Decisions that come back

Two hundred and forty decisions across eighteen organisations were tracked for twelve months each, between January 2024 and June 2025, to test a simple question: how often does a decision that was made actually stay made? Records were read before anyone was interviewed, deliberately in that order, because retrospective accounts of decision-making are shaped by what happened afterwards — a decision that held gets remembered as clear, and the same decision, had it unravelled, would be remembered as rushed. In fourteen of the eighteen organisations, the record disagreed with the account given afterwards, usually about how many times something had actually been revisited.

The headline number is worse than the organisations themselves believed: 137 of 240 decisions returned for further discussion at least once within a year, against a median self-estimate of 14% who could recall it happening. The report tested several variables that seemed likely to explain the gap — whether the decision was formally minuted, the seniority of the decision-maker, whether the organisation had a RACI, meeting size — and none of them separated durable decisions from ones that came back. Having a RACI in particular made no measurable difference, which the report treats as its most useful negative finding, since RACIs are a documented governance practice organisations already invest in specifically to prevent this problem.

What did separate the outcomes was narrower and cheaper: whether the record stated what the decision's owner could do without asking again. Eighty-two of the 240 decisions carried a written scope, limit or category of exception the owner could approve alone; 17% of those returned. The other 158 carried no such record; 78% returned. The gap held within every kind of ownership the study coded — a named individual with written authority returned 11% of the time, the same individual without it returned 34% — a larger separation than ownership type alone produces. The report's plausible mechanism, offered rather than proven: a decision without stated authority has no defence against its first obstacle, so the owner takes it back to the group at the first sign of friction, and a group, being a group, discusses it rather than closing it.

A second, smaller finding concerns reversal criteria — a stated condition or date under which a decision should be revisited. Only 29 of 240 decisions recorded one, and just two of those returned, the cleanest single result in the study and, the report notes plainly, the smallest sample too. Where all four record features it identifies — a named owner, written authority, a stated escalation destination, and reversal criteria — were present together, decisions returned 5% of the time; where none were present, 95% did.

The practical takeaway the report lands on is not a governance programme. It's four specific lines that together cost about five minutes per decision to write: name a person, not a role or committee; state what they can do without asking again; state where the decision goes if they can't resolve it themselves; and state what would make the decision wrong. The report calls the last of these the highest-value two minutes available in the whole study, and the cheapest to test — on decisions an organisation already has on record, this afternoon.