Methodology
This is not a black box. Here is exactly how a value becomes a result, how we grade its exactness, and how we decide precision — so you can judge whether a figure is fit for your purpose.
Three separate layers
Exact units engine
Dimension-internal conversions with no material context — joules, watt-hours, calories, BTU, mass, volume, time. These run on fixed, testable factors and are exact (or standard-definition where the unit itself is a convention).
Context-dependent fuel engine
Conversions that only make sense with a material: litres of diesel to kilograms, m³ of gas to kWh, kg of hydrogen to energy. These need a density and/or heating value from data, carry assumptions, and are source-based or estimated.
Sources, assumptions & emissions engine
Every non-exact result explains which source and year it used, on what basis (LHV/HHV), for which scope and region, and with what uncertainty. CO₂ and CO₂e stay separate; missing data is marked "not available".
Exactness levels
Every result carries one of these labels. A result takes the least exact level of any input in its calculation path — exactness is a floor, set by the weakest link. These are the badges you see throughout the app.
- exact Follows from an SI or definitional identity — no material assumption.
- standard definition Fixed by a published standard or convention (toe, therm, IT cal/BTU).
- source-based A measured/tabulated value from a cited source; varies in reality.
- estimate A representative value with genuine spread — shown with a ~ marker.
- region + year Correct only for a stated region and year (e.g. grid electricity).
- your assumption Depends on a value you supplied or accepted (density, basis, condition).
- context required Answerable once you supply one more thing — a prompt, not an error.
- not available Not meaningful, not in scope, or impossible even in principle.
Precision & the ~ marker
Showing six significant figures for an estimate is itself an error — false precision. We compute internally at full precision and only round at display time, capped by exactness:
| Level | Max sig. figures | Marker |
|---|---|---|
| exact / standard definition | up to 6 (or the constant's own) | none |
| source-based | 3–4 | none |
| estimate | 2–3 | ~ |
| region + year | 2–3 | ~ + label |
The leading ~ means "this is a representative value with genuine spread" — it
appears only on estimates and region/year figures, never on exact identities. Where a
property genuinely spans a range, we show ~A–B rather than a false point value.
Heating value basis (LHV vs HHV)
Fuel energy depends on whether you count the heat released when water vapour condenses. The default is LHV/NCV (lower / net) — the dominant convention in international energy statistics — and every fuel-energy result is labeled with its basis. Where the data has an HHV/GCV value too, we show it alongside; we never derive one basis from the other with a generic factor. Read more →
What the tool refuses to do
- Convert power to energy without a duration (kW is not kWh).
- Give a grid-electricity CO₂e figure without a region and year.
- Turn CO₂ into CO₂e, or derive a missing heating-value basis.
- Invent a number when the data is missing — it says not available instead.
Refusing is a feature: a context required prompt means the question is answerable once you pin down one more thing, which is more useful than a confident wrong answer.
For the honest scope of what v0.1 answers exactly, estimates, and refuses — and what you must not use it for — see the limitations summary, and follow every result's sources back to the primary document.