Ten thousand downloads. How many fields changed?
You can show what was commissioned, what was published and who turned up. The hard part is the last mile: which farmer met the evidence, whether anything changed, whether it was done properly, and whether it worked. Proof holds that trail as records, and shows the gaps as gaps.
Strong at the start. Thin where it counts.
Every organisation in this position sits in the middle of the same chain. Visibility is excellent for the first four steps and drops away for the rest, which is exactly where the value is.
Everything above the line is activity. Everything below it is impact. Most annual reporting stops at the line, and everyone in the room knows it.
Follow one recommendation from publication to the field.
Pick a recommendation, a group of holdings and a run of seasons. Each step carries its own count and its own denominator. The step where the trail breaks is marked, because that is the step worth funding.
Not what the study concluded. Whether it works on a farm like mine.
The same recommendation behaves differently on different ground. Describe the farm and the record answers honestly, including the descriptions where too few holdings match to say anything at all.
The questions coming back off the farm, and whether anyone is working on them.
These are not survey responses. They are counted from records: questions asked and left unresolved, practices quietly dropped, and recommendations resting on thinner evidence than they appear to.
What a levy payer asks, answered without a single farm being named.
One recommendation, taken end to end, in the form a board or a levy payer can actually check.
Every figure carries its denominator. Any group with fewer than five records is withheld before anyone sees it, including from you. Synthetic demo data.
A trial answers the question under the conditions it was run in. The farmer asks whether it holds on their ground, in their season, with their kit.
Good research gets rejected for the wrong reason, and nobody records why.
Research reports, trial data, benchmarking, market work and knowledge exchange sit in separate systems that were never meant to talk.
Nobody can say what is already known, where the evidence conflicts, or which two programmes are answering the same question.
Benchmarking depends on farmers typing in detailed figures, which means partial participation, self-selection and comparison groups too small to trust.
The arithmetic is right and the comparison is still wrong for that farm.
A farmer attends nothing and adopts the recommendation through their agronomist. Another attends everything and changes nothing. A third adopts it two years later.
Attendance and downloads count all three the same way.
One record, carried the whole way down the chain.
The record starts on the farm
What was applied, drilled, cut or measured, written down when it happened with the conditions attached. Not recalled at the end of the season for a form.
Your guidance is a record too
The recommendation, the conditions it applies under and the research behind it sit in the same structure, so a decision can point back at it.
Comparison happens on like for like
Soil, rainfall, system and scale are held with every figure. Where the match is poor, Proof says so rather than averaging past it.
The gaps stay visible
Where the trail breaks, it is shown as broken. Absence is reported as absence, never filled in with an assumption.
Three commitments the farmer can read as easily as you can.
Proof does not sell identified farm records. Organisations pay for the aggregated, anonymised picture, and the farmer can see exactly what that picture contains.
Why farmers take partBelow five records in any group, nothing is published. Not to you, not to a member, not in an export. Fail closed, every time.
The withholding ruleProof produces no score, no ranking and no verdict on any farm, any adviser or any recommendation. It holds the record. You draw the conclusion.
What Proof will not doBring one recommendation you already publish.
Twenty minutes. We will take it end to end: who met it, what changed, whether it was done properly, what happened next, and the conditions where nobody has enough evidence to say. Including the parts that do not flatter it.