MRV, Explained for Engineers

A credit is a promise. MRV decides whether to believe it.

A carbon credit represents one ton of CO2 either removed from the atmosphere or kept out of it. You buy one, you've funded a ton's worth of climate benefit. Simple.

Except — how do you know the ton is real?

That question is the entire ballgame, and it has a name: MRV — Measurement, Reporting, and Verification. It's the system that turns "we planted some trees" or "this farm changed how it tills its soil" into a number a buyer will pay for. When MRV is strong, a credit is trustworthy. When it's weak, the credit is worth nothing — and worse, it quietly discredits every other credit in the market.

So carbon markets don't have a carbon problem. They have a trust problem. And once you look closely, the trust problem is an engineering problem.

Let me make the case the way I'd want it made to me — as a systems problem.

MRV is a pipeline you already recognize

Break the acronym apart and it stops being jargon:

• Measurement — turn a physical reality (a farm, a forest, a field of soil) into data.
• Reporting — structure that data into a standardized, defensible claim.
• Verification — have an independent third party confirm the claim is true.

Input, transform, audit. If you've built an ETL pipeline, you already understand the shape. The twist is the last stage: most pipelines you've built didn't end with an adversary reviewing every row, looking for reasons to reject it.

That single difference — an audit at the end, run by someone whose job is to be skeptical — changes how you have to build everything upstream.

The hard part is the engineering, not the chemistry

Here's what surprised me coming in as an engineer: the science of carbon is largely settled. We broadly know how much carbon a hectare of healthy soil or a stand of trees holds. The hard part is everything around the science.

You're measuring what you can't directly observe. You can't dig up and weigh the carbon in ten thousand hectares of soil. You sample a little, then model the rest. Every number carries uncertainty, and you have to quantify and defend that uncertainty — not hide it.

The data comes from the physical world, so it's messy. Sensors drop out. Satellite passes are cloudy. Smallholder farms — the ones that need this income most — often have no digital infrastructure at all. You're building on sparse, irregular, partially-missing inputs and still have to produce a defensible answer.

Every number has to be reproducible. When a verifier challenges a claim, "trust me" is not an answer. You need provenance: which raw inputs, which model version, which assumptions, on which date, produced this exact figure. That's data lineage and versioning treated as a first-class requirement, not a nice-to-have. If you've ever had to reproduce a result from six months ago, you know the discipline this demands at the source.

And it has to resist gaming. Carbon markets have been burned by inflated baselines, double-counted credits, and projects claiming benefit that would have happened anyway. A credible MRV system makes those failure modes hard by design — auditable trails, conservative defaults, no silent overwrites. You're not just building a pipeline; you're building one that holds up under adversarial pressure.

Put it together: it looks like data plumbing, but it's plumbing where every value must survive an audit. That's a real engineering problem, and a genuinely interesting one.

Why software is the bottleneck

If the science is understood, why isn't this solved?

Because the traditional way of doing MRV is manual — consultants, site visits, spreadsheets, PDFs. It's slow, expensive, and it doesn't scale to millions of smallholder farms. The cost of verifying a credit can rival the value of the credit itself. That's the bottleneck, and it isn't a science bottleneck. It's a software one.

Software is what makes verification cheap enough, fast enough, and consistent enough to work at scale. Remote sensing plus models replaces some of the site visits. A well-designed data layer makes an audit a query instead of an archaeology project. AI fills the gaps where ground data is thin.

This is the part I find motivating. The blocker to moving real money toward real climate outcomes isn't a mystery of nature — it's tooling that doesn't exist yet, or doesn't exist well enough. That's a builder's problem. It's why I'm spending this decade on it, and it's the problem HiTerra exists to solve.

The takeaway

MRV is the trust infrastructure of carbon markets. Strip away the climate vocabulary and what's underneath is a familiar engineering challenge: build a data pipeline whose every output is sparse-input-tolerant, reproducible, and able to survive an adversarial audit — at a cost low enough to scale.

That's an engineering problem hiding inside a climate problem. If you can build verifiable data systems, you can move this market — and the market, in turn, moves carbon.

I'll go a layer deeper in future posts — the actual stack, the model choices, the tradeoffs. For now, keep one thing: in carbon markets, the credibility of the whole system rests on the boring infrastructure underneath. Boring infrastructure is exactly what engineers are for.