I ask boards this question often: what process governs your people decisions. Most cannot answer it. They can describe the interview stages. They can point to a scorecard nobody actually fills in properly. What they do not have is a process for people decisions with the same rigour as a capital allocation decision, a pricing decision or a market entry decision. That gap is where good executives fail without ever noticing it happening.
Peter Drucker named this gap in 1985, in a Harvard Business Review essay called "How to Make People Decisions." His argument still holds because the underlying mechanism has not changed. Human beings are worse at judging other human beings than they believe they are, and the cost of that misjudgement compounds for years, not months. Drucker used a baseball metaphor to make the number land. A .333 batting average is an excellent season in baseball. It is a disastrous result when it describes how often your hiring and promotion decisions turn out right.
Why You Need a Process for People Decisions
Executives resist this framing at first. Hiring feels personal, intuitive, a judgement you either have or you do not. Drucker's point was the opposite. The people who are best at this treat it as a discipline, not a gift. Read what he actually wrote, because most people who quote the .333 figure have never sat with the full sentence.
Sit with that number against your own decisions. If a third of significant hiring and promotion calls are outright failures, and you make even ten of them a year, three are going to end in a rehire, a resignation, a difficult exit, or a manager who stays for years while the team underneath them slowly disengages. A further third are only minimally effective. The seat gets filled. The box gets ticked. No capability compounds underneath it. That seat is occupied, not built.
This is not a talent problem. It is a process problem, and that distinction is the argument of this entire article. Executives who are excellent at reading a balance sheet, structuring a deal or fixing a stalled sales pipeline will still make people decisions on instinct: a good interview, a warm reference from someone they trust, a feeling about chemistry in the room. None of that is wrong exactly. It is simply not enough, on its own, to beat a one-in-three failure rate, and the number has not been meaningfully challenged in forty years because the underlying mechanism has not changed either.
There is a version of this argument executives accept in theory and reject in practice. They will nod at the .333 figure in a room, then walk straight back to a hiring process built around a single interview, a warm recommendation and a decision made in the car park afterward. The gap between what a leadership team believes about decision quality and what its actual process produces is often the widest gap in the whole organisation, wider than the gap in almost any other function, because nobody is measuring it. Finance gets audited. Sales gets a pipeline review. People decisions get a debrief that lasts ten minutes and rarely gets written down.
What Gut Feel Actually Costs You
Instinct is not useless. It is under-specified. A hiring manager's gut reaction draws on real pattern recognition, built from years of working with people. The problem is that the pattern recognition is trained on a small, biased sample, namely the people that manager has personally worked closely with, filtered through whatever made an impression on them in the room. That is a narrow training set for a decision this consequential. I wrote about what the managers who consistently get this right actually do differently in How the Best Managers Find and Develop Talent, and the short version is that they deliberately widen the sample instead of trusting the strongest impression in the room.
The cost of skipping a process shows up late, which is exactly why it survives so long unchallenged. A bad hiring decision rarely fails in the first month. It fails in month eight, when the gap between the role's real demands and the person's real capability has become impossible to ignore, and by then you have sunk cost, a team that has adjusted around the wrong person, and a rehire that costs multiples of the original salary. None of that shows up on the interview scorecard. It shows up eighteen months later, in your attrition numbers, your engagement scores and your own calendar, now full of one-to-ones that exist to manage a decision you made too quickly.
- The same two or three people make every significant hiring and promotion call, with no structured input from anyone who will actually work with the new hire.
- Interview notes exist, but nobody compares them against a written definition of what the role needs to produce in its first year.
- References are checked as a formality after the decision has already been made, not as evidence that could change it.
- There is no fixed point, ninety days or twelve months out, where you formally review whether the decision is turning out right.
- Promotion decisions are made faster than hiring decisions, on the assumption that internal familiarity is a substitute for evidence.
There is also a scale problem hiding in the .333 figure. In a small business, a bad people decision is painful but contained, one team feels it, one manager absorbs the fallout. In a scaling organisation, the same failure rate gets multiplied across every layer that is being built at once: new managers hiring their own reports, new functions standing up their own leadership, a structure trying to grow faster than anyone can properly vet the people filling it. Without a process, the failure rate does not stay steady as headcount grows. It compounds, because each bad decision is now hiring underneath it.
The Three Buckets Drucker Was Describing
It helps to name the three outcomes Drucker was pointing at, because most organisations only ever measure two of them properly. They track the failures, eventually, once someone has to be moved or managed out. They almost never measure the middle third at all.
- The Right Third: The person is not just competent, they are growing into more capability than the role strictly required. This is what a real process should be producing more of, and it is the only bucket most organisations actually notice.
- The Minimal Third: The role gets done. Nothing compounds. No bench strength is created underneath this person, and no one is being stretched. This bucket rarely gets flagged because nothing is visibly broken, which is exactly why it is so expensive over time.
- The Failure Third: A decision that eventually has to be unwound, usually within eighteen months, at a cost that is almost always higher than the org chart suggests once you count the rehire, the disruption and the manager hours spent managing the situation instead of the business.
The reason this matters now, not just in 1985, is that judgement itself is becoming a scarcer, more valuable resource inside organisations that are automating the routine parts of decision-making. I have argued elsewhere that judgement has to become infrastructure, something you deliberately build and protect, rather than something you simply hope certain people happen to have. People decisions are the clearest test of whether that infrastructure actually exists, because they are the decisions where a wrong call is hardest to reverse and slowest to show up.
The People Decision Protocol
You do not need to reinvent how you hire to do this. It requires treating the decision with the same discipline you would apply to any other decision above a certain size. This is the process I use with executive teams, built around five checkpoints rather than five interview rounds.
- Define the decision before you meet anyone: Write down, in one page, what the role actually needs to produce in its first twelve months, not a list of duties copied from the last job description. If you cannot describe the outcome, you cannot judge a candidate against it, and every interview afterward becomes an exercise in vibes.
- Insist on more than one qualified option: A decision made against a single acceptable candidate is not really a decision, it is a relief. Comparison is what forces you to be explicit about trade-offs instead of talking yourself into the person in front of you.
- Separate evidence from impression: Score candidates against demonstrated performance on comparable problems, gathered separately from how they performed in the room with you. The two signals are different, and the second one is far weaker than most executives assume.
- Take structured reference calls before the decision, not after: Ask what the person actually did on a specific project, not whether the referee liked working with them. A reference call taken after you have already decided is not due diligence, it is confirmation you were already looking for.
- Put a review date on the decision itself: Every significant hire or promotion gets a ninety day and a twelve month checkpoint on the calendar at the moment the decision is made, not when something starts to go wrong. This is what turns a one-in-three failure rate into information you can act on early, instead of a surprise you absorb late.
This is the same discipline behind good succession planning, which is really just a people decision made in advance of the vacancy instead of under the pressure of one. And it is the same discipline behind improving executive team decision making generally: the quality of a decision is a function of the process that produced it, not the seniority or confidence of the person who made the call.
The five checkpoints above are neither expensive nor slow to run. Together they add perhaps a week to a process that, for a senior hire, was already going to take a month. What they buy in exchange is evidence you can actually stand behind if the decision is questioned later, and a paper trail that tells you which part of the process broke down when a decision does end up in the failure third. Without that record, every bad hire becomes a mystery you solve from memory, months after the decision was made, with everyone involved already defending their own judgement rather than examining it.
The uncomfortable part of Drucker's number is not that executives get people decisions wrong. It is that the best executives in the world get them wrong at roughly the same rate as everyone else, because instinct does not scale and it does not improve much with seniority. A process does both. Build the process once, and you are no longer betting the next twelve months of a team's performance on how convincing someone was for forty-five minutes in a room.
If you want a structured way to test judgement and capability before you make the call, rather than after, that is exactly what a proper leadership assessment is built to do, and it is the kind of diagnostic infrastructure I build into Capability AI for the executive teams I work with.
This does not remove judgement from the decision. It gives judgement something to work with. A process does not choose the candidate for you, and it should not try to. What it does is force the evidence onto the table before the decision gets made, instead of letting a strong first impression stand in for evidence that was never gathered. That is the difference between an executive with good instincts and an executive who has simply never had a bad hire prove them wrong yet.