Value creation runs on evidence operators do not have

The tailwinds that used to do the operating partner’s job are gone. What is left is an evidence problem — and it sits inside the customer base.

Natalie McCrae, Founder, Keystone IQ · 15 min read

Operating partners have never mattered more to returns. The market has taken away the tailwinds that used to do their job. What is left is a value creation problem, and it sits inside the customer base.

The value creation evidence problem

Operators are asked to move EBITDA they cannot see

According to the McKinsey 2026 Global Private Markets Report, for the decade ending 2022, leverage and multiple expansion did most of the work in private equity. Between them they drove 59% of buyout returns; operational value creation supplied the remaining 41%. That balance has not simply shifted. The components that supported it have been removed. Entry multiples reached a new high in 2025 at 11.8 times EBITDA, edging past the 2022 peak and far exceeding the 2010 to 2022 average of 9.1 times. Over the same period, debt’s share of entry multiples fell from 44% in 2016 to 37% in 2025. The cheap fuel is gone. What is left is the engine.

KPMG calls the result operational alpha, and calls it the price of admission. The 2025 Value Creation in Private Equity report makes the same argument with sharper arithmetic: revenue growth accounts for between 56% and 70% of enterprise value uplift depending on outcome quartile. Margin does far less work than its reputation suggests. The top-quartile margin expansion achieved by high-MOIC deals is broadly the same as that achieved by the bottom quartile. If revenue is the lever, then the quality of that revenue is the thing to underwrite.

The BDO Private Equity Value Report 2026 confirms the same shift on the ground. Across 245 of the highest-growth portfolio companies tracked between 2022 and 2024, average EBITDA CAGR reached 62%. That growth did not come from the market. It came from inside the business.

An evidence problem, not a resourcing problem

The operating partner model was built for a market that no longer exists. That much is widely acknowledged across the industry. The usual prescription follows quickly: give operators more authority, clearer mandates, earlier involvement, better playbooks.

Authority is half the problem. The other half is evidence. An operating partner is handed a mandate to move margin without a true picture of where margin sits. McKinsey’s data shows value creation is back-loaded: for deals exited since 2019, around 6% of the final EBITDA margin is generated in the last year of the hold, 4% in the year before, and roughly 1% in each year prior. The effort is crammed at the end. Earlier and sustained execution produces better returns. You cannot execute early against a picture you do not have.

This is the operating partner’s real constraint. Not the org chart. Not the authority structure. Not even talent or change management, though both matter. The absence of customer-level evidence at the moment it is most needed.

Revenue quality is a customer question

Reported revenue describes the past in aggregate. It hides who carries the margin. It understates churn at value. It flatters the quality of the new customers coming in. The headline can rise while the structure beneath it weakens. None of that appears in the P&L. Evidenced revenue quality changes how the asset should be valued.

Revenue growth is the primary driver of enterprise value uplift and the most reliable path to earning a higher exit multiple. Cosmetic improvements no longer carry exits. If the revenue picture is flattering rather than true, buyers will find it when they look closely.

Customer evidence is not the same as customer analytics. AI is collapsing the cost of analytics. It is not collapsing the cost of the interpretation that reads them. That interpretation is the operator judgement layer: the experienced reading that turns a pattern into a costed lever, with a named owner and a date against it.

What the headline cannot show

Consider this European ecommerce business, anonymised. Around €1.92bn in reported revenue. Customers up 32% year on year. Every headline metric pointing the right way.

When every customer was scored using Keystone IQ’s Revenue Quality Architecture™ and tracked year on year, the structure told a different story.

The customers leaving were disproportionately the valuable ones, so the reported churn rate understated the damage by almost eight points. A concentration of revenue sat in a minority of customers, plain in the movement data and invisible in the count. And most of the new growth was arriving weak. The groups here were not imposed. They emerged from natural breaks in the data, from how customers actually behaved, not from tiers decided in advance or modelled proxies.

That is the difference between reading a customer base and forcing it into a template. The headline said one thing. The base said another. Only one of them is useful in developing a winning value creation strategy.

From advice to a costed lever

A 2026 survey of more than 100 senior PE leaders found operating partners describing value creation as a continuous intelligence layer rather than a transactional event. The same leaders named revenue quality scorecards and unit economics as the foundation of effective monitoring, and ranked leading indicators, including pipeline, conversion and customer retention, ahead of lagging financial outcomes. Identifying early signals of performance risk was their single biggest monitoring challenge.

This is evidence, not forecast, a mirror, not a model. It is descriptive, built from what customers have actually done, not a model of what they might do next. The distinction matters. A model of what customers might do requires assumptions about behaviour that have not yet occurred. A record of what they have done requires nothing of the kind. It is the difference between projection and proof.

Give an operator the real structure of the customer base, who carries the margin, where it is quietly shifting, what the reported numbers miss, and the brief becomes a set of costed moves with owners and dates. That is the difference between an operating partner and an expensive bystander.

The same evidence defends the exit

What an operator builds across the hold is what management has to defend at sale. EY’s 2026 exit-readiness study found that 60% of GPs name developing a robust set of data and KPIs as their hardest finance-function challenge at exit. Evidencing value creation initiatives in exit EBITDA is named both the top challenge, cited by 55% of respondents, and the single biggest determinant of exit outcomes, cited by 59%.

The exit preparation runway amplifies this. Around half of GPs who began preparation 12 to 24 months before sale reported “much” or “a great deal” of improvement in exit outcomes. Those starting six months out reported materially weaker results. Customer evidence built over the hold is what lets the equity story hold up when a buyer probes it. Built the week before sale, it does not.

The multiple you exit on is the market’s grade on how convincingly you built a better business. BDO’s deal partners and their portfolio companies confirm the same pattern: narrative alone is no longer enough. Buyers want evidence of delivery and a management team that has proven it can execute.

The brief is sound. The evidence is missing.

The operating partner brief is sound. The people are often excellent. The model fails when it asks them to create value they were never shown, then judges them on EBITDA that moved too late to count.

The fix is not another playbook. It is the evidence underneath one.

Keystone IQ’s Value Creation Map identifies the customer-base levers an operator needs to turn a value creation brief into costed, defensible moves across the hold.

Sources

FAQs

The problem is rarely capability. It is timing and evidence. Operating partners typically receive a mandate before they have a clear picture of where value actually sits in the customer base. Without that picture, early execution is guesswork. The work accelerates toward exit not because operators are slow, but because the evidence they need arrives late, if it arrives at all.

Revenue quality is a measure of how structurally sound the revenue line is, not just how large it is. It distinguishes between revenue carried by a loyal, high-value core and revenue that depends on constant replacement of customers who leave. A business with strong reported revenue but high value-weighted churn and weak acquisition cohorts is a structurally different asset from one where the same headline is supported by a stable, concentrated core.

Reported revenue is an aggregate. It records what came in, not who brought it, how reliably they will return, or whether the mix is improving or deteriorating. Churn in aggregate understates the damage when the customers leaving are disproportionately valuable. Growth in aggregate flatters the business when new customers spend less and stay for shorter periods. The P&L does not distinguish between these. Customer-level transaction data does.

CDD assesses the market: size, growth, competitive dynamics, management capability. It does not typically assess the asset at individual customer level. These are different questions. A market can be attractive while the specific customer base being acquired is quietly deteriorating. Customer-base analysis and CDD are complementary, not substitutes.

It matters from the outset. McKinsey’s analysis of deals exited since 2019 shows that only around 1% of EBITDA margin improvement accrues in each year prior to the final two years. Operators who wait until year four or five to build a customer evidence base are compressing their execution window to the point where meaningful improvement is very difficult to deliver, and even harder to evidence at exit.

AI can identify patterns in customer data at speed and at scale. It cannot supply the commercial context that determines what a pattern means in a specific deal. Whether a shift in customer composition is a structural risk or a temporary effect of an acquisition cohort requires a reading that draws on operator experience. The judgement layer is what converts a data output into a costed decision with a named owner and a date.

Buyers underwrite the revenue they expect to inherit. If the evidence for that revenue is narrative rather than data, they discount it. A business that can show year-on-year cohort movement, retention trajectories, and acquisition quality at customer level gives buyers less to challenge and less reason to widen the valuation gap. The equity story holds in diligence when it is built on evidence rather than assertion.

Reported churn counts the proportion of customers who leave. Value-weighted churn measures the proportion of revenue those customers represented. When the customers leaving are disproportionately high-value, the headline churn rate understates the damage. In the anonymised case in this article, reported churn was 6.2%. Revenue-weighted churn was 14.1%. The difference did not appear anywhere in the P&L.

Talk to us about your customer base

Days not weeks. Fixed scope. No PII required.

Get in Touch