Innovation by Spin-Out: The Sister-Company Play

Everything in this series so far has happened inside the walls. Ambidextrous units, labs, Kickbox grants, growth boards — all of it is machinery for protecting a new bet from the parent company’s immune system while keeping it inside the parent company. This post is about the moment that stops working: when the conflict between the new business and the core isn’t a process problem you can wall off, but a values problem you can’t. When that’s true, the answer isn’t a better org chart. It’s a different company — separate P&L, separate brand, separate leadership, and sometimes a separate cap table.

Christensen himself said this plainly, and it’s the part of his prescription people quote least. The RPV framework from part one sorts an organization’s capabilities into resources, processes, and values, and his 2000 HBR piece with Michael Overdorf turns that into a placement chart: if the innovation fits your processes but not your values, or neither, don’t run it inside — spin it out, because the resource-allocation machinery will starve it no matter how sincere the sponsorship. The disk-drive study’s clearest positive finding was the same point from the other direction: the rare incumbents that survived a disruptive transition were almost all ones that set up an autonomous organization, with its own economics, to chase the new market. Separation wasn’t a nice-to-have in the data. It was close to the whole treatment effect.

When does an ambidextrous unit stop being enough?

The ambidexterity evidence from part two says a separate unit integrated at the executive level beats both embedding and full isolation, and by a wide margin. So why would you ever go further and leave the building? Because the O’Reilly–Tushman design has a load-bearing assumption hiding in it: that executive integration is an asset. The shared senior sponsor exists so the new unit can borrow the parent’s genuinely useful resources — technology, brand, distribution, customer access — while being shielded from its processes.

That assumption holds when the fights between core and new unit are about processes: stage gates, budgeting cadence, hiring bands, brand review. Those, a strong sponsor can wave away. It breaks when the fights are about values in Christensen’s sense — the criteria by which the whole organization decides what’s worth doing. Three of them recur so predictably that I treat them as the decision line:

Margin dilution. If the new business’s steady-state gross margins are structurally below the corporate threshold, every quarterly review inside the parent re-litigates its right to exist. You can defend a pre-revenue experiment with innovation accounting; you cannot defend a scaled business that permanently drags the blended margin. Kodak’s digital business had exactly this problem, and no internal structure fixes arithmetic.

Channel conflict. If the new business sells direct while the core lives or dies by retail and distributor relationships — or undercuts the core’s flagship pricing — then the parent’s own salesforce and account managers are structurally its enemies. A sponsor can protect a team from a process. Nobody can protect it from ten thousand commissioned colleagues whose customers are angry.

Talent economics. If the venture needs people who price themselves in equity, an internal unit can’t pay them. Corporate comp bands are a value, not a process: they encode what the institution believes a person’s contribution can be worth. You can’t give an internal team founder-grade upside without detonating the pay structure around it — which is precisely why Cisco’s workaround, below, was both brilliant and corrosive.

If none of these three apply — the fights are about cadence and control, and the parent’s channel and technology genuinely help — stay inside and build the ambidextrous unit. If two or three apply, integration isn’t your asset. It’s your anchor.

Nespresso: the walled-off sister nobody talks about correctly

The pop telling of Nespresso is a triumph-of-innovation story: Nestlé invents the capsule, George Clooney appears, billions follow. The actual history is a twenty-year lesson in why some businesses survive only outside the parent’s walls.

The technology came out of Nestlé’s own labs — Eric Favre’s capsule system, patented in 1976 — and then spent a decade nearly dying inside the company, because everything about it violated Nestlé’s values. Nestlé sold coffee as a mass grocery product: Nescafé in jars, moved through supermarkets at supermarket margins, marketed to everyone. Nespresso proposed premium coffee at a multiple of the per-cup price, sold with a machine, direct to consumers. The internal antibodies read it, correctly by their own lights, as a rounding error with channel-conflict risk.

What saved it was exile. In 1986 Nestlé launched Nespresso as a separate company — its own legal entity, its own offices away from the Vevey headquarters, eventually its own CEO — explicitly so the ordinary Nestlé machinery couldn’t reach it. The first business model (office coffee systems, sold through a partner) flopped; the unit came within a board meeting of being shut. Jean-Paul Gaillard, brought in around 1988, made the moves an internal unit could never have made: pivot from offices to affluent households, and sell through a direct mail-order “Club Nespresso” — bypassing Nestlé’s retail channel entirely, because that channel’s economics and relationships were the threat. Later came the boutiques, again a channel Nestlé the grocery company had no processes for and no values that endorsed.

And then — the part every impatient executive should sit with — it took roughly two decades. Nespresso didn’t become a meaningful pillar of Nestlé’s results until the 2000s; the billion-franc milestones and the double-digit growth streak the business became famous for arrived some twenty years after the 1986 founding. No internal unit survives twenty years of sub-threshold performance inside a public company’s planning cycle. A sister company with its own P&L, reporting scoreboard, and a thin executive tether can. That’s the whole argument for the structure, compressed into one case.

Alphabet: spin-out as an operating system, with an honest scoreboard

If Nespresso is spin-out as a one-off act of protection, Alphabet is the attempt to institutionalize it. The 2015 restructuring turned Google’s speculative portfolio into literal sister companies: Google became one subsidiary of a new holding company, and the “Other Bets” — Verily, Waymo, Wing, and the rest — became separately reported entities with their own CEOs, their own finances, and Alphabet as a shareholder rather than a boss. X, the moonshot factory, formalized “graduation” as the pipeline: Verily left in 2015, Waymo in 2016, each becoming a standalone company rather than a Google division.

The mechanism is worth naming precisely, because it’s the values logic again. A self-driving program inside Google’s P&L is a cost center competing with search advertising for headcount and patience — the exact starvation dynamic from part one. Waymo the sister company has its own board, its own comp structure with venture-style equity, and — this is the underrated move — outside investors. Alphabet raised external rounds for Waymo and Verily from firms like Silver Lake and Andreessen Horowitz not because it needed the cash, but because an external price is a discipline no internal review can fake. Someone with no loyalty to the strategy priced the bet.

Now the honest scoreboard, because this series doesn’t do hagiography. Other Bets has lost billions of dollars essentially every year since the segment was created — the cumulative operating losses across a decade run well past $30 billion — against revenue that remains a rounding error on Alphabet’s statements. Several bets died or were reabsorbed (Loon shut down in 2021; Nest was folded back into Google in 2018, a graduation formally reversed). Waymo, founded as a Google project in 2009, needed more than fifteen years to reach the point of scaling a real paid robotaxi service, and its economics remain unproven at the scale that would justify the spend. The structure demonstrably keeps long-horizon bets alive in ways no internal unit could. Whether it makes them pay is, on the public evidence so far, still an open question — with Waymo as the one bet that might settle it. Spin-out solves the survival problem. It does not repeal the base rates of new ventures.

Cisco’s spin-in: the trick that worked three times and poisoned the well

The strangest variant on record is Cisco’s “spin-in,” and it inverts the whole play: instead of pushing an internal team out, you fund an external startup — staffed by your own best people — with a pre-agreed acquisition waiting at the end.

Cisco ran it three times with essentially the same trio of star engineers — Mario Mazzola, Prem Jain, Luca Cafiero. Andiamo Systems built storage-networking switches and was bought back in a deal announced in 2002 at up to roughly $750 million. Nuova Systems built what became the Nexus 5000 line and the UCS server business — arguably the most consequential product Cisco shipped that decade — and was bought back in 2008 for several hundred million more. Insieme Networks built the Nexus 9000 and Cisco’s software-defined networking answer, acquired in 2013 for up to $863 million. As machinery, it’s elegant: the team gets a real startup — separate entity, separate equity, founder-grade upside if they hit the milestones — without leaving the mothership’s orbit, and Cisco gets a guaranteed window on the output, having pre-negotiated the exit price. It solves the talent-economics problem head-on: you can pay people like founders, if they’re founders.

Why did Cisco retire it? Not because the products failed — they didn’t. Because the structure answered the talent-values conflict by creating a fairness conflict. Engineers inside Cisco watched colleagues walk across the street, work on Cisco-adjacent problems with Cisco’s money, and come back tens of millions of dollars richer, while the people who kept the core running got refresh grants. The resentment was an open secret in the Valley, and it compounded with each round. The kicker is what happened after the model ended: the same team left for real in 2017, founded Pensando without a pre-agreed Cisco exit, and sold it to AMD in 2022 for $1.9 billion. The spin-in’s lesson cuts both ways — the venture upside was real and the parent could harness it, but a mechanism that mints founders from favorites inside a company of salaried peers spends cultural capital every time you run it.

The cautionary pattern: the digital attacker that came home

One failure family deserves a paragraph, because it’s the most common way this play is botched today: the corporate “digital attacker.” A bank or insurer, told that it can’t disrupt itself from within, launches a separate digital brand — new name, new app, new office — to attack its own market. JPMorgan’s Finn lasted about a year before being folded back into Chase in 2019; NatWest’s Bó closed within months in 2020; the pattern repeats across insurance and telecom with less famous names. The autopsy is usually the same and it’s diagnostic: these weren’t spin-outs, they were costumes. Same balance sheet, same risk committees, same executives one reorg away, no independent governance, no outside capital, no equity, and — fatally — no answer to why a customer should prefer the attacker over either the parent or a genuine startup. Separation of brand without separation of values is the showroom problem from the labs post wearing a fintech hoodie. The org chart moved; the resource-allocation process didn’t.

What a real spin-out requires

Strip the cases down and a real spin-out has five properties. Miss two and you’ve built a costume.

Independent governance. Its own board, with members who don’t report into the parent’s hierarchy — ideally including outsiders whose reputations ride on the venture, not on the parent’s comfort. If the parent’s CFO can reach into the venture’s budget between board meetings, it isn’t separate.

Founder-grade equity. Real ownership in the entity for the operating team, vesting against the venture’s outcomes, not the parent’s stock. This is the point of the structure, not a perk: it’s how you hire people the comp bands can’t, and how you make the venture’s survival somebody’s personal core business — the only durable fix part one identified.

Arm’s-length commercial contracts. Whatever the venture buys from or sells to the parent — technology licenses, data, distribution — gets papered at market terms, as if between strangers. Cozy internal transfer pricing feels generous and is actually a leash: it makes the venture’s economics fictional and gives the parent a valve to quietly strangle it later.

A declared endgame. IPO, sale back to the parent at a formula or market price, or independence — decided and written down at formation. Cisco’s spin-ins worked mechanically because the exit was pre-agreed; Alphabet’s bets drift partly because “graduate to… what, exactly?” was never fully answered. An undeclared endgame defaults, under pressure, to reabsorption on the parent’s terms.

Escape velocity from the parent-as-first-customer trap. The parent as anchor customer is the most seductive asset a spin-out has and the most reliable killer. It fills the pipeline, flatters the revenue line, and quietly reshapes the product around one atypical, politically loud buyer — until the venture is a captive supplier with a startup’s cost of capital and none of a startup’s market discipline. Take the parent’s revenue if it’s there, but cap it: set a date and a percentage by which most revenue must come from strangers, and treat missing it as the kill signal it is. Nespresso’s entire channel strategy was, in effect, a refusal of this trap — it built its own route to customers precisely because borrowing the parent’s would have meant inheriting the parent’s values.

When you can’t build it or spin it out

The spin-out is the last structural move available for a bet you intend to own. But some futures you can’t build inside and can’t credibly staff outside either — the capability, the founders, and the insight all live in someone else’s startup, and no amount of separation engineering conjures them. For those, incumbents reach for a different instrument entirely: put money into other people’s ventures and buy a window seat on the future instead of a deed to it. That’s corporate venture capital — a tool with a genuinely terrible pop reputation and a more interesting real record — and it’s where this series goes next.

Further reading

• Christensen & Overdorf, “Meeting the Challenge of Disruptive Change” (HBR, 2000) — the RPV fit chart that tells you when to spin out rather than incubate.
• Kashani & Miller, “Innovation and Renovation: The Nespresso Story” (IMD case) — the standard account of the exile, the near-death, and the Gaillard pivot.
• Alphabet’s 10-K filings, segment reporting — the primary source on Other Bets economics; read the numbers, not the keynotes.
• “Cisco’s ‘spin-in’ strategy” coverage in Fortune (2013–2016) on Insieme and the MPLS team — the mechanics and the resentment, both on the record.