Cannibalization: when to launch the thing that eats your best business
Apple retired the iPod on its own schedule. BlackBerry refused twice and went to 0.5% of the market. Sixty companies on record, and what separated the ones that committed from the ones that announced.
The question is almost never whether
By the time this reaches a room, the argument about whether the new thing will take revenue from the old one is over. It will. The live question is narrower and harder: does the old business get retired on your schedule or on somebody else's.
That distinction decides what you are actually choosing between. Retiring it yourself means you set the date, keep the customer, and carry the margin drop in a quarter you picked. Having it retired for you means a competitor sets the date, the customer moves first, and you find out from the share number. Both end in the same place. Only one of them is a decision.
What sixty companies did
The record holds 60 verified cannibalization decisions. Forty-two were never reversed. That is not the same as forty-two working, and the clearest proof is BlackBerry, which appears twice.
In 2008 BlackBerry chose to hedge with the Storm rather than fully build the touchscreen future, to protect the keyboard margin. In 2012 it refused to take BBM cross-platform to iPhone and Android, because the handset was the empire. Both decisions are recorded as held. Neither was undone. BlackBerry's global smartphone share fell for seventeen consecutive quarters to 0.5% by early 2014.
A decision holding means the company never changed its mind. It says nothing about whether the company was right.
The reversals share one shape
Eleven of the sixty were undone, and they do not look like failures of nerve at the moment of choice. They look like decisions that were never funded.
BP said in 2020 it would cannibalize oil and gas for a clean-energy pivot. By 2025 it had told the market its faith in the transition was misplaced, that it had gone too far too fast, and it was growing production to 2.3 to 2.5 million barrels a day rather than shrinking it. The record's own summary of the choice is: announce without committing.
Gap announced an Old Navy spinoff in January 2020 and canceled it within eleven months, citing the cost and complexity of splitting the company against softer performance. Alibaba announced six independently listable business groups in March 2023; by November the cloud spin-off was canceled and more than $20 billion of market value went with it, and the Cainiao IPO application was withdrawn the following March. Under Armour reached up for fashion and down for off-price at the same time, posted its first North American revenue decline after years of 20%-plus growth, and restructured back toward performance.
In each case the announcement was treated as the decision. Nothing irreversible happened in the first year, which meant there was nothing to stop the company walking back.
The ones that stuck spent something they could not get back
Apple shipped a product in 2007 that made its own roughly 40%-of-revenue product redundant. The iPod line was not killed by press release; it was allowed to decline and was finally retired in 2022, fifteen years later, on Apple's schedule.
Philip Morris International paid about $16 billion for Swedish Match in 2022 rather than build the smoke-free future internally, on the stated reasoning that it could not build it fast enough. ZYN U.S. shipments reached nearly 165 million cans in the fourth quarter of 2024, up nearly 42%, and 202 million cans the following quarter.
Disney built the streaming replacement for ESPN's affiliate business while the affiliate business was still paying. ESPN household penetration fell to 71 million in fiscal 2023 from 74 million the year before, and the direct-to-consumer businesses reached $143 million of operating income on $22.8 billion of revenue in 2024 — far thinner than the old toll, and in existence.
The pattern is not courage. It is that each of them moved capital, product or people in a way that could not be quietly undone the following quarter.
How to run this decision
Drawn from what separated the records that stuck from the ones that were undone. A step nobody in the record can be seen taking is not here.
- 01
Write down when the old business stops, in a date
Not a trend, a date. The year the revenue line crosses below the level that pays for the organization around it. If nobody will put a date on it, the decision is not ready to be made and the rest of these steps will produce theater.
- 02
Take the new thing away from the person who owns the old number
BlackBerry's refusal to unlock BBM is described in the record as leadership being unable to stomach loosening it from the handset, because the handset was the empire. That is not a character flaw, it is an org chart. Whoever is measured on the old margin will protect it, correctly, right up until it is gone.
- 03
Price the delay, not the move
The paper always favors waiting, because the cost of the new thing is immediate and specific and the cost of waiting is distant and vague. Force the comparison: what does one more year of the current position cost if the market moves as it is already moving? BlackBerry's seventeen consecutive quarters of share decline is what that number looks like when nobody calculates it.
- 04
Commit something irreversible in the first four quarters
Capital, a shipped product, an acquisition, a public retirement date. The eleven reversals in the record are distinguished from the forty-two that stuck by whether anything happened in the first year that could not be walked back. An announcement can always be walked back, and the record shows that it usually is.
- 05
Set the old product's retirement date before you launch the new one
Apple let the iPod run for fifteen years after the iPhone and retired it deliberately. The point of setting the date early is not to hit it; it is that the conversation about when the old thing ends happens once, at the start, rather than every quarter for a decade.
- 06
Name the number that would make you stop
Write it down before you start, with the date you will read it. A decision with no stopping condition is not a bet, it is a position, and positions in this record get defended long past the point where the evidence turned.
What it looks like when this is going wrong
- The plan is an announcement and no capital is attached to it in the first year.BP, 2020. Gap, 2020. Alibaba, 2023.
- The new business reports to the executive whose compensation depends on the old one.BlackBerry, 2012.
- You are reaching in two directions at once, up for premium and down for volume.Under Armour, 2016.
- The old business's decline is being described internally as cyclical or temporary.BlackBerry, 2008.
- The stated reason for not moving is complexity and cost of separation rather than a number.Gap, 2020.