The clock is gone: how longer hold periods change the buy-and-build platform

When I ask an operating partner how many definitions of a qualified lead exist across their group, I usually get silence. For years that didn't matter, because the clock was running: buy, tidy up, sell in five. That clock is gone, and it changes the arithmetic of every buy-and-build platform. Holding periods at exit now run around seven years, against five to six between 2010 and 2021 ^[1]. Unifying marketing and sales costs from day one and pays after years. That's why the hold period decides whether it's worth doing at all.

What actually changed about hold periods?

Two developments run in parallel and land in the same place. The first one wasn't chosen. More than 16,000 companies have sat in portfolios for over four years, 52 percent of all buyout-backed inventory and the highest on record ^[2]. Even the 2025 rebound left the queue intact. Exit value jumped 47 percent to 717 billion dollars while the number of exits fell 2 percent ^[1].

The second development is deliberate, and it rarely makes it into the conversation. 123 new evergreen funds launched in 2025, 32 of them in private equity, against 50 launches in all of 2020 ^[3]. These funds have no end date. They never have to sell.

Which of the two you believe in decides the whole calculation, and almost nobody says which. Here's my read. The backlog is cyclical and clears with the next open exit window. The evergreen structures are structural and stay. A decision resting on the backlog alone survives until the next good selling year.

What follows applies to both roads. To earn the return that 5 percent annual EBITDA growth delivered a decade ago, you now need 10 to 12 percent, with debt costing 8 to 9 percent ^[4]. Those points come out of the operating business. In the mid-market, bolt-on acquisitions are gaining ground on large transformational deals for the same reason ^[5].

What unification in a buy-and-build platform actually buys

Unification usually gets justified on cost. One master agreement, one tool stack, one procurement lever, each pays back once. In software transactions, 52 percent of value creation came from growth and 6 percent from margin, with the remainder from multiple expansion and cheap debt ^[6]. Both of those have largely gone, which shifts the whole weight onto the 52 percent.

Learning speed compounds every quarter. Take twelve holdings, each testing four things a quarter: a new segment, a different price anchor, an added channel, a reworked onboarding flow. With separate data models, each company learns from its own four experiments. With one shared model, all of them learn from 48. Across eight years that's 128 usable experiments per company against 1,536 for the group (own calculation).

That calculation assumes everything travels, and it doesn't. Price anchors and onboarding flows travel almost always, segment tests almost never. Assume a third and you still land above 500 (own calculation).

There's a documented reason it pays late. Revenue synergies take three to five years to reach full effect, and the first measurable wins appear in the first 90 days at the earliest ^[13]. Under a five-year clock the return falls behind the exit. Across eight years, five years of compounding remain.

Once two companies use the same definition, a campaign that worked in one becomes portable to the other. Before that it was an anecdote from a board meeting. What happens next I've seen several times. The sales leads start calling each other unprompted, because they're looking at the same numbers, and they stop explaining to one another why their market is different.

What belongs to the group and what stays with the company?

The line worth drawing runs through each individual company.

Everything the customer sees stays local. The brand, the vocabulary of the industry, the contracting party, the person whose mobile number the customer actually has. That's where the reason for buying the company sits, and that's what gets destroyed fastest.

Everything the customer never sees belongs to the group. This is where it gets concrete, because "shared data model" sounds like a mega-project and isn't one. Every company keeps its own system. What matches are the objects and their mandatory fields: contact, account, opportunity, campaign, quote, order, each with the same status values and the same currency logic. Do that for six objects and you can report across twelve companies with twelve different CRMs underneath.

Nobody has to invent the definition either. Forrester's B2B Revenue Waterfall is the most widely used standard and counts opportunities together with the buying group that decides ^[11]. An opportunity counts once need, decision authority, a budget or a credible path to one, and a timeline are on the table. Which standard a group picks is up to the group. That everyone picks the same one is the entire point.

The yardstick can be named too. For marketing and sales, two numbers are enough to start: qualified opportunities per quarter, and new business as a share of revenue. Both calculated by the same formula everywhere, monthly, in the same format, owned by one person at group level.

Constellation Software supports half of this. Brands, products and acquisition decisions sit entirely with the business units, and every unit is still measured on the same number, return on invested capital plus organic revenue growth, with compensation attached ^[7]. A shared marketing and sales data model is not part of that picture. So my claim goes further than the example does, and the reason sits in what gets bought. Software arrives with a sales engine built in and recurring revenue. Services businesses arrive with sales teams whose knowledge lives in people's heads. Buy the second kind and you pay for it twice while you run them side by side.

Two cases sit right on the line and resist a clean answer. Price belongs to the company, the band belongs to the group, otherwise two holdings end up bidding against each other at the same buyer. Access to a sister company's customer list stays a negotiation and never becomes a rule. Leave both open in year one and you'll settle them in front of the board later.

Some portfolios are exempt from all of this. If your companies sell to genuinely different buyers, a tool distributor to procurement and a software vendor to the head of IT, campaigns and content will not travel. The shared yardstick still holds. Shared campaign production does not.

What does unification cost when it goes wrong?

What I keep seeing in these programmes is an inverted sequence. System migration starts before anyone has defined what gets measured. Months of project work then tie up the people who bring in revenue, and the output is an interface everyone can operate and nobody needs for the job they're paid to do. After that the whole topic is dead for years.

The order of magnitude is documented. Gartner puts the annual cost of poor data quality at an average of 12.9 million dollars per organisation, and names inconsistency across sources as the hardest data quality problem to solve ^[12]. That's per company, not per group.

It gets more expensive when the reason you bought the company takes the hit. A holding acquired for its closeness to an industry, then fed central campaigns that sound corporate inside that industry, loses exactly what was paid for. A sales team that can no longer write to customers in their own language rarely protests loudly. It simply stops contributing.

Politics enter the moment a shared yardstick exposes gaps that used to disappear inside different calculation methods. Two companies with identical revenue growth suddenly look different. Raise that before the first report lands on the table, because afterwards you'll be arguing from the back foot.

Where AI actually earns money inside a portfolio

One question does more for a group than any list of AI ideas. Where do we do the same thing twenty times, slightly differently? Quoting, tender responses, onboarding sequences, win-back campaigns, special pricing approvals. That list is simultaneously the AI plan and the unification plan.

FTI surveyed 200 decision-makers at private equity firms running at least a billion dollars. 36 percent say their portfolio companies use AI in daily operations, 7 percent across the whole business ^[10]. BCG attributes roughly 70 percent of the value in AI programmes to people and process, 20 percent to data and 10 percent to algorithms ^[8]. Leave the process untouched and you've bought software and earned a slide deck.

What holds it back is people. 35 percent point to the shortage of AI and IT specialists ^[10]. A group hires that capability once. A standalone company hires it every time. The obvious objection is that one central team becomes a bottleneck with twelve queues, and for day-to-day operations that's fair. For building the first shared reporting layer it isn't.

Everyone buys tools. William Barnett described the race in which competitors drive each other to run faster without gaining ground ^[9]. When the same models are available to everyone, advantage shifts to what is hard to buy: your own data model and your rebuilt workflows.

Governance sits in the same drawer. Build the guardrails once for the group and adapt them per company. That saves the repetition and speeds adoption up, because nobody waits on a matter of principle any more.

How to draw the line in four steps

1. Put in writing first what the managing directors keep: brand, customer relationship, pricing authority inside agreed bands. A board resolution plus an addendum to the managing director's contract, owned by the board chair. Skip this at position one and you'll negotiate the next three twice.
2. Measure how far the definitions have drifted. Send every company the same five anonymised deals and have them tick which ones sit in the pipeline and at what value. One week, one spreadsheet. The spread is your answer.
3. Set the yardstick before you touch a tool. Two metrics, one formula, monthly in the same format, owned by one person at group level. Only once those numbers run stable for three months do you talk about systems.
4. Define the abort criterion before you start. If after two quarters on a shared yardstick no company has adopted anything from another, the yardstick is paper. Stop there. Rolling out on a paper yardstick costs more than the pause.

My Take

What interests me most here is a number that appears in no data room today. I expect buyers of secondary portfolios and continuation vehicles to start asking, within two or three years, how many companies in a group measure on the same data model. That share is the cheapest available signal of whether a group can still learn, and it'll move price long before anyone sells a maturity model for it. The stiffest resistance I'd expect somewhere nobody looks for it, inside the fund's own deal team. A shared yardstick reveals which add-ons carried the growth story and which ones only lengthened the list. Well, and whoever makes that visible rarely wins friends.

For how the wall everyone builds differs from the moat that pays, see the moat that decides value. On what open interfaces do to software business models, read MCP and software business models, and for the single-company view, building your B2B SaaS castle.