Spreadsheet MRV vs dMRV: Why Excel Breaks Down at Audit Time

Most carbon projects start in spreadsheets, and many run in them for years. The problem is not that spreadsheets calculate badly, because they calculate fine. It is that verification does not test the calculation. It tests whether you can walk backward from the headline tonnage to the field measurement that produced it, and spreadsheets are structurally bad at preserving that path.

A verifier picks a number, for example, the biomass total for one stratum in one monitoring period, and asks how it was derived. Then they follow it down: which plots, which trees, which allometric equation, which wood density values, and which expansion factors.

At each step they are checking two things: that the step is methodologically correct, and that the inputs are the ones actually recorded in the field.

A spreadsheet can answer the first. It frequently cannot answer the second, because the link between a cell and its field origin is not stored anywhere.

Version divergence. `inventory_final_v3_REVISED.xlsx` exists on three laptops in three states. Which one produced the submitted number is unanswerable six months later.

Silent formula breakage. A row inserted outside a range, a fill-down that stopped early, a reference that shifted. Spreadsheets fail quietly, so the number still looks like a number.

Manual transcription. Field sheets typed in by hand. Every transcription is an error opportunity, and none of them is logged.

No change history. Who changed which value, when, and why. Standard practice in any audited system; absent by default in a workbook.

Lost provenance for constants. A wood density value sits in a cell with no record of which database it came from or which species it was matched to.

Findings at verification do not usually say "your carbon is wrong." They say "we cannot confirm this is right." The consequence is the same, resulting in a conservative deduction or rework.

Rework is the expensive outcome, because it can mean re-measuring plots. And because the finding arrives after the monitoring period, the project may no longer be able to reconstruct what it did.

The second cost is recurring. A crediting period involves many verification cycles, and a project that rebuilds its evidentiary argument from scratch each time pays that cost repeatedly.

The distinguishing feature is not software. It is that every derived value can name its inputs.

A field observation is recorded once, with its observer, timestamp and location, and is never retyped. Conversions reference a specific named equation version rather than an inlined formula. Constants reference their source. Aggregations are reproducible from the records rather than stored as values. And every change is attributable.

Projects that run this way answer a verifier's question by querying, not by reconstructing. That difference compounds across a crediting period.

For exploratory analysis, scenario modelling and checking someone else's arithmetic, a spreadsheet is the right tool and always will be.

The argument is narrower than "stop using Excel": it should not be the system of record for data that must survive independent audit over decades.

1. Verra VCS Validation and Verification requirements — https://verra.org/ (accessed 16 Sep 2026)

2. ICVCM Core Carbon Principles, traceability expectations — https://icvcm.org/core-carbon-principles/ (accessed 16 Sep 2026)

3. IPCC 2006 Guidelines Vol. 4 (AFOLU) — https://www.ipcc-nggip.iges.or.jp/public/2006gl/vol4.html (accessed 16 Sep 2026)