CEO succession: what separated the handoffs that held from the ones the board undid

Ten of sixteen held, and holding is not the same as working

Twenty-six companies in this corpus faced the question of who takes the seat after the person who defined it. Eleven of the seventeen settled decisions were never reversed, and nine more are too early to call — the highest share of any decision type here, because a board's choice of successor is reported the day it is made and its consequences are not.

HP shows why that number needs reading carefully. Its board ousted Léo Apotheker in September 2011 after the stock fell roughly 40% in his tenure and reached inward for one of its own directors. Three fired chief executives, $83 million in severance and an $8.8 billion write-down later, the same reflex produced the same move: in February 2026 Enrique Lores stepped down and the board again installed one of its own members as interim and opened another search. That record is classified as held. Nothing was reversed because the board kept doing the thing that was not working.

So the verdict column says whether a decision stuck. The question worth asking of the record is a narrower one: what did the boards whose choices stuck do differently from the three whose choices were undone?

A runway is a test, not a coronation

Costco's board appointed Craig Jelinek president and chief operating officer on 1 February 2010, put him on the board, and folded him into an Office of the President alongside the co-founder and the chairman. He became chief executive two years later. The structure existed, in the words of the record, to make one specific person the obvious next chief executive in public, with enough time to notice if it was not working. Revenue went from $99 billion in 2012 to $242.3 billion by 2023 and the warehouse count from 617 to 861.

Coca-Cola had a designated heir too. When Roberto Goizueta died, the board elected Douglas Ivester chairman and chief executive within five days, because he was president and chief operating officer and already doing the job, so the handover was treated as a formality. He announced his retirement roughly two and a half years later and relinquished the role in April 2000, with the stock falling sharply on the news. Being named the successor and being tested as the successor are different things, and Coca-Cola had done only the first.

The fast handoffs that held were fast because the work was already finished. McDonald's named Charlie Bell chief executive within hours of Jim Cantalupo's death in April 2004; Bell had effectively been running the company since a promotion sixteen months earlier. When Bell resigned seven months later, the board named Jim Skinner the same day and the turnaround program continued through both. Micron's bylaws auto-elevated Mark Durcan when Steve Appleton was killed in February 2012, and the board ratified it; Durcan abandoned a planned retirement to take it. Total replaced a chairman and chief executive killed in a runway crash within 36 hours, using a groomed insider and a former chairman held in reserve. None of these boards were quick. They were early.

The three reversals failed in three different ways

Coca-Cola's was an untested heir. Nike's was a different mistake: in October 2019 the board chose John Donahoe, a sitting director of five years with a technology résumé, to lead a digital pivot. He had run companies, and he had not run this one. In September 2024 the board reached back inside and un-retired Elliott Hill, a thirty-two-year insider, undoing the direction the 2019 choice had set.

Yahoo's failure was not about the person at all. The board hired Marissa Mayer in July 2012 as the most credible product visionary available, to transform a core business the record describes as an unwinnable job. She stayed to the end; Verizon bought the company for about $4.48 billion in June 2017, marked down $350 million after the breach disclosures. The board had asked who, when the open question was whether.

Run those three together and the diagnostic is short. Has this person done this job here, under observation, for long enough that the board could have changed its mind? Is the job one a chief executive can actually do? If the answer to either is no, the appointment is a bet the board will be asked to settle again.

Microsoft's 2014 choice is the useful counter-example, because the process was nothing like Costco's and it held anyway. The search ran about five months and reached outside — Ford's Alan Mulally, Qualcomm's Steve Mollenkopf, Ericsson's chief among six serious candidates — and the record's account of why it landed inside is not that the insider won the argument but that the board could not agree on an outsider and the outsiders withdrew. Satya Nadella had been at Microsoft twenty-two years. A board that cannot agree on a stranger ends up choosing the person it already knows, which is a weaker reason than Costco's and produced a comparable result. Process is the thing you can control, not the thing that guarantees the outcome.

The title and the control are two different handoffs

Larry Page and Sergey Brin gave Sundar Pichai the Alphabet chief executive title in December 2019 and the supervoting structure did not change. The Lee family paid the largest inheritance tax bill in Korean history — about 12 trillion won — by borrowing against the group rather than selling into it, and came out with a larger direct stake in Samsung Electronics, 1.67% against 0.70%, and in Samsung Life, 10.44% against 0.06%, followed by the executive chairman title in October 2022. In both cases the operating job moved and the control did not, which is a legitimate structure and an illegitimate thing to leave unstated.

A mandate can also run the other way, and Sony is the case for it. Kazuo Hirai was named president on 1 February 2012 with a television business heading for its eighth straight losing year, and he took personal charge of it rather than handing it to a division head, unwinding the Samsung LCD joint venture to strip roughly 50 billion yen of cost. The numbers that followed were not a rescue: a record 457 billion yen annual loss for the year ended March 2012, an eighth consecutive quarterly loss by early 2013, and a TV unit only nearing an annual profit by 2016 after more than $7 billion of losses. A successor who owns the worst problem personally is buying years, and the board should price them before it hands the job over.

The successions that held also tended to say out loud what the new person was not there to change. Jelinek's public commitment was to the $1.50 hot dog. Bell and Skinner carried on Cantalupo's plan without amendment. Apple's board named Tim Cook on Steve Jobs's written recommendation in August 2011 and the mandate was continuity; the stock rose roughly 2,000% against the S&P 500's 503%, services reached about $96 billion a year by fiscal 2024, and the company's share in China slipped out of the top tier over the same period. A handoff that held is not a handoff that solved everything, and the board that says which parts are fixed has given the successor a shorter list to be wrong about.

One more cost is worth pricing. GE ran a public multi-year horse race, a deliberate Jack Welch method, and it held. The two losing finalists left immediately for chief executive jobs elsewhere, and GE's stock fell more than 30% over Jeff Immelt's tenure, roughly $150 billion. A tournament identifies a winner by manufacturing departures.