model risk management (MRM) review. Transition and physical risk output from the platform is built with that review in mind:

Traceable assumptions

Every scenario run states which pathway (NGFS, IEA, SSP), which vintage, and which parameters were used — nothing is a mystery input.

Documented, not proprietary-black-box

Calculation methodology is written down and shared with your risk, compliance or assurance team on request — not held back as trade secret.

Reproducible outputs

Re-running the same inputs against the same scenario produces the same output — a baseline requirement for any model going in front of a validation committee.

Human review before filing

Nothing generated by an agent reaches a regulatory filing or a credit decision without a named person on your team signing off first.


Emission factors & data quality

A continually maintained library, not a static import.

Activity-based, regional and supplier-specific emission factors, refreshed as source databases update, with support for your own licensed or custom factors. Every activity-to-factor mapping is reviewable and reused consistently across periods rather than silently re-guessed each cycle.

Ask your risk or assurance team to review the methodology.

We're glad to walk your internal model review, risk, or audit team through the calculation methods directly, before you touch real data.