Our experiments dashboard is full of results and empty of decisions. Six months later nobody remembers why we stopped the pricing test. Is anyone using an agent to write the decision record at the end of each experiment, and where does that record live?
We use the weekly report agent to draft a decision entry every time an experiment changes state. The record lives in the same doc as the definitions, so the decision and the metric definition it depended on are never separated.
The important field is "what we would need to see to reopen this". Without it every stopped test gets re-proposed by a new hire.
Adding a reopen condition to the template today.