Boards Vote on a Read They Cannot Check

Two hours a year on the call that cannot be delegated

Start with the number that should end the argument. A board spends, on average, two hours a year on CEO succession. Ram Charan calls candidate selection the one duty a board cannot hand off, and the same boards that would never approve a budget on two hours of work approve a successor on it. The result follows mechanically. Half of large companies cannot name a successor today, two of five new CEOs fail inside eighteen months, and poor succession is estimated to destroy close to a trillion dollars a year across the S&P 1500. This is what you get when the longest-tailed decision in corporate life is run on the least time and the worst information.

The dollar figures hide the mechanism, so be precise about the structure. The board's entire picture of the bench arrives through one pipe: the incumbent CEO, plus a few rehearsed presentations a year. Directors do not see what a candidate does when a number comes in wrong, who would walk the day a rival is named, or whether the slate is the real bench or the CEO's edit. The board is structurally the last party to know the thing it most needs to know, and then it votes.

What the canon actually measured

The cleanest demolition of director confidence comes from the survey program David Larcker and Stephen Miles ran out of Stanford^[1]. Their 2014 study asked directors how well they know senior management, then measured the exposure behind the answer. About 55 percent said they understood their executives' strengths and weaknesses well. Only 28 percent had spent unscripted time with those executives, without the CEO in the room, on a quarterly basis. The confidence is unbacked. A majority claim a strong read of people they have mostly seen perform. That is not knowledge. It is the memory of a presentation, and the presentation was the point.

The earlier Stanford survey, run with The Miles Group, sized the rest of the gap. Boards average two hours a year on succession^[2]; 39 percent of companies report zero viable internal candidates; 69 percent say the next CEO must be ready now while only 54 percent are grooming anyone. The board is asked to choose ready people it has not developed, from a bench it cannot see. Miles's prescription is diagnostic, not procedural^[10]: count the unscripted hours each director has spent with each candidate. Most cannot, and the count is the finding.

Charan, across Ending the CEO Succession Crisis^[3] and The Leadership Pipeline, names what the staged exposure hides. He argues succession fails at three points at once: development that never isolates the handful of true CEO candidates from thousands of high potentials, boards that do not put in the hours, and recruiters who recycle the same charismatic outsiders. The pipeline gives the diagnostic edge: the standard failure is a leader still doing the previous level's work while wearing the next level's title. A board reading polish cannot see the passage. It sees the title and the deck, and promotes the person the presentation built.

Noel Tichy sharpens the same point. His weighting is the line worth keeping: roughly 70 percent of real leadership development is crucible jobs, the rest coaching and training. A candidate who has never run something whole through a bad cycle is a paper candidate no matter how the deck reads. Claudio Fernandez-Araoz and colleagues put the cost near a trillion dollars a year^[4] in the S&P 1500, traced to three engines: ill-suited outside CEOs hired in a hurry, capital walking out after a botched choice, and unprepared internal promotions. Every engine is a read failure, a picture the board never checked against the people who work for the candidate.

The board the checklist could not save

If the surveys measure the gap, Jeffrey Sonnenfeld explains why structure alone never closes it. After Enron, WorldCom, and Tyco, he pointed out that the failed boards looked structurally perfect on every governance checklist. Structure did not predict failure; the social system did^[5]. A board is healthy or sick by its climate: whether directors challenge assumptions, whether management shares bad news on time, whether dissent is safe. This is the structural cause I keep returning to. The pathology is not a missing committee. It is a room that has confused collegiality with health, and a collegial room does not interrogate a polished slate.

The 2025 director survey shows the social system still failing in exactly his terms. Fifty-five percent of directors say at least one fellow director should be replaced and nothing happens^[6], and about three quarters skip individual director reviews entirely. A board that cannot have the hard conversation about itself will not have it about a candidate. The deference that lets a weak director sit lets a rehearsed slate pass, and the incumbent, reading the same room, is built every structural way to manage it.

Why the read never reaches the room

Pull the threads together and they converge on one mechanism, which is the thing I actually believe. The information that predicts whether a candidate can lead does not live in the talent system. It lives in the heads of the people who work for the candidate and under the incumbent, and every channel built to move it upward filters it out by design. A direct report knows exactly what a candidate does when a number comes in wrong. Saying so through any official route is career-ending, so the manager softens it, the function rolls it into a rating, the CHRO negotiates it into a slate, and by the boardroom it is a clean nine-box cell. Each step is locally rational. The sum is a board choosing on a fiction.

The nine-box is the artifact where the filtering hardens into method. The grid promises an objective read of performance against potential, and the critiques are well established: the placements encode the rater's bias, and the tidiness of the picture launders away the episode that carried the signal. A candidate who shoots the messenger lands in the same high-potential cell as one who absorbs bad news, because the grid records the boss's rating, not the messenger's experience. The board reads it as evidence. It is where the evidence went to be smoothed.

And the trap is self-sealing. The candidate who punishes the bearer of bad news produces, over a year or two, a team that has learned to bring only good news, and that team looks calm and aligned in every exposure the board sees. The behavior that should disqualify the candidate is the same behavior that erases its own evidence. A board reading the surface reads alignment; the reports read the silence underneath it, and the reports are usually right.

The same shape governs the slate itself. The list the board sees is rarely the CHRO's honest read; it is a negotiated artifact, and the negotiation lives only in the CHRO's memory. A name softened, an assessment briefed gentle, a stronger candidate who vanished between drafts. GE is the canonical demonstration^[7]: Welch ran a public horse race and the deferential board picked Immelt, a complement to himself rather than a match to where the business had to go. When Flannery was fired after roughly fourteen months, a 23 billion dollar writedown blindsided the board, the cleanest case of directors who did not know the state of their own businesses well enough to know their CEO was failing. The information diet was the disease. The famous names were symptoms.

Family transfer bends the curve hardest, and there the people who can see the next failure are the least likely to say so. Manfred Kets de Vries calls the founder's grip the retirement syndrome^[8]: the incumbent clings because exit means lost recognition, lost control, and a confrontation with mortality, and the markers show in slipped dates and hostility to the conversation. Of failed transitions, on the order of 60 percent fail on trust and communication breakdown^[11] and 25 percent on unprepared heirs. Both causes live in heads, at the dinner table, in old fairness wounds the charter records as a truce rather than a war. The board learns which one was operative only after the announcement it cannot unmake.

The emergency case removes even the illusion of time. Roughly 58 percent of boards have no emergency succession plan^[12], so the modal response to a CEO dying, falling ill, or being taken out by scandal is improvisation on a bench the board barely knows. Strategy& puts the price on the failure^[9]: a forced turnover costs a median large company about 1.8 billion dollars more in shareholder value than a planned one. The board that skipped the unscripted hours pays for them at the worst possible moment.

The instrument the literature keeps naming

The grim joke in all of it is that the truth is not missing. Every expert above, underneath the framework, is describing the same recovery move from a different angle. Charan wants the unscripted hours counted. Tichy wants the crucible found, not the title read. Sonnenfeld wants the social system diagnosed by climate rather than checklist. None of it is exotic. The reason boards do not have the read is not that it is unknowable. It is that the only channels they built for asking are the ones the organization learned long ago to perform into.

Which points, finally, at the kind of instrument this calls for. Not another nine-box, which records the rating and loses the episode. Not another rehearsed presentation, which delivers polish by design. Something closer to what the canon has described for twenty years: a neutral, confidential conversation with the people who work for and around each candidate, anchored to a specific recent moment rather than a rating on request, run for enough voices that no answer traces back to one person. That is the kind of instrument we are building at Latent Variables. The literature already told the board where the read sits. The open problem was only ever reaching it before the vote.