Errata
What we got wrong
Anything a reader could have relied on and that turned out to be wrong: a figure, a coverage claim, a provenance line, a description of how something works. Listed rather than corrected silently, so anyone who copied the earlier version can find out it moved.
10 corrections so far. Entries are added, never edited away. How the figures are made is on the methodology page.
- TrueGaugecorrection 203fd0b
- What was wrong
- The prototype's process actions — cut a step's energy, cut or regrind its scrap, electrify it — were worked out over the whole product rather than the parts the process acts on, and priced scrap at the product's average material intensity. On a moulded enclosure with a steel pin and a cardboard carton, the energy action claimed about a quarter more than the conversion energy the same result reported.
- How it was found
- The first-saving figure disagreed with the plant-energy line on the same screen.
- What changed
- Each action now applies only to the lines its process acts on, the same rule the totals use, and prices scrap at those lines' own intensity. A test holds the two figures together.
- TrueGaugecorrection a063133
- What was wrong
- The matcher's documentation said every match was decided by either a wide margin or none, with no uncertain middle. Measured over a labelled set, the margins fall into four clusters and the middle is populated.
- How it was found
- The first run of the published accuracy benchmark. The earlier claim rested on examples chosen by the person who wrote the matcher.
- What changed
- The claim and the test that asserted it were corrected, and the accuracy is now published.
- CarbonSpeccorrection 3092569
- What was wrong
- The public worked example labelled its breakdown "By NRM element". The groups were CarbonSpec's design elements, including an Envelope group that NRM does not have — so a quantity surveyor setting it beside a cost plan would have been comparing different scopes.
- How it was found
- Building a reconciliation to the real NRM groups.
- What changed
- The label was removed, the two classifications are reconciled explicitly, and a test fails the build if the design grouping claims to be NRM again.
- TrueGaugecorrection 48ce2f8
- What was wrong
- The only steel factor is for steel cans, and it matched the bare word "steel" — so a 320 kg fabricated frame was priced as beverage cans and shown as a direct answer.
- How it was found
- A report that traced a wrong figure to the factor library rather than the reader.
- What changed
- A match now records whether the factor describes the same kind of article the client named, and a mismatch is shown as a proxy on both the question and the result.
- TrueGaugecorrection e45384a
- What was wrong
- In the signed-in application, a material matched to a UK government factor contributed nothing to the total rather than its figure, because those factors had no per-indicator rows in the shared database. An earlier note had put the effect at about a sixfold understatement of fuel and freight lines; the real effect was their omission.
- How it was found
- Reading the calculator rather than the factor table.
- What changed
- A global-warming row was written for each reachable UK factor at its full-cycle value, taking the set from 0 of 528 computable factors to 525.
- TrueGaugecorrection 4d04681
- What was wrong
- 61 non-polymer processes — composting, sand casting, fermentation, papermaking and others — carried a provenance line citing a polymer-processing benchmarking source, which did not describe them.
- How it was found
- Loading the process library into the shared reference database.
- What changed
- Those processes now carry the general wording the process library can support. The specific claim stays on the twelve operations it is true of.
- Sitecorrection 8cba1d3
- What was wrong
- The showcase claimed 102 processes against the 98 then modelled, and five selectable geographies against six.
- How it was found
- Checking the page's own claims against the running product.
- What changed
- The figures were corrected. Counts shown in prose are now read from the data wherever possible, so they cannot drift from it.
- Sitecorrection 2a6217d
- What was wrong
- The European data survey described the INIES database as undocumented. It is paid.
- How it was found
- Checking access terms before deciding whether it could be used.
- What changed
- The survey entry was corrected. INIES is not used.
- TrueGaugecorrection 6a6ebff
- What was wrong
- The geography picker showed the size of each published source where a reader expects coverage. The UK advertised 8,742 when it could price 35 materials — out by 250 times — and every other geography overstated in the same way.
- How it was found
- A test written for a smaller miscount in the French figure.
- What changed
- Coverage and source size are now separate fields, and a test asserts the coverage shown equals what the lookup can actually price.
- TrueGaugecorrection d2910a1
- What was wrong
- UK electricity was priced at 0.177 kg CO2e per kWh after the 2026 conversion factors had put generation at 0.13096 — about a quarter too high, which moved every electrification comparison. Separately, three polymer proxies were reasoned from chemistry rather than read: LDPE was assumed to sit above HDPE (it is 4% below), PVC was refused as a proxy on a difference that is not there (it is 5% below HDPE), and polypropylene was called marginally below HDPE (it is 16.7% below).
- How it was found
- Reading the full published DESNZ material set instead of a transcribed subset.
- What changed
- The library now reads 63 factors source-read from the 2026 edition, the grid figure was updated, and the proxy rules were removed. Only refusals for materials the set genuinely lacks remain.