Hey all,

I recovered from my UK jetlag just in time to be discombobulated by another unexpected change. The beloved Sunny - who cuts my hair and both of my boys is moving to Houston! (and with equivalent housing 25% of the cost compared to Northern Virginia, who could blame her?)

It's harder than you might anticipate to find a barber who meets the bar for teenage boys but let me tell you - in the era of 'looksmaxxing' we have a tough search on our hands to replace this 'single point of success' in our lives.

Single points of success are also a feature in almost all businesses, not least my own. It's always such a hard balance to embrace the friction and slow-down of systemizing something that is taken care of so adeptly by the one person who performs it so effortlessly.

You delay documentation and automation as a lower priority...until, of course, it is thrust upon you by a diligence process or a sale.

Anyway, I'm looking for NoVA teenage barber recommendations.

Enjoy the rest of this week's memo!

Cheers,

Graeme

Three Things I Learned This Week

The 90-day feedback loop is a data problem in disguise

Alex Hormozi wrote something this week that reads like a data strategy memo, even though it never mentions data. He opens with a Buffett line. "If you get nine women pregnant, you don't have a baby in a month." Some things take time. Fine. Then he makes the point that matters for anyone running a portfolio company.

If you catch yourself saying "we will not know if this worked for another 90 days," that is not a patience problem. It is a missing leading indicator. No process in a business should take three months to tell you whether it is working, because that caps you at four iterations a year. The competitor who found an earlier signal is running twenty.

His fix is to find the earlier step you can measure. If churn is the problem, activation is the leading indicator. If the sales cycle is long, qualified calls with a budget-holding decision-maker is the earlier signal. You still wait the 90 days for the final read, but you iterate weekly on the leading metric.

Here is the translation for operating partners. The reason most value creation plans run on annual feedback loops is that the data to support a faster loop either does not exist or cannot be trusted. You cannot iterate on activation if activation is not instrumented. You cannot watch qualified-call conversion if the CRM is a graveyard. The faster feedback loop Hormozi is describing is a data readiness outcome. The firms that iterate twenty times to your three are not smarter. They built the measurement layer first.

The Pygmalion Razor and the data standard you set

Sahil Bloom published a piece this week on what he calls the Pygmalion Razor. The Pygmalion Effect is the documented phenomenon that people rise to the expectations placed on them. Bloom's version is to spend your time around people who hold high expectations of you, because you grow into the standard they assume you can meet. The discomfort of that room is the cost of entry for growth.

I read it through an operating partner lens. The expectation you set for a portfolio company's data is the expectation it grows into. If the board only ever asks for a top-line revenue number, a top-line revenue number is all anyone builds the muscle to produce. If you expect a number that traces to source systems, survives a buyer's question, and ties out across the GL from quarter one, the team builds that capability long before it is under diligence pressure.

Most operating partners set the investor-grade data expectation about three months before exit. By then it is a fire drill. The Pygmalion-aware ones set it three years out, when it is just the house standard, and the company quietly grows into it. The expectation is the intervention.

Link: Sahil Bloom on the Pygmalion Razor (sahilbloom.substack.com/p/the-most-powerful-decision-making)

95% say AI works. 31% use it where it matters most.

Two numbers from this week's PE research that only make sense side by side. FTI Consulting's 2026 PE AI Radar reports that 95% of funds say their AI initiatives are meeting or exceeding the original business case. AI works when it is deployed. Then S&P Global's survey of general partners found that only 31% have AI somewhat or fully integrated into diligence, and the single biggest barrier, cited by 49% of respondents, is lack of expertise.

Sit with that. The technology is delivering against its business case almost universally where firms actually use it. And most firms still have not put it into one of their most data-intensive core processes. The constraint was never whether AI works. The constraint is expertise and the readiness of the underlying data to feed it.

This is the same point from a different direction every week in this newsletter. The bottleneck is not the model. It is whether the organization has the data foundations and the people to put it to work. The firms closing that gap are not buying better AI. They are doing the unglamorous data and capability work that lets the AI they already own actually run.

Two News Stories From This Week in Mid-Market PE and Data

Medallia just wiped out $5.1B in equity. The post-mortem is about the AI story it never had.

Sources: Insight Innovation Ventures, May 15 edition

What happened. Thoma Bravo has handed control of Medallia, the experience-management software company it took private for $6.4B in 2021, to a group of lenders led by Blackstone that holds much of the roughly $2.8B in outstanding debt. The PIK relief window expired, Thoma Bravo declined to inject further equity, and roughly $5.1B of equity value was wiped out in the restructuring.

Why you should care. The reason this matters beyond the headline is the read on why it happened. The market is no longer patient with PE-backed, heavily leveraged software businesses that lack a credible AI infrastructure story. Medallia sits on enormous volumes of customer-experience data, exactly the kind of proprietary asset that should be worth more in an AI world, not less. But owning the data and being able to turn it into a governed, AI-ready, defensible advantage are two different things. A $6.4B take-private carrying $2.8B of debt has no room for "we own a lot of data but cannot yet make it work." The lesson for any leveraged portfolio company is blunt. Your data assets only protect your valuation if you can demonstrate they are AI-ready. The debt does not wait for the data project to finish.

Buyers are paying record multiples, and underwriting them on data advantages

Sources: McKinsey Global Private Markets Report 2026

What happened. McKinsey's 2026 report shows buyout multiples hit a record median of 11.8x EBITDA in 2025, edging past the 2022 peak. The rebound in deal value came because acquirers are paying more, not doing more deals. Deals over $500M now carry a five-year median multiple of 15.8x against 11.5x overall. The more important line sits in the underwriting section. McKinsey notes that over the past 12 months AI has shifted how dealmakers assess risk, away from broad sector-disruption narratives and toward a granular assessment of asset quality based on "verifiable competitive moats, data advantages, embedded workflows, and execution capability."

Why you should care. Read that list of what buyers now underwrite on. Verifiable competitive moats. Data advantages. Embedded workflows. Three of the four are data questions. The premium multiples are real, but they are concentrated, and they are increasingly awarded to companies that can prove a data advantage rather than assert one. Verifiable is the operative word. A buyer paying a record multiple is going to verify the moat, and the verification happens in the data room. The companies earning the 15.8x are the ones whose data advantage holds up when someone checks. This is the bifurcation, now written into how the largest firm in the business describes underwriting.

Free Tool of the Week - The Buyer's Private Scorecard

Most sellers find out how a buyer graded their data room after the offer lands lighter than expected. The Buyer's Private Scorecard is the rubric diligence teams actually use to evaluate a data room, across six categories. Worth a look before you build the data room, not after.

See the scorecard here

Sign-off

If any of this lands, and there is something you think we can help with, just reply. We read everything that comes through.

As always, forward this on to your favorite PE-backed friend.

Cheers,

Graeme