Every replatform proposal carries a completion date. Almost none carry the date the sponsor expects to sell. That gap is where enterprise value moves quietly from the current owner to the next one.
Table of Contents
Research Grounding
Bain's 2026 Global Private Equity Report, published February 22, sets out the new deal math under the heading "12 is the new 5." A 2015 buyout with roughly half the purchase price borrowed at 6% to 7% needed about 5% annual EBITDA growth to reach a 2.5x multiple on invested capital over a five-year hold. With borrowing now at 8% to 9%, leverage nearer 30% to 40%, and purchase multiples at record levels but no longer rising, Bain puts the requirement for that same 2.5x over five years closer to 10% to 12%.
Read the assumption rather than the number. Both versions of that model run on a five-year hold.
The same report puts average holding periods at exit around seven years, with almost 40% of companies held longer than five, up from 29% in 2019. Bain's analysis of buyout vintages from 2000 to 2015 shows IRR stagnating around year seven and declining after that.
So the return is underwritten on five years, the asset is held for seven, and the returns clock starts working against you at the moment the extra time was supposed to help. Modernization programs live inside that gap.
The PE Translation
Modernization gets scoped on engineering timelines and judged on fund timelines. Engineering answers how long it takes to build correctly, which produces an honest multi-year number. The fund asks how much EBITDA arrives before the exit window and what a buyer sees in the data room. Those two clocks rarely appear side by side in the same meeting.
A replatform with a 30-month payback started inside a hold with 24 months remaining is a capital transfer. This sponsor funds it, absorbs the delivery risk, and books the cost. The next sponsor receives the finished asset and the margin it produces. The work can be right for the business and still be financed by the wrong owner.
McKinsey's research on modernization sharpens the risk. Companies with the most severe technical debt are 40% more likely to end up with incomplete or canceled IT modernizations than those with the least, and McKinsey notes that tech modernization programs often outlast the tenure of the leadership that approved them. In a PE-backed company, that tenure has a name and a date attached to it.
Half finished is the hardest state to underwrite at exit. It is also, in the deals I have looked at, where these programs most often stop.
The Four Enterprise Debts
Modernization proposals price one debt and inherit four. That is why the estimates are wrong and why programs stall partway through.
Technology debt carries the only engineering estimate anyone produces: legacy architecture, unsupported dependencies, brittle integrations, undocumented services. Left alone, it shows up as delivery velocity that degrades every quarter while headcount stays flat.
Process debt is the operating logic encoded in systems and never written down anywhere else. Approval chains, exception handling, the manual reconciliation someone performs before the board pack goes out. The cost lands when a replatform cannot be specified, because nobody in the building can describe what the current system does.
Talent debt covers two things: institutional knowledge concentrated in a handful of people, and the skills gap between the stack you run and the stack you are proposing to run. Programs built without pricing it discover that the engineers who can safely change the legacy system are the same engineers needed to build its replacement.
Data debt is the absence of a shared, trustworthy definition of the entities the business runs on. This one usually detonates a full quarter after the code is ready, when billing, provisioning, support, and reporting each turn out to carry a different definition of what a customer is.
A proposal that prices only the first category will be wrong by whatever the other three cost, and those three surface late, after the spend is committed.
Operator Experience
Inside a PE-backed software company, the replatform line item reappears in the budget with the same confidence every cycle, and the completion date slides by roughly the length of the cycle.
What kills these programs is the middle state. During an incremental migration you run both systems at once: two deployment pipelines, two on-call rotations, and a synchronization layer between them that nobody planned to own. That sync layer becomes a product in its own right, with its own reconciliation logic, its own failure modes, and no place on the roadmap. The operating cost curve through all of this is a hump. Spend peaks while both systems are still live, which is precisely when the program is most likely to be paused for budget reasons.
McKinsey documents the shape of this in a large B2B business that found 70% of a $2 billion margin opportunity depended on technology costing $400 million, cut the investment to roughly $300 million, and walked away from a quarter of the margin. Two and a half years later the team had completed half the planned work.
The debt that stalls a migration is rarely the technology debt that was estimated. I have watched schedules absorb a full quarter over entity definitions that were assumed to be settled, and staffing plans collapse because the four people who understood the legacy billing path were also the four required to validate the new one.
In diligence, the middle state is what I look for first. Two systems live, a sync layer with no named owner, a completion date that has already moved twice. Sell-side materials describe that condition as "modernization underway," which reads as progress and prices as risk. A clean legacy stack can be underwritten, and so can a finished platform. The half-migrated estate leaves a buyer with no reliable basis for a completion estimate. The buyer prices the remaining capex and then applies an execution discount because the seller cannot bound either the cost or the risk.
The Remaining Hold Test
Five questions to run against any technology investment proposal before it enters a value creation plan. Any question without a clear answer is an unpriced risk, and the initiative goes back for decomposition.
Clock. What is the cash payback period, how many months of hold remain, and how much of the value gets recognized at exit rather than in cash before it? An initiative that delivers neither cash returns inside the hold nor a defensible exit-value benefit does not qualify in its current form.
Debt. Which of the four enterprise debts does this initiative genuinely require paid down to deliver its outcome, and which is it paying down out of habit? Fund only the ones on the critical path.
Mechanism. What is the named revenue or margin mechanism, and which KPI moves when it works? "Improved scalability" fails this question. "Removes the seat ceiling blocking six enterprise renewals" passes it.
Increment. What is the smallest increment that delivers the mechanism, and does that increment stand alone if funding stops the day after it ships? An increment that only has value when the next three land is not an increment.
Exposure. If this program is half done on the day the company is marketed, what does a buyer price? Remaining capex, execution risk, or both.
The rule that falls out: choose the value initiative first, then pay down only the enterprise debt that the initiative requires. Sequencing is the discipline here, and restraint is a poor substitute for it.
Four Overrides to the Remaining Hold Test
Four situations outrank the payback calculation, and treating them as optional is its own kind of value destruction.
A regulatory forcing function with a date on it removes the choice. The deadline sets the payback period, and missing it arrives as some combination of fines, lost certification, breached contracts, and revenue you are prohibited from recognizing.
Security exposure that will surface in buy-side diligence gets paid for either way. Remediate on your own timeline at your own cost, or have a buyer convert it into a discount, an escrow, a remediation condition, or a reason to walk. The second path costs more and lands at the worst moment.
Integration debt inside a buy-and-build thesis behaves differently from other technology debt, because the payback is denominated per acquisition rather than once. A platform that cuts integration time from nine months to three pays back again with every add-on, so the arithmetic improves as deal pace increases. Where the thesis is built on add-ons, integration capability is the thesis.
A genuine revenue ceiling can justify investment that pays back past the remaining hold. The test is whether contracts are being lost or refused, with named deals attached, and whether the resulting revenue opportunity can be demonstrated to a buyer at exit. Architectural preference dressed as a ceiling is the most expensive proposal in the building.
AI as a Diagnostic Wedge
Gartner predicts that over 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value, and inadequate risk controls. Those are management and architecture failures more than model failures. Escalating cost is usually data preparation and integration work that nobody scoped. Unclear business value tends to mean a process got automated whose output nobody downstream depended on. Risk-control failures expose process and talent debt together: decision rights were never assigned, control ownership sat with nobody, and the team lacked the capability to operate the system safely.
Every one of those is an enterprise debt surfacing under load, which makes an AI initiative a useful diagnostic wedge. It can surface the binding constraint on a far shorter cycle than a multi-year replatform, before the larger capital is committed.
Point an agent at a workflow and the data debt announces itself, because the agent cannot resolve the four definitions of customer that a human employee resolves from memory without noticing. Sequenced properly, the initiative names the debt and the paydown gets scoped to what it requires. Sequenced badly, it becomes another program with a multi-year dependency chain underneath it and a business case that expires before the dependencies clear.
Boardroom Question: Which technology initiatives will not pay back inside the remaining hold, and for each one, what value can we credibly capture at exit, or what smaller increment will pay back before then?
Three decisions
Put remaining hold on every technology business case as a required field, alongside payback period, and require the two numbers to appear on the same page. The change costs nothing and surfaces the mismatch before the capital is committed.
Reclassify every modernization program sitting in the migration middle state, defined by observable conditions rather than a completion estimate: both systems still live, a completion date that has moved more than once, or an inability to state what it would cost to retire the legacy environment. For each one, decide deliberately whether to fund it to completion, cut it back to an increment that stands alone, or reverse it. Drifting is the option that guarantees the worst exit position.
Require that any AI initiative approved this year name the enterprise debt it expects to expose, before it starts. Then hold the debt paydown budget against what the initiative proves is binding, rather than against the modernization roadmap that was written before anyone tested it.
One number
Five years. That is the holding period Bain's 2.5x return math assumes, in both the 2015 version and today's. Average holding periods at exit have drifted to around seven. Roadmaps get built against the seven, returns get measured against the five, and the two years in between are where half-finished programs accumulate.
Board Takeaway: A replatform whose value arrives after the sale is the next owner's asset and this fund's expense.
Portco Brief translates one technology signal a week into an enterprise-value decision for operating partners and portfolio CEOs, CFOs, COOs, CTOs, and Technology Leaders. Forward it to the operator who keeps saying the audit came back clean.
Subscribe at portcobrief.com.