Product Cannibalization: When to Launch the Product That Eats Yours

Why the cannibalization argument in the room is not the one that decides it

The room argues about whether to launch. The argument that decides it is about something else: whether the substitution is going to happen whatever you do. That is a question about your customers, and it has an answer outside your building.

Whether you should be the one supplying the substitute is a separate question, about your arithmetic. Companies answer the second and assume the first is theirs to decide. That is what makes the conversation feel like a test of nerve rather than a calculation.

The four positions, and only two are cannibalization decisions

Two of the four positions are not launch decisions at all, and they are the ones most often left with no plan. A company that concludes it should not launch usually records that conclusion and nothing else, and the declining business is then managed by neglect.

It is happening, and you can supply it

Apple is the case everyone cites, and its record is specific. In Apple's fiscal 2006 the iPod was about 40% of revenue, $7,676 million of $19,315 million. Apple announced the iPhone on January 9, 2007 and shipped it on June 29, 2007 with an iPod built in. Apple discontinued the last iPod, the iPod Touch, in 2022. Steve Jobs's rule was "if you don't cannibalize yourself, someone else will."

Meta faced the same position with TikTok and paid for it openly. Meta launched Instagram Reels in the U.S. and more than 50 countries on August 5, 2020, knowing the minutes would come partly out of Feed and Stories. In its third-quarter 2022 call Meta said Reels was "negative overall to revenue by about $500 million in the quarter." Reels reached about 20% of Instagram time spent in the first quarter of 2022 and more than half by January 2025. In April 2025 Zuckerberg still testified that users continue to engage more on TikTok than with his apps.

Sometimes the substitution is happening whatever you do, and the only choice is whose app the customers leave for. Disney launched Disney+ on November 12, 2019, while ESPN's affiliate business was already eroding. ESPN household penetration fell to 71 million in fiscal 2023 from 74 million the year before. Disney's streaming business made its first quarterly operating profit, $47 million, in the quarter ending March 2024, and combined operating income of $1.327 billion on $24.614 billion of revenue in fiscal 2025.

Comcast soft-launched Peacock on April 15, 2020, having decided it would rather lose customers to its own app than to someone else's. Peacock's revenue reached $4.9 billion in 2024, up from $3.4 billion, while its annual EBITDA loss narrowed to $1.79 billion from $2.75 billion. Peacock's cumulative losses from 2020 through the first quarter of 2026 exceed $11 billion.

Supplying it badly, or protecting the margin instead

The common failure in the second position is to supply the substitute on terms set by the old business. Intel sold its XScale mobile processor business to Marvell for $600 million in June 2006 to focus on x86. As TechCrunch put it in 2016, Intel "basically couldn't allow low-margin x86 products to cannibalize its high-margin x86 products." Intel later subsidized Atom tablet chips and shipped 46 million in 2014, but the effort hurt profitability and was not repeated for smartphones. Intel's share of the x86 market fell from a peak of about 89% around 2014 to 70.8% in the fourth quarter of 2025.

Xerox built the personal computer at PARC and shipped the Star in 1981, then priced it like a copier: a complete Star office system cost up to $100,000 and was leased to corporations. Xerox sold roughly 25,000 Stars by most counts, and some sources put the figure far lower.

Netflix was in the right position and nearly broke itself on the method. In September 2011 Netflix announced it would split DVD-by-mail into a separate company called Qwikster. Netflix lost about 800,000 U.S. subscribers in the third quarter of 2011, canceled Qwikster 23 days after announcing it, and saw its share price fall about 80% from its July peak. Netflix kept the DVD business inside the company and shipped its last DVD on September 29, 2023.

It is happening, and somebody else has it

When customers are switching and you cannot build the substitute in time, the move is not a launch. It is buying the capability, harvesting with dates on it, or selling while the decline is visible to you and not yet to a buyer.

Philip Morris International bought the substitute rather than build it. PMI's purchase of Swedish Match, the maker of ZYN, closed on November 25, 2022 and valued it at roughly US$16 billion. In the first quarter of 2025 PMI's U.S. ZYN shipments exceeded 200 million cans, up 53% on a year earlier.

Walmart bought a team. For years Walmart executives feared online would cannibalize the stores. In August 2016 Walmart agreed to buy Jet.com for about $3 billion in cash plus $300 million in shares. Walmart's U.S. e-commerce business became profitable for the first time in the quarter ended April 30, 2025, and in fiscal 2026 Walmart's global e-commerce sales grew 24% to $150.4 billion.

When it is not cannibalization at all, and the share it needs when it is

Some launches argued as cannibalization are not. Microsoft's shift to cloud under Satya Nadella from February 2014 is often told as self-disruption. In fiscal 2025 Microsoft's Windows OEM and Devices revenue rose 3%, Azure passed $75 billion in annual revenue, and total revenue reached $281.7 billion. Read closely, it was "a portfolio call, not a sacrifice."

Enphase designed its storage product so it could not cannibalize its microinverters: each IQ Battery 10 carries twelve embedded microinverters, and production shipments began in July 2020. In 2024 Enphase's microinverter units still fell about 58% and its revenue fell 42% to $1,330.4 million, while battery shipments rose 48%. The design protected the old product from the new one, not from the market.

Where the substitution is real and you can supply it, the decision reduces to one number: the capture share, the proportion of switching customers who must buy the substitute from you for launching to beat not launching. Two corrections change it more than anything else. The base case must decline, because the switching happens whether or not you launch. And the old fixed cost does not follow volume down; in most manufacturing and network businesses under half of it can be removed, over years.

Nothing in that says do not launch. It says the plan rests on producing a result the company has never produced, and that belief belongs in the paper as an assumption rather than inside a share number.

What these cases show

Each case is drawn from the Stratrix case library and checked against the filings and reporting listed under Sources. The last column says what each one does not establish.

Running the decision

What this method cannot tell you

It does not tell you whether your substitution is real. That is a question about your customers, answered outside the method, and it decides everything else.

This is also the decision most damaged by hindsight. Every documented case is told afterward as though the substitution was obvious at the time, and the companies that correctly declined to cannibalize a business nothing was threatening leave no record. None of the cases above sits in the fourth position, where no one can yet supply the substitute profitably, and that gap is part of the same bias. The cases support recognizing the positions. They cannot support a success rate, and this page does not claim one.

Questions