You have a statutory target. Most of your evidence for it is a model.
Agricultural emissions reach a national inventory as activity data multiplied by an emission factor. Very little of it is measured on a farm. Proof adds the measured layer underneath, so a carbon budget, a net zero pathway and a scheme outcome can all be traced to records that exist in fields.
Detail at the bottom. Comparability at the top. Never both at once.
Agricultural evidence weakens as it climbs. By the time a figure reaches a national target it is standardised, delayed and stripped of the context that made it meaningful. Proof keeps the context attached all the way up.
A national dashboard can create the appearance of comparability without delivering a valid comparison. That is the risk Proof is built to remove.
The pathway you report. The pathway the fields support.
The dashed line is the required trajectory. The grey line is the modelled inventory position. The teal line is what measured farm records say, with the evidenced share stated rather than assumed. Change the target, the measure or the basis.
Emissions fell after the scheme. Now prove the scheme did it.
Participants against a matched comparison group, both drawn from the same evidence base. Change the intervention, the comparison basis or the period and watch how much of the movement survives.
Alternative explanations · tested, not waved away
Two carbon numbers, one target, and a boundary that quietly differs.
Carbon figures are only comparable when boundary, method, warming potential and baseline all match. Proof checks that before anything is placed on the same axis, and says so plainly when it cannot be.
Every published figure should answer these before it is published.
This is the flagship indicator from the demo programme, answered the way an official should be able to answer it at a select committee, without exposing a single farm.
The programme reports 1,412 cover-crop actions claimed. Nobody can say how many were established, how many survived to spring, or how many would have happened anyway.
You are accountable for an outcome and holding a count of activities.
Two delivery partners report the same indicator. One counts a practice as adopted on declaration, the other on evidence of establishment.
The better reporter looks like the better performer, and the funding follows.
The evaluation lands two years after the policy decision it was meant to inform, describing a scheme that has already been redesigned twice.
Evidence that arrives after the decision is a history, not an input.
Outcomes in months
Measured while you can still act
The demo programme reads real outcomes eight months after launch instead of three years, because the evidence is created in season by the people doing the work rather than collected afterwards by survey.
Paid on evidence
Claims cite records, not declarations
In the demo payment window, £2.3m was paid against farm-created dated evidence and nothing on self-declaration alone. Gaps are named per claim and visible to the claimant, so assurance replaces inspection lottery.
Publishable safely
Suppression that proves itself
No group under five records is ever published, on any cut, and the suppression register is itself public quarter by quarter. That register is what makes the figures defensible to statisticians and to the press.
Four institutional fears, and what changes about each.
Funding something ineffective for years. The outcome signal arrives in season, so a programme that is not working shows it in year one rather than at final evaluation.
Cancelling something effective because measurement was poor. The more expensive mistake, and the one nobody defends against. Measured outcomes with denominators protect good programmes from bad evidence.
Rewarding providers who report well rather than perform well. One definition, applied identically, removes the advantage of a well-written return.
Discovering a problem during audit or political challenge. Every figure already carries its method, denominator, exclusions and suppression register. There is nothing left to discover.
Proof never scores or ranks a farm, never pools regions whose definitions or methods differ, never publishes a group of fewer than five records on any cut, and never presents a modelled estimate as a measurement. Drill-down ends at governed sources and never reaches a private holding. Where attribution is uncertain, the uncertainty is published with the figure rather than resolved in its favour.
Proof draws no verdict. It holds the record.
“Does this replace our statistical service?”
No. It gives them a source layer that carries method, baseline and provenance with every value, so harmonised definitions can be applied rather than reconstructed. The analysis, the publication and the judgement stay exactly where they are.
“Why would farms and partners record to one standard?”
Because the same records answer their buyer, their bank and their insurer. Scheme reporting stops being a separate chore and becomes a by-product of evidence they already keep, which is the only version of harmonisation that has ever held.
“What coverage would a programme actually have?”
Less than full, and the honest answer matters more than a large number. On the call we will show real counts by region and action, and where coverage is too thin to support a claim, the demo withholds the figure in front of you rather than filling the gap.
Proof is working with a first group of programme and evaluation teams through the 2026 season. There are no testimonials on this page, because we would rather earn real ones than write our own. Early programmes help set the definitions their sector will inherit, which is the part that outlasts any single scheme.
Bring one target you are accountable for.
20 minutes. We will take the indicator behind it and walk the eight questions in order: what it means, how it was calculated, what it can honestly be compared with, and what a committee would find if they pulled the thread.