By market share losses as the measure, Kraft Heinz is starting to work again. They narrowed from 90 basis points at the start of 2025 to 30 in the first half of this year. The company beat on Q2, raised its full-year outlook, and put another $100M behind its brands.
But the most vexing part of it all is the decision driving the recovery was made five months late, by accident, by a man the board hired to do the opposite. Could a simulation have predicted this?
The decision(s)
Three decisions, actually.
July 2015 - 3G Capital and Berkshire Hathaway close the Kraft-Heinz merger. Bernardo Hees runs it, Alex Behring chairs it, and zero-based budgeting runs everything else. $1.5B in promised synergies. 2,500 layoffs a month after close. Print double-sided. No more free Jell-O at headquarters.
September 2, 2025 - The board votes unanimously to take it apart with Global Taste Elevation Co. on one side and North American Grocery Co. on the other.
February 11, 2026 - Steve Cahillane, hired six weeks earlier for the express purpose of running that split, pauses it and redirects $600M into marketing, R&D, and product.
The cognitive composition
The 2015 room had one dominant vector: cost. 3G’s operating identity was buy scale, lever it, cut it, repeat. Genuinely excellent cognition, at least for a business whose problem is solely bloat. But Kraft Heinz’s problem was demand. And nobody seemingly at the table had the job of creating it.
You can read exactly where the missing seat was, because the 2019 writedown named it. $15.4B, concentrated on the Kraft and Oscar Mayer trademarks. Those are not random assets. Those are the two brands you impact when you stop feeding them. And when the cutting ran out of things to cut, the room did the only thing its cognition knew how to do: in February 2017 it bid $143B for Unilever. More scale to cut. Unilever said no in 48 hours.
Now the second room. By July 2022, 3G held zero board seats, down from three. By the fourth quarter of 2023 it had sold out entirely. The board that voted in 2025 to reverse the merger contained none of the people who built it. No memory, no accountability, nobody who had to defend the original thesis before dismantling it.
The dissent did not live inside that room. It lived on a phone call with CNBC. Buffett owned 27.5% of the company and said this, in one sentence: “It certainly didn’t turn out to be a brilliant idea to put them together, but I don’t think taking them apart will fix it.” He called the merger wrong and the un-merger wrong, objected to $300M in separation costs and to shareholders getting no vote, and the stock fell 7% that day. A unanimous board vote with your largest shareholder dissenting in public is not consensus but rather a room with no chair for an objection it already knew was coming.
Then the hire. The board recruited Cahillane whose signature credential was leading Kellogg through the Kellanova and WK Kellogg separation and pre-assigned him the CEO seat of one of the two future companies before he had spent a day inside the building. They hired the splitter to split.
The part everyone gets backwards
Six weeks in, the splitter looked at the business, said the problems were “fixable and within our control,” and stopped.
He appears to have been right. Share-loss trends improved, guidance went up, reinvestment now runs near $700M. Hold it loosely though as sales are still down 1.4% year over year, volume and mix are still a 2.6-point drag papered over by pricing, and Q2 carried another $7.4B non-cash impairment. But the reinvestment thesis is beating the separation thesis.
Which means the September 2025 vote was wrong. Unanimously wrong, at the board level, on the largest structural decision the company had available.
And nothing inside Kraft Heinz caught it. Not the board that voted for it. Not the process that produced it. Not the three months between the vote and the CEO announcement. It got caught because an outsider walked in with different eyes, before he had been socialized into the answer.
That is not governance. That is luck wearing a suit.
What a Decision Simulation would have surfaced
The stress-test for September 2025 was never “should we separate?” It was: “What does a room produce when it holds zero merger-architects, a publicly dissenting 27.5% owner, and pressure to show shareholders motion?” The predicted output is the one that shipped: the largest structural move available, announced before anyone was chosen to own it.
The hire is the cleanest demonstration of the fix. The board sequenced it backwards: picked the decision, then the person, and the decision only got tested when the person walked in the door. Run Cahillane’s cognition against the separation thesis before the offer letter and you get the February answer in November without burning a quarter, a CEO transition, and $300M of separation cost to get there.
The lesson
The right answer arrived. It just arrived by accident, and a board that cannot tell the difference between a good decision and a lucky one will keep running the same room and calling the result strategy.