Before
Freeze the case
Definitions, lineage, model limits, expected outcome and the proposed move stay attached.
Completed decision proof
The evidence survives the recommendation.
In 90 seconds: the evidence available at the time, the proposed move, human approval, implementation and what was reported after the measurement window.
90 seconds · representative demo data
Reported outcomes remain distinct from causal estimates throughout.
During
Make approval accountable
The reviewer, note and implementation hand-off become part of the record.
After
Record, do not overclaim
Reported change is compared with expectation. Causal credit requires an identified model or experiment.
One guided 30-day cycle
Bring one recurring decision, real evidence and the finance approver.
We agree a written success criterion before starting. The goal is one completed, accountable customer loop—not dashboard adoption.