A note from Itô
A limit is useful before it is tested. By writing down what we will not tolerate first, we make the difficult decision while the room is quiet—not when pressure is asking us to improvise.
The exercise begins with plain language. What would invalidate the research? What level of uncertainty is acceptable? Which operational conditions require us to pause? A useful limit must be observable, owned by someone, and connected to an action.
Numbers alone are not a risk process. A boundary without monitoring is a wish; monitoring without a predetermined response is a dashboard. The work is to connect what we measure to what we are prepared to do.
Writing the limits first narrows the space for rationalization. It gives a team a shared reference point when the evidence is uncomfortable and the easy answer is to wait.