Custom benchmarks & targets

private by default · reviewed before shared synthetic demo data

Build a benchmark

live · computed from 780 authorised records · groups under 5 always withheld

Measure

Crop

Season

Region

Min evidence standing

Compatibility

Group A · practice

Compare with · Group B

Open matching records

Method is attached to every figure: metric, denominator and inclusion rules travel with the benchmark. A saved definition is private until reviewed. It can never rank named farms, and no group under 5 is ever shown.

Definitions

one engine for every workspace — a target can never masquerade as a standard
DefinitionOwner · versionTargetStatus

Every definition retains: owner · source · version · effective dates · metric, unit and basis · method and system boundary · denominator · included and excluded records · limitations · status. Target value, pathway and deadline are optional fields.

Target tracking · actual vs pathway

gap-to-target per measure · coverage beside every status
target pathway actual · gap to pathway called out, not graded deadline baseline year shaded band = share of the base with supporting evidence — thin evidence can't masquerade as success
on pathway
status per measure · never composite
68%
evidence coverage behind the line
n=1,046
measured records in the denominator

Roll-ups never silently pool incompatible boundaries — the custom target can sit beside Proof, national and international references with the different methods kept visible (see benchmarks).

Private → reusable · the review gate

A definition starts private. It becomes a reusable Proof benchmark only after explicit mapping and review — metric, unit, basis, method and boundary aligned to canonical vocabularies, inclusions and limitations checked. Until then it can guide your own tracking but never appears to anyone else as an established standard.

Engine rules (identical in every workspace): coverage shown next to every status; no pooling across incompatible boundaries; no composite score of any kind; a status is never declared where the evidence is too thin; k=5 throughout. Each workspace adds only its own forbidden-score guardrail on top.