What are you actually buying?
Products sold as dMRV platforms cluster into three groups that answer different questions, and the most common procurement mistake is comparing tools from different groups on a single feature list. A measurement tool and a project system can both be excellent and still not be alternatives to one another.
| Layer | The question it answers | Bought by |
|---|---|---|
| Measurement | Is this project measurable and monitorable from a distance? | Teams that need a scientific estimate of biomass or land-cover change |
| Diligence and investment | Is this project credible and worth financing? | Buyers and investors screening a portfolio of projects |
| Execution and evidence | Can this project be designed, evidenced, verified and issued? | Developers building and running the project itself |
Work out which question is in front of you before shortlisting anything. A developer who needs to reach issuance will find a pure measurement layer leaves most of the work undone; an investor screening fifty projects has no use for field data capture.
What should you ask a dMRV vendor?
The questions that separate platforms are rarely the ones on the feature comparison. These are the ones worth asking early, because the answers are expensive to discover late:
- How is a measurement traced back to the observation behind it, and can that still be done at a verification cycle eight years from now?
- Which registries and methodologies has the platform actually produced accepted documentation for, as opposed to listing as supported?
- What happens where there is no connectivity — is field capture genuinely offline-capable, or does it assume a signal?
- How is uncertainty reported: as a single headline accuracy number, or propagated through to the estimate a verifier will scrutinise?
- Who owns the data, and can you export the full evidence base if you leave?
- What does re-monitoring cost in cycle two, once the baseline exists — not what did onboarding cost?
How do you read an accuracy claim?
An accuracy claim means very little without the conditions attached to it. A single percentage figure quoted without context is a marketing number, not a measurement. What makes a claim checkable is the validation behind it: how many ground plots it was compared against, in which ecosystems, and whether the comparison was independent.
- Ask what the figure was validated against — ground plots, and how many, in which biomes.
- Ask whether uncertainty is reported per estimate or as one headline number across everything.
- Be wary of accuracy quoted without a stated ecosystem: a model tuned on temperate forest tells you little about a mangrove.
- Check whether below-ground and soil carbon are measured or simply excluded — this matters enormously for coastal projects.
The same scepticism is worth applying to us. Where we have not published a validated figure, we would rather say so than quote one that cannot be checked.
Where does Flora Carbon AI fit?
Flora Carbon AI sits at the execution and evidence layer: the work of taking a project from field observation through to registry-ready documentation. FloraScope handles land screening and land-use and land-cover mapping, FloraTrace captures per-tree ground evidence bound to GPS and time, and FloraGPT supports methodology selection, PDD drafting and audit preparation. Our work aligns with Gold Standard, Verra (VCS), Plan Vivo, ICR and CDM.
If what you need is a pure remote-sensing measurement layer, or portfolio screening for investment, a tool built for that layer will serve you better than we will. If you are building and running the project and need the evidence to hold up at verification, that is the problem we are built for. More detail is on what a dMRV platform does and carbon project development.