The Separation Cost Model: What the Deal Model Usually Misses

By the MeridianCogent team — this perspective draws on published carve-out cost research across the industry. BCG frames separation economics across one-time costs, transitional services, dis-synergies and stranded costs; for execution management, we find it useful to separate two additional mechanisms — standalone run-rate uplift and cost created purely by schedule slippage.

---

The purchase price in a carve-out is negotiated once, in a room, over weeks or months, by people who scrutinise every input. Separation economics are different. They move every week after signing, across several distinct cost mechanisms that rarely get modelled together, and by the time all of them are visible, the deal that looked attractive in the transaction model can look materially different once the actual cost of separating the business is added up.

None of the figures below are new — several appear elsewhere in this library, attached to specific problems. What's useful is putting them in one structure, because a deal model that treats "separation cost" as a single contingency line is missing the fact that it's actually several distinct mechanisms, each with its own driver and its own timeline.

One-off separation costs

Temporary costs needed to execute the separation itself: legal and entity setup, IT migration, data migration, PMO and advisory fees, HR transition work, and facilities changes. BCG's analysis of separation costs across more than fifty divestitures found these typically running 1 to 5 percent of the divested business's revenue, with complex or large carve-outs reaching or exceeding 10 percent. Because the underlying effort is often similar in absolute terms regardless of deal size, smaller divestitures tend to sit at the higher end of that range as a percentage of revenue.

TSA costs

Temporary operating costs paid during transition — not just the base service charges quoted at signing, but the step-up and extension pricing that activates if a service runs past its base term. TSA exposure shouldn't be modelled only at the base rate: extension escalators mean schedule risk has a direct economic consequence, so the cost model should carry a downside case for delayed exits, not just the on-schedule figure.

Standalone cost uplift

The permanent increase in the standalone entity's ongoing run rate — the gap between what a service cost at the parent's scale and what it costs to run independently. McKinsey's own account of its carve-out due diligence work is specific here: it has observed cases where the cost of replacing centrally allocated services ran as high as 200 percent of the original allocation — meaning a shared-services allocation that looked like $10 million in the parent's books can cost up to $20 million to stand up alone. This shows up in lost purchasing scale, standalone software licensing negotiated at a fraction of the parent's volume, and functions that have to be duplicated rather than simply continued.

Stranded costs

Seller-side costs that remain after the corresponding business leaves. Practitioner guidance from BD Emerson puts stranded costs at roughly 20 to 40 percent of the cost previously allocated to the divested business. Deloitte's research on divestiture outcomes separately found that 51 percent of divesting companies saw a profitability drop greater than 3.3 percentage points in the year after the deal — a figure that captures the scale of the problem even where the precise mechanism varies by company. This is a seller-side cost, and it's the one most likely to be missing from a buyer-focused deal model entirely, even though it directly affects what the seller actually nets from the transaction.

Dis-synergies

The economic benefit of combination that disappears on separation, and isn't fully captured in the other categories above. Loss of shared procurement scale and shared logistics that both sides previously benefited from and neither can fully replicate alone. These are harder to quantify precisely than the other mechanisms, which is exactly why they're the most likely to be assumed away rather than modelled.

Execution slippage

The cost of the plan not holding. Extra months of TSA at step-up pricing, duplicate running costs where old and new systems overlap longer than intended, and delayed value capture pushed later than the deal model assumed. This category doesn't exist as a fixed number — it's a function of how well the separation is actually executed against the timeline the other categories assumed.

Why these don't show up in one place

Ownership of these costs is fragmented across functions — Legal owns entity and TSA structuring, IT owns migration cost, Finance owns standalone uplift if anyone owns it explicitly at all — which makes it easy for the consolidated exposure to become stale or inconsistent even when each function is genuinely tracking its own piece carefully. The issue isn't that nobody is doing the work; it's that the full picture requires someone whose job spans all of it, and that role often doesn't exist on the deal team.

Whose economics are they?

Worth being explicit that "separation cost" isn't one symmetric number shared by both sides — each mechanism lands differently depending on who's exposed to it:

An illustrative example

Purely illustrative, not a claim about typical figures — a €300 million-revenue carved-out business might see something like:

The point isn't that these numbers are typical — it's that the original "€12M separation budget" was never really a €12M economic exposure. Each of the other categories hits a different P&L, in a different period, owned by a different function, and none of them show up if the deal model only tracked the first line.

The thesis underneath this

The purchase price is a single negotiated number, fixed at signing. Separation economics are not a number — they're a set of mechanisms that keep moving for months or years after the price is agreed, each with its own driver, and each capable of eroding the value the price was based on if nobody is tracking it against the original assumption. The deal model establishes the baseline. Execution needs a live cost model, not a static contingency line reviewed once at signing.

Lessons learned

Six mechanisms, not one. A single "integration cost" line in a deal model becomes a real gap the moment any one of these categories runs ahead of what was assumed.

Stranded cost is a seller-side number, and gets left out of buyer-built models for exactly that reason. If nobody on the deal team is explicitly responsible for the seller's side of the ledger, it doesn't appear anywhere.

TSA exposure modelled only at the base rate is modelled optimistically. Carry a downside case for extension, not just the on-schedule figure.

Execution slippage is the mechanism that makes all the others worse or better. It isn't a fixed cost; it's the multiplier that determines whether the other categories land close to their estimate or well past it.

---

Sources: BCG, "The 2021 M&A Report: Mastering the Art of Breaking Up" and "6 Technology Due Diligence Imperatives in Carve-outs" (separation cost analysis); McKinsey & Company, "Solving the carve-out conundrum" (standalone cost replacement analysis); Deloitte, 2026 Global Divestiture Survey and divestiture profitability research; BD Emerson (TSA step-up pricing and stranded cost sizing analysis).