Bijuraj Sarangi

Every board eventually asks the same question about a struggling or complex part of its portfolio: what do we need to change to make this work?

As potential answers, both structural separation and transformation get treated as points on the same spectrum, as if separation is simply what a board does when transformation has failed for long enough. Interestingly, they aren’t answers to the same problem. Transformation assumes the businesses belong together and the job is to run them better. Separation questions the premise itself, that these businesses were ever one business to begin with, rather than a reporting convenience that has outlived its usefulness.

The question a board actually needs to answer is this: when should boards decide to structurally separate a business for better returns, and when should they not?

Once you get that call right, a company can unlock more enterprise value in a single transaction than years of operational improvement would have produced. If you get it wrong, a board may end up spending a transformation budget trying to fix a structural problem no amount of operational work was ever going to solve, or worse, break apart a business whose combined value depended on staying together in the first place.

The public record of the last five years is now large enough to answer this properly with disclosed financial outcomes a board could stress-test today. Some of the largest breakups in recent corporate history prove that separation, done for the right reason, creates value transformation alone cannot reach. Just as tellingly, some of the most closely watched companies in the market considered the exact same move and chose not to make it, and were right not to.

What separates the two isn’t size, industry, or how bold the board is willing to be. It’s a much narrower, more testable question: are these businesses actually correlated in a way that compounds when kept together, or only bundled in a way that confuses the market about what each part is worth?

The case for separation, in three public examples

These are the largest, most scrutinized corporate separations of the past five years, each with public financial disclosures a board could stress-test easily.

General Electric: three-way breakup, 2021–2024

For two decades, GE traded as a diversified industrial conglomerate with businesses across aviation, healthcare, energy and financial services, bundled into a single equity story that analysts increasingly found impossible to model. In 2021, under CEO Larry Culp, the board committed to splitting the company into three standalone businesses: GE Aerospace, GE HealthCare, and GE Vernova.

The logic was not related to operational improvement, since each business was already reasonably well run. But, it was about a diversified holding company obscuring capital allocation, blending distinct investor bases into one confused one, and preventing each business from being valued on its own growth and margin profile. Once separated, each entity could pursue a capital structure, growth mandate, and management incentive plan suited to its own industry, instead of one compromise structure serving three unrelated end markets.

Results: Combined market cap rose from roughly $290bn pre-separation to $474bn post-separation, an implied uplift of about 64%. Total shareholder return since the breakup began has been approximately 400%.

Kellogg Company: split into Kellanova & WK Kellogg Co, 2023

Kellogg’s separation is the more instructive case, precisely because it wasn’t as dramatic a headline as GE was. The company split its high-growth global snacking business (Kellanova) from its mature, structurally challenged North American cereal business (WK Kellogg Co), a business facing declining volumes and requiring a costly manufacturing reset.

To improve its health, WK Kellogg still had to close a plant, cut roughly 17% of its workforce, and take real charges across it’s spend areas. The separation didn’t avoid that hard operational work; it made the work legible. Once the two businesses were standalone and cleanly valued, each became an easier, more obvious acquisition target than it ever was buried inside a single conglomerate.

Results: Kellanova was acquired by Mars in a deal valuing it near $36bn. WK Kellogg was subsequently acquired by Ferrero. The structural separation converted two entangled businesses into two clean, saleable assets, each attractive to a different acquirer with a different strategic logic.

Honeywell: Aerospace / Automation separation, 2025–2026

In February 2025, Honeywell’s board, after a year-long portfolio review led by CEO Vimal Kapur, announced it would separate into three independent, industry-leading companies: Honeywell Aerospace, a pure-play automation business (Honeywell Technologies), and a previously announced spin-off of its Advanced Materials unit. The board formally approved the Aerospace spin-off in June 2026, and the separation completed that same month.

What is notable is the explicit board language: the goal was not to fix an underperforming business, but to give each business “tailored growth strategies” and “distinct capital profiles.” Aerospace and industrial automation have almost nothing in common operationally, different customers, different cycles, different capital intensity. Bundling them had never made the businesses worse. It had simply made them harder to value, harder to run, and harder for investors to underwrite conviction in either one.

Results: Three companies now exist where one did, Aerospace (over $15bn in 2024 revenue), Honeywell Technologies (automation), and Advanced Materials, each structured to preserve full shareholder value through a tax-free separation.

The counter narrative is true, but there is a catch

Separation is not automatically value-accretive, and boards should be skeptical of anyone who claims otherwise. Kenvue, spun out of Johnson & Johnson in 2023, has traded down more than 20% since separation. Solventum, spun out of 3M in 2024, fell roughly 17% in its first year as a standalone company. Research from Trivariate points to a pattern: spin-offs into genuinely different industries from the parent tend to outperform, while separating an already high-quality, well-integrated business can destroy the synergies that made it valuable in the first place. Separation is a tool, not a strategy, it works when the businesses being separated were never actually one business in an operational sense, only in a reporting one.

So what actually decides it: correlation

Every example above shares a hidden variable that matters more than revenue, geography, or how long the businesses have been under one roof: correlation. Do the businesses share a customer, an asset, or a value mechanism that compounds when the businesses stay together, or are they sharing nothing but a balance sheet and a logo?

GE’s aviation, healthcare and energy businesses sold to different buyers, ran on different cycles, and competed in industries that had nothing to do with each other. Honeywell’s aerospace and automation businesses shared a name and a balance sheet, and little else that mattered operationally. Kellogg’s snacking and cereal businesses had diverging growth curves and needed different levels of capital investment to fix. In every one of these cases, the correlation between the businesses was low, and the case for separation was really a case for stopping the pretense that they were one business at all. That is precisely the test that fails for some of the most-discussed separation candidates in the market today, and it’s worth looking at why boards in those cases chose the opposite path.

When it doesn’t make sense: Amazon and AWS

Analysts and markets have been proposing a separation of AWS business from its parent retail business Amazon for a while now. The logic looks identical to GE’s on paper: AWS is a high-margin, high-growth cloud platform generating a large share of Amazon’s operating income, while retail runs on thin margins and heavy capital intensity. Sum-of-the-parts models built by analysts have repeatedly shown AWS alone would command a materially higher standalone valuation than it receives buried inside Amazon’s blended stock.

Amazon has never pursued it, and the reason is the correlation test. AWS was built on infrastructure Amazon originally stood up to run its own retail operation at scale, and AWS’s profits, in turn, fund the low-margin logistics, devices and pricing strategy that keep the retail flywheel spinning. The two businesses are not sharing a reporting line, they are sharing a genuine, ongoing economic dependency: retail generates the scale and data that make AWS valuable, and AWS generates the margin that lets retail keep underpricing to grow. Sever that link to satisfy a valuation model, and you may get a cleaner-looking AWS on day one and a structurally weaker retail business by day two, one that no longer has the internal capital to subsidize its own growth.

The lesson: a business throwing off superior margins next to one that doesn’t is not, on its own, evidence of a conglomerate discount. It’s only evidence of one if the two businesses aren’t actually funding, feeding, or depending on each other in ways that would survive the separation.

Disney and ESPN

Disney is the cleaner illustration because the company actually got close to doing it, then reversed course. Under CEO Bob Chapek, Disney seriously explored spinning off or selling ESPN and ABC, and Wall Street analysts spent much of 2022 and 2023 building the case publicly: linear television and sports rights, they argued, had almost no logical connection left to a company that was really a franchise IP business built on Marvel, Pixar, and Star Wars. One widely cited analyst note put it plainly, that there was very little reason for Disney and ESPN to remain together given how media consumption had evolved.

When Bob Iger returned as CEO in late 2022, he reversed the direction Chapek had been moving in. Rather than spin ESPN and ABC out as separate public companies, he reorganized the company around what he has called Disney’s flywheel: content created by the studios drives attendance at the parks, sells consumer products, and fills the streaming pipeline, and each of those channels in turn funds the next round of content investment. ESPN was made a standalone internal operating unit for accountability, but was kept inside the same public company, because sports content, live audiences and rights bundling still contribute to that same cross-monetization engine, even if the connection is less obvious on a segment-reporting table than aerospace and automation ever were unrelated at Honeywell.

The lesson: the fact that two businesses look statistically unrelated in a segment report doesn’t mean they are operationally unrelated. If the real driver of enterprise value is a flywheel that runs through both businesses, separating them on the strength of an analyst’s sum-of-the-parts model can destroy the very mechanism that made the sum worth more than its parts.

IBM and Kyndryl

IBM’s 2021 spin-off of Kyndryl, its low-margin, low-growth managed infrastructure services business, is a useful example precisely because it shows separation doesn’t create value symmetrically for both resulting companies. The correlation test supported the separation: IBM’s remaining hybrid cloud and software business needed a very different investment profile, growth expectation and multiple than a commodity IT-outsourcing business tied to zero-margin legacy contracts. Separating let IBM re-rate as a software and AI company, unburdened by a business that was structurally dragging its valuation down.

But Kyndryl itself had a much harder road as a standalone company. Without IBM’s balance sheet and internal cross-subsidy, it had to renegotiate a large share of its contracts out of break-even pricing on its own, and it took years, and a change in leadership confidence from the market, before the turnaround narrative held.

The lesson: correlation being low enough to justify separation for the parent doesn’t automatically mean the spun-off entity was built to thrive alone. Boards approving a separation owe the same rigor to the entity leaving as the one staying, since the newco doesn’t inherit the parent’s balance sheet, only its former problems.

So, what can be a working framework for boards?

1. Is there a real economic flywheel, or just a shared income statement?

Ask whether one business is genuinely funding, feeding, or depending on the other’s customers, data, or infrastructure, the way AWS and Amazon retail do, or Disney’s content and its parks and merchandise do. If the dependency is real and ongoing, separation severs value instead of unlocking it. If the businesses only intersect on a consolidated balance sheet, that dependency was never real to begin with.

2. Do the businesses need genuinely different capital structures and growth mandates?

GE’s three businesses, Honeywell’s aerospace and automation units, and Kellogg’s snacking and cereal businesses all needed different capital intensity, different growth expectations, and different investor bases to be run well. When one part of the business is quietly subsidising or constraining the capital allocation of another that doesn’t need the same treatment, that is a structural cost, not an operational one, and it won’t be fixed by a transformation program.

3. Is the complexity destroying the valuation independent of how well either part performs?

GE’s individual businesses were not badly run before 2021. Honeywell’s aerospace and automation divisions were not underperforming. In both cases, the businesses were fine, and the combination was the problem, because a blended equity story prevented either business from being valued, financed or incentivized on its own terms. If good management inside a bad structure is still producing a discount, it’s a separation question and not a transformation question.

4. Can the entity being separated actually survive without the parent’s balance sheet?

Kyndryl’s rocky first years as a standalone company are the reminder that a separation decision made correctly for the parent can still leave the spun-off business undercapitalized, under-resourced, or structurally disadvantaged in its own market. A board approving a separation has to underwrite both resulting companies and not just the one it’s keeping.

Not every business needs to be broken up, and not every business with a lower-margin segment sitting next to a higher-margin one is a candidate for separation. The diagnostic question most boards don’t ask until an someone forces it onto the agenda is this: are these businesses worth more to their own investors, their own management teams, and their own capital markets as separate entities, or is the value they create together something that genuinely disappears the moment they’re apart?

Transformation asks: how do we make this company better?

Separation asks a prior, more fundamental question: should this be one company at all?

The four examples above show that the answer isn’t determined by size, industry, or how uncomfortable the conversation is, it’s determined by whether the businesses are actually correlated in a way that compounds or only bundled in a way that confuses. Boards need to answer this question before moving with a transformation or separaton decision.