In the app
A guided setup, not a blank spreadsheet.
GHG inventory tooling for Taiwan SMEs
Most spreadsheet templates hand you a coefficient and ask you to trust it. Carbon Explorer shows the citation next to the number — the government gazette, the IPCC table, the dataset DOI — because the people who'll eventually audit your report are going to ask for exactly that.
Factor ledger
Factor Management groups every coefficient by where it came from, not by which form uses it. Turn a database off and every picker in the app stops offering it — but activity records already calculated from it keep their original number, because disabling a source should never rewrite history.
Coverage
Scope 1 and 2 are the base case. For Scope 3, the app covers eight of the GHG Protocol's fifteen categories — the ones an SME with no upstream investment portfolio or franchise network is actually likely to need. The rest simply aren't built yet, and the app doesn't pretend otherwise.
Combustion, refrigerant leakage, septic BOD/COD, and industrial process emissions — each with its own official factor table.
Taiwan's official "industrial" grid factor by default, with location-based grid factors for four more countries.
TWD spend ÷ annual USD rate × USEEIO cradle-to-gate factor, across 17 SME-relevant sectors.
Computers, office furniture, production equipment — same spend-based mechanism as Cat. 1.
Auto-calculated well-to-tank emissions from the Scope 1 fuel you already logged. Nothing to type in.
Freight, courier, and forwarding spend, same sector list as Cat. 9.
Weight × waste type × treatment method — landfill, incineration, recycling, or composting.
Sixteen transport modes; distance entered directly or as trips × average length.
Same transport-mode table as business travel, tagged by purpose instead of duplicated.
Shares its sector list and factors with Cat. 4 — the calculation doesn't care which direction the truck is going.
What we checked, and rejected
Every factor source in the ledger above was chosen after a rejected alternative. That process — not the calculator — is the actual work of building something an auditor will trust.
The plan was EXIOBASE for purchase-category factors. Three free sources were checked; the one usable table (ExioML factor_accounting) turned out to divide each sector's own emissions by its own output — no upstream supply chain included. Using it would have systematically undercounted every purchase. Switched to the EPA's USEEIO model instead, whose "with margins" column is already a full cradle-to-gate figure.
IGES publishes free national grid factors, but reading the actual file showed Operating Margin / Build Margin / Combined Margin columns and a "CDM crediting period" field — that's a baseline methodology for individual clean-development projects, not a national annual average. Ember's Global Electricity Review measures the thing a corporate inventory actually needs, so that's what shipped.
DEFRA's fuel well-to-tank factors are used as a worldwide proxy — burning diesel has roughly the same upstream footprint no matter which country sells it. Its UK electricity generation and transmission-loss factors were left out on purpose: those numbers are tied to the UK's specific power mix and grid losses, and would misrepresent any other country's electricity.
R-410A's GWP is widely quoted online as 2088. Recomputing it from the IPCC's own AR5 component values gives 1923.5 — 2088 is the older AR4 figure, uncredited, still circulating. The app's refrigerant table is AR5 throughout; mixing report versions inside one inventory is exactly the kind of error a reviewer should be able to catch by following the citation.
In the app
How it's built to be used
Everything lives in a local Room database. Nothing about a company's emissions data touches a network unless the user chooses to export it themselves.
The boundary table and totals — enough for a first look or an internal report — export for free. Line-item detail with full factor traceability is a one-time unlock, priced at the point people actually need an audit trail.
A single JSON file — activity records, equipment, site boundary, custom factors, preferences, even the floor plan photo — restorable on any device, saved wherever the user likes.
The xlsx round-trip is for bulk-editing activity data, not a backup — the app says so in the same screen, rather than let a misleading label cause a surprise later.
Under the hood
Domain use cases depend only on repository interfaces — swapping Room for something else, or adding a sync backend later, wouldn't touch a single calculation. It's also why 106 unit tests can exercise the entire calculation layer with in-memory fakes and no Android framework at all.