There is a number in the McKinsey research on value creation that gets read backwards by almost everyone who sees it.
When you look at where EBITDA-margin improvement actually shows up across a hold, most of it lands late. The final year and the year before it carry the bulk of the gains. The early years contribute far less per year. The chart looks like a hockey stick, flat for a while and then a sharp rise near the exit.
The obvious conclusion is that value creation is a late-hold activity. Get the company stable, run it for a few years, then turn on the operational improvement engine in the back half when the exit is in sight. The data seems to say so.
That reading is wrong, and it costs firms real multiple.
What the late spike actually tells you
Late margin improvement is not proof that value creation belongs late. It is what you get when you start late.
The same McKinsey work that shows the late spike argues the gains do not have to arrive late. The firms that break the pattern shift execution earlier in the hold and sustain it. They do not wait for the back half. The improvement curve starts rising sooner and keeps rising, which means by the time they reach the exit window they are compounding gains on top of gains rather than scrambling to manufacture them.
The average firm produces the hockey stick because the average firm spends the early years getting organized. Integration noise, a settling forecast, a management team finding its feet. The operational levers do not get pulled hard until the exit clock forces the issue. So the improvement clusters at the end, and everyone mistakes a behavior pattern for a law of physics.
It is not a law. It is a choice about when you start. And the firms that choose to start earlier do not give up the late gains. They keep those and add the early ones on top.
The math the curve hides
Think about what compounding does to a margin improvement.
A margin point added in year two of a seven-year hold works for five years. A margin point added in year six works for one. The early point carries more weight, because every operational gain you bank early keeps producing while you stack the next one on top of it.
Run the two paths side by side. Two firms, same company, same set of levers, same total improvement available. One starts pulling in year two and adds gains steadily. The other waits until the exit is in sight and crams the same work into the final two years. Both arrive at a similar margin on paper at the exit. But the early mover spent four extra years operating at a higher margin, throwing off more cash, funding more add-ons, and walking into the data room with a track record instead of a recent spike. A buyer looks at a long, steady climb and sees a durable business. A buyer looks at a sharp jump in the last eighteen months and asks whether it is real or whether it was engineered for the sale. The shape of the curve is itself a diligence signal.
This is the same logic I wrote about in the 2026 numbers that matter. The deal math has changed. Required EBITDA growth has roughly doubled, financial engineering carries less of the return, and every operational lever now has to contribute. When you needed 5% growth, a late push was enough. When you need 10 to 12%, you cannot afford to leave the early years on the table. There are not enough months in the back half to produce that much improvement from a standing start.
Hold periods make this sharper, not softer. McKinsey puts the average hold at roughly six and a half to seven years now. A longer hold sounds like more time to fix things later. In practice it means more years of compounding available to the firm that starts early, and more years of drift available to the firm that waits. The same stretch in the timeline rewards early movers and punishes late ones. It does not even out.
Why most firms cannot start early even when they want to
Here is the part that ties the timing argument to the work I actually do.
You cannot execute operational improvement early if you cannot measure the business early. And most portfolio companies cannot measure themselves cleanly in the early hold. They have three versions of revenue, a margin number that takes two people and three days to assemble, and no way to attribute a quarter of growth to the initiatives that produced it.
A management team in that state cannot pull operational levers with any confidence in year two. They do not know which segment carries the margin, which product line is quietly losing money, or which initiative moved the number versus which one just consumed budget. So they do the safe thing. They run the company, keep the lights on, and wait for the exit to force a reckoning. The improvement clusters late because the visibility to act early was never there.
This is the constructive other half of the year-two problem I described in diagnosing a stalled value creation plan. That post was about what to do when the plan has already stalled and you need to find the real constraint. This one is about not stalling in the first place. The mechanism is the same in both directions. A company that cannot see its own performance cannot act on it, whether you catch the problem at year two or design around it from day one.
The firms that produce the early-and-sustained curve are the firms whose portfolio companies can read their own instruments from the start. That is not a coincidence. It is the prerequisite.
The objection, and why it does not hold
The usual pushback is that the early hold is too noisy to act on. The integration is not done. The team is new. The systems are mid-migration. Pulling operational levers in that environment is premature, so the disciplined move is to stabilize first and improve later.
There is a real point buried in that, and it is also where the timing argument gets misused. Stabilizing the company and making it measurable are not the same task, and they do not have to happen in sequence. You can establish trusted definitions for a handful of metrics while the integration is still in flight. In fact that is the better time to do it, because you are already touching the systems and you can set the standard once rather than reconciling three versions of it later.
What you cannot do is treat measurement as something that comes after stabilization. If you wait for the company to settle before you build the ability to see it, you will be most of the way through the hold before you have a number you trust. By then the compounding window has closed and you are back to the late spike, telling yourself you waited because the early years were noisy. The noise is exactly why you build the instruments early. You cannot navigate noise you cannot measure.
What starting early actually requires
Starting early does not mean launching a twelve-month transformation in the first 100 days. It means making the business measurable fast enough that you can act on it while the compounding window is still wide open.
Pick the five to seven metrics the value creation plan depends on. Revenue by segment, gross margin by product or service line, retention, pipeline conversion, and a couple of operational drivers specific to the business. Not the forty on the dashboard. The handful that, if they move, mean the thesis is working.
Establish one source and one definition for each, signed off by finance, sales, and operations. The win here is killing the second and third versions of the truth, not buying new tooling. One number per metric that nobody relitigates is what lets a management team act in year two instead of arguing.
Then connect every initiative in the value creation plan to one of those numbers, with a baseline and a target. The initiatives that cannot be connected to a metric are the ones to question first. They were never measurable, which means you would never have known whether they worked.
This is the discipline behind a data-driven value creation plan. When the plan and the measurement share the same small set of numbers, you can start pulling levers early and watch them land in the actuals, quarter by quarter, instead of hoping it all comes together in the final two years.
The diagnosis is a few weeks. Establishing trusted definitions for a handful of metrics is usually a quarter. That is a small investment against a hold that runs six or seven years, and it is the thing that moves you from the average curve to the top-performer curve.
The honest version of the data
The McKinsey numbers are right. Most margin improvement does land late. If you want to know roughly where your portfolio companies sit on the timing question, the VCP Data Score is a quick way to gauge whether a company can support early execution or is set up to repeat the late-spike pattern by default.
What the late spike does not tell you is that late is good. Late is what average looks like. Late is the signature of a firm that started the operational work when the exit forced it to, rather than when the value was greatest.
The firms beating the market are not finding some secret lever in the back half. They are pulling the same levers everyone else pulls, just sooner, and letting time do the compounding. They can do that because their companies can measure themselves from the start.
If your value creation curve is flat for years and then spikes at the end, that is not a strategy. It is a tell. It says the early years were spent getting ready to act instead of acting. The fix is not to push harder at the exit. It is to make the business measurable early enough that you never have to.
Start early, and the margin compounds in your favor. Start late, and you pay for it in the multiple. The data only looks like an argument for waiting if you read it backwards.