At a management meeting, we made a decision: stop shipping to any client with more than thirty days overdue. Owner — the commercial director. Deadline — immediate. We wrote it into the minutes, everyone nodded, everyone left. Yesterday I pulled the receivables report: since then it's up eighteen percent. I didn't go to the commercial director — I went to the operations clerk who actually generates the shipments. Twenty minutes in, I had three answers. First: her screen doesn't show client overdue status at all — accounting closes the books with a two-week lag. Second: the one time she did hold a shipment, a sales manager told her "ship it, I signed off" — and she shipped it, because arguing with him isn't her call to make. Third: that manager's bonus is calculated on shipments, not on payments received. The decision died on day four, at the first real conflict. Nobody ever came back to report it.
Notice what was actually in place: there was an owner. There was a decision. There were minutes. What wasn't there: the person at the point of execution had no data to see the trigger, no authority to hold the line against a verbal "I signed off," and an active incentive that paid her, in effect, to break the rule.
A decision on paper and a decision at work are two different objects. The paper version answers "what do we want." The working version answers "what does this specific person see on their screen, what are they authorized to do, and what's in it for them the moment a conflict shows up." Between those two objects sit five recurring breakpoints: data, authority, rules, process — and the decision itself, if it never specified what to do in the disputed case. Naming an owner closes none of these. It only determines who gets asked at the next meeting.
I see the same mechanism with clients across completely different industries. A construction company: the decision was "no crew goes on site without a signed estimate." Every single foreman went out without one — because an idle crew costs money every day, and signing off an estimate took a week. The rule collided with the economics of the process and predictably lost; it only started working once the sign-off cycle was cut to a day. An education company: the decision was "call every student who drops out within forty-eight hours." Owner named, script written. Except the dropout list was only generated once a week — the decision was unworkable the moment it was made, because the data physically arrived after its own deadline. A clinic: front-desk staff were required to confirm appointments the day before — but couldn't see changes to doctors' schedules, so they kept confirming slots that no longer existed.
Every one of these conversations started the same way: "people aren't executing, we need tighter control." And every one of them ended not with control, but with a specific point you could put your finger on: here the data arrives later than it's needed; here the person has no right to say no; here the current incentive system pays against the decision. The breakpoint is always specific. It takes days to find, not quarters — if what you're looking for is the point, not someone to blame.
That's probably the core distinction. The question "who didn't do it" produces explanations. The question "at what point could this not have been done" produces a result.
Minutes record intent. It becomes a decision at the point where one specific person, at one specific moment, has the data, the authority, and the motive all lined up. Everything else is correspondence.