Paul Nutt spent twenty years doing something almost nobody in management research does: he followed real decisions in real organizations from the moment they were made to the point, two years later, where you could tell whether they had stuck. Four hundred of them, in companies, hospitals and public agencies. His 1999 paper in the Academy of Management Executive has a title that still reads like a provocation: surprising but true, half the decisions in organizations fail.
I first read it a decade into running companies, and the honest reaction was that half sounded about right. What surprised me was not the rate. It was how consistently the successful half had done the same few things.
What "fail" means
Nutt's definition is stricter than "it went badly". A decision failed if it was not fully adopted within two years, and about a third of the four hundred were never used at all. These were not marginal calls. They were the decisions the organizations had chosen to make: new products, reorganizations, purchases, market entries. Half of them evaporated.
The failures were rarely about intelligence or information. Nutt kept finding the same pattern in how the decision had been made.
What the successful half did differently
Two things stood out across the whole sample.
The first was alternatives. In fewer than one decision in five had the decision-maker seriously considered more than one option. Most decisions were made by finding an idea that seemed to work and then gathering support for it, which Nutt called the idea-imposition process. When decision-makers did weigh several options, the success rate went from 56 percent to 70 percent. That is the single largest effect in the study, and it comes from a practice that costs an afternoon.
The second was how the decision was carried. Decision-makers who issued the decision as an edict, or who tried to persuade people of a conclusion already reached, saw far higher failure rates than those who involved the people who would have to live with it before the decision was fixed. Participation was the mechanism by which the decision found out what was wrong with it while there was still time.
Process beats analysis, six to one
A decade later, Dan Lovallo and Olivier Sibony at McKinsey looked at 1,048 strategic decisions, mostly investments and product launches, and asked what predicted their financial return. They separated two things that are usually lumped together: the quality of the analysis (the models, the market data, the sensitivity tables, the comparables) and the quality of the decision process (whether the team had explicitly explored the uncertainties, sought out views that contradicted the leader's, included people whose incentives cut against the proposal, and discussed the downside as seriously as the upside). The process explained six times as much of the variance in returns as the analysis did.
That number changed how I thought about my own leadership meetings. We had good analysis. What we did not have was a way to make the analyst's doubts audible once the CEO had signalled which way he was leaning. Lovallo and Sibony's data says that is the expensive gap, not the model.
The link to results
Bain's research on decision effectiveness (Blenko, Mankins and Rogers, 2010) surveyed more than 750 companies and found that the organizations that make and execute key decisions well are the ones with top-tier financial results, and the relationship held in every industry and country they looked at. McKinsey's 2019 survey of more than 1,200 managers puts a cost on the other side of the ledger: managers spend 37 percent of their time making decisions and judge more than half of that time wasted. For a Fortune 500 company that works out to about 530,000 lost working days a year. For a company of forty, scale it down, and it is still the biggest line item nobody has on a budget.
What this means on Monday
Three of Nutt's findings turn directly into practice for a company of your size.
Before you decide anything that matters, write down at least one real alternative, with its own case. Not a straw man. The thing you would do if the first idea were unavailable. Half the time the exercise confirms the first idea. The other half is where the 14 points of success rate come from.
Ask the people who will carry the decision for their read before it is fixed, and ask them in a way that does not tell them what you hope to hear. Nutt's participation finding and Lovallo and Sibony's contradiction finding are the same finding from two angles: the decision that hears its objections early survives.
And write the decision down as claims that could turn out to be false, so that two years later there is something to check against. Nutt could only count failures because he went back. Most companies never do.
Where Midfire fits
Midfire is built around Nutt's pattern. It asks for the paths before it drafts the conditions, so a big bet cannot be committed with one option on the table; it gathers each colleague's read independently, before the decision-maker's view is visible; and it books the look-back at commit, so the two-year question gets asked. The findings are Nutt's, Lovallo's and Sibony's. The product's job is to make the successful half's habits the path of least resistance.
The number I keep coming back to
Fifty-six to seventy. That is what seriously considering a second option did to the success rate in Nutt's sample, and it is the cheapest improvement in the whole literature. If you take one thing from four hundred decisions and twenty years of following them, take that.