Quality of Earnings Ends Where the Hold Begins.

The return has moved inside the company, into the revenue system that now has to carry the entire hold, and the one document every buyer trusts was built to describe the past.

There is one document a buyer almost never skips. Before the wire goes out, before the committee signs, an independent firm is paid to open the target's books and produce a Quality of Earnings report. It normalizes the profit, strips the one-time items, tests whether the earnings turn into cash. It is the closest thing private equity has to a verdict, and serious buyers commission it on nearly every deal. They are right to.

It is also, in every line, a report about the past.

For most of the last fifteen years, that was the correct thing to buy, because the past was a fair guide to the return. That is the part that is no longer true. The instrument did not change. The thing it was pointed at moved.

Here is what the verdict was quietly standing on. For fifteen years, a buyer did not have to be excellent at revenue. They had to be excellent at two other things: the price paid going in, and the cost of the debt used to get there. McKinsey put a number on it that should be read twice. Of the total return on buyout deals entered in 2010 or later and exited by 2021, roughly two-thirds came from just two sources: the market multiple rising, and leverage.

Neither of those is a property of the company. One is a property of the market you happened to sell into. The other is a property of the rate you happened to borrow at. For a decade, most of the return was manufactured outside the business, and the business only had to not get in the way.

That is the decade the Quality of Earnings report was built for. When the engine of the return sits in the entry price and the exit market, a rigorous account of past earnings is a reasonable thing to trust, because you are not really betting on the company. You are betting on the multiple, and holding a clean receipt. Both of those bets have been called.

You can watch the old engine stall in a single line: the cash actually coming back to investors. Bain's Global Private Equity Report 2026 puts distributions to limited partners at 14 percent of net asset value in 2025. That is the fourth year in a row below 15 percent, which the report calls an industry record, and a level last seen in the 2008 to 2009 financial crisis. The people who committed the capital are not getting it back.

They are not getting it back because it cannot get out. Roughly 32,000 companies, worth about 3.8 trillion dollars, are sitting unsold in portfolios right now. The average hold has stretched to about seven years at exit, up from the five to six that held across 2010 to 2021. And the debt that used to be nearly free now costs on the order of 8 to 9 percent.

Read those four facts as one sentence. The exit is closed, the multiple is flat, the loan is expensive, and the clock has added two years. Every escape route the old model used to reach for is blocked at once.

An extra two years of hold is not a delay. It is a different bet entirely. It lifts the whole weight of the return off the exit and sets it down on the operating years. Which is to say, on the revenue.

Bain's own shorthand for the new arithmetic is blunt, and it is worth carrying out of this brief if nothing else is: twelve is the new five. A deal that used to need about 5 percent average annual EBITDA growth to clear the benchmark return now needs something closer to 10 to 12 percent. The report's own words are that attractive returns "now requires significantly more operational improvement and revenue growth." That is not a tightening. It roughly doubles the growth the operating business has to produce on its own, with no help from the multiple and a headwind from the rate.

So the return has moved. It used to live in the entry price and the exit market, the two things a backward report can speak to. It now lives inside the company, in the system that produces the revenue, for seven years, at a rate the business has rarely had to hit unassisted.

McKinsey's data confirms where the edge went. General partners that genuinely emphasize asset operations earn internal rates of return two to three points higher than peers who do not. The premium has moved to the operators. Which leaves one question standing in the diligence room, and it is not a question about the past: before you buy, how do you know the revenue system can actually carry the number.

Now go back to the one document no one skips, and read it for what it is. A Quality of Earnings report normalizes historical earnings. It removes one-time items and discretionary expenses, tests whether reported profit converts to cash, examines working capital, checks that revenue was recognized under real standards. Its stated purpose, from the advisory practice that sells it, is to establish "the degree to which reported profits reflect a company's true, sustainable, and recurring economic performance."

Read the verbs. Normalize. Remove. Test. Examine. Verify. Every one of them points backward, because that is the honest and difficult job the instrument was built to do: prove the earnings were real. It does that job well. It was simply never asked the second question, the one that used to be answered by the multiple and now is not: will the system that produced those earnings reproduce them, and grow them 10 to 12 percent a year, for seven years, alone.

This is the quiet inversion at the center of the deal. The question the Quality of Earnings report answers remains essential. Proving the earnings were real is not optional, and it never will be. It simply no longer carries the return by itself. The pack scores the accounting, the contracts, the legal exposure, the tax position, and the reported performance, and it scores them well. What it does not produce is a standardized, independent, comparable rating of the one thing now expected to carry the return. That thing has a name. It is the revenue architecture. The pack validates its outputs. It does not rate the system.

The natural objection is that the pack already handles this, because it looks hard at retention and growth. It does look at them. It looks at the numbers. The numbers are outputs. That is the whole trap.

Take the metric that carries the most weight in a modern software diligence: net revenue retention. McKinsey's work on B2B technology shows exactly why it earns that weight. Top-quartile companies run net revenue retention near 113 percent and trade at roughly 24 times enterprise value to revenue. Bottom-quartile companies sit near 98 percent and trade at about 5 times. The best of them, in the report's phrase, "grow 13 percent without adding any new business." Retention is not a vanity number. It is most of the valuation.

But a 113 is a reading, not a mechanism. It tells you expansion happened. It cannot tell you why. It cannot tell you whether the company has an engineered motion that produces expansion on purpose, or whether two large accounts simply grew last year and will not grow next year. One of those is a durable system worth 24 times revenue. The other is a good number balanced on top of a fragile one, and it will read exactly the same on the page. The diligence pack, reading the output, cannot separate them. It records that the number is 113 and moves on to the next tab.

Under the old model, this was survivable, because the return was sitting safely in the multiple and the revenue only had to not collapse. Under the new one, with 10 to 12 percent riding on that motion for seven years, the difference between an engineered 113 and an accidental one is the difference between the fund's best deal and its worst. And it is invisible to every instrument currently in the room.

This is not a gap of methodology. The methods in the pack are excellent at the jobs they were built for. It is a gap of instrument. There is no standardized, independent, comparable external instrument for rating the revenue architecture itself, the system that now has to produce the growth the whole deal depends on.

Notice that every other asset in the stack has one. Credit has a rating. Public equities have coverage. The building has an appraisal. The earnings have a Quality of Earnings report. Each is external, standardized, and comparable across companies and over time. The revenue system is examined, often with real care, by commercial due diligence, market studies, growth and go-to-market assessments, and operating-partner review. That work matters, and much of it is good. But it is bespoke to each deal, scoped to the single target in front of it, and not built to place one company's revenue architecture against another's on a common scale. There is assessment. There is not yet an instrument.

The market is what turns this from an academic gap into a priced one. When two-thirds of the return came from outside the company, an unrated revenue system was a tolerable blind spot; you were not really betting on it. Now that essentially all of the return has to come from inside the company, across a hold two years longer than planned, the unrated revenue system is the single largest unpriced risk in the deal, and it sits outside the scope of the one report every buyer trusts.

There is a name for the difference between the two documents this brief has described. The Quality of Earnings report is an autopsy: a rigorous, certified account of what was true, performed on a period that is already closed. What the deal now requires is a diagnosis: an examination of a living system that has to run, hard, for seven more years. The industry has spent a decade perfecting the autopsy because the autopsy was enough. The autopsy is still necessary. It is no longer sufficient.

The report that only looks backward was built for a market that paid you to be right about price and patient about time. That market is gone. Distributions are at crisis-era lows, holds are stretching past seven years, and the return has moved inside the company, into the revenue architecture, which existing diligence examines but no standardized instrument yet rates.

Quality of Earnings will keep telling you the revenue was real, and you will still need it to. That question remains essential. It simply no longer carries the return by itself. The harder question, the one the next seven years turns on, is whether the machine that made the revenue can make more. That question has no instrument yet. Building one is the work.