Separate participation from progress
Attendance and completed lessons describe activity. They may matter operationally, but they do not show whether a venture resolved an important risk. Ask what belief changed, what evidence supports that change, and which decision followed.
Keep comparable records
Use a shared vocabulary for assumptions, tests, observations, decisions, and next actions. Comparable structure does not require identical experiments. A hardware venture and a software venture can share a decision record while using very different validation methods.
Intervene at a named blocker
A useful intervention has a trigger, an owner, and a desired change. Stale evidence may require new customer access; delayed reviews may require mentor capacity. Avoid sending every founder to the same workshop when the underlying constraints differ.
Report limitations honestly
Describe missing evidence, unresolved questions, and selection bias alongside outcomes. Do not attribute every positive change to the program. Review latency and blocker resolution can guide program operations, but their meaning depends on the cohort context and the quality of underlying records.