CIMs Are Advocacy Documents

The document that was written to sell you the company

A deal partner takes a company to the investment committee with a thesis, an entry multiple, a 100-day plan, and a return already promised to the fund's limited partners. The committee approves. The wire goes out. The entire case rests on a confidential information memorandum and a management presentation, both written by the seller's bankers for the express purpose of getting that wire sent. Everyone in the room knows the CIM is an advocacy document. The thing that should bother anyone who underwrites for a living is how little of the underwriting ever reaches past it.

The thesis is a claim about a future, and the future runs on things two and three levels below the CEO: whether the team really delivered the margin recovery the deck credits to them, or whether two directors did and are now fielding recruiter calls; whether the plant numbers are real or the dashboard is configured to flatter; whether the monthly close survives the day the one controller quits. None of that is in the data room, which records outputs. The people who know which parts of the story are true have a paycheck to protect, a reference to keep, and a job that just became a question mark the day the company went into play. So they say the safe thing, or nothing, and the safe thing rolls up into the clean fact base the committee underwrites.

What the canon actually found about reading a team

Start with the most measurable part, which should embarrass the median deal process. The cleanest finding in the management-assessment literature is a base rate. Geoff Smart and Randy Street's structured method^[1], the Who interview that ghSMART has now run more than twenty-nine thousand times, reports that typical unstructured hiring lands the right executive roughly half the time, while a structured, chronological, reference-checked assessment moves the hit rate toward ninety percent. Half to ninety is the difference between a coin flip and a near-certainty, sitting on the highest-leverage decision in the deal, and most diligence still reads management off a two-hour presentation built to impress. The method is not exotic. Walk the full career in strict order, and for each job, in the move that does the work, get the boss's name spelled out, then “when we talk to her, what will she say.” When, not if. The threat of the reference check is the truth serum.

The same shape is older still. Brad Smart's Topgrading^[2], the industrial-psychology root the ghSMART style grew out of, runs a three-to-four-hour tandem interview across fifty competencies, with the candidate arranging the reference calls themselves, because making them arrange it is both truth serum and test. Smart's published cost of a senior mis-hire is the number that should govern how much a fund spends here: up to fifteen times base salary for a VP and above. A fund will spend a quarter of a million dollars on quality-of-earnings to confirm the EBITDA, then bet the whole thesis on a management read it bought for the price of lunch.

Here the literature gets a clean academic spine. Kaplan, Klebanov and Sorensen^[3] took the ghSMART data on more than three hundred buyout and venture CEO candidates and asked which characteristics actually predicted success. The answer was uncomfortable for how boards behave: execution skills, being fast, aggressive, persistent, predicted buyout outcomes more strongly than the interpersonal skills interviews reward. The CEO Genome work of Elena Botelho and Kim Powell^[4], built from seventeen thousand assessments, lands in the same place: decisive CEOs were twelve times more likely to be high performers, and the trait boards underweight most is reliability, the dull one. The polished presentation most deals read a leader off is tuned to reward exactly the charisma the evidence says does not predict the outcome, and to miss the follow-through that does.

The value does not live in the boxes

Even a perfect read on the CEO underwrites the wrong thing, because the equity case is not concentrated in the corner office. Sandy Ogg, the former Blackstone operating partner, and McKinsey's talent-to-value work^[5] make the point with numbers that should reorganize how a fund spends its retention dollars. Locate the value in roles rather than titles and it lands in thirty to fifty roles carrying seventy to eighty percent of the value agenda, and those roles are not where the org chart suggests: about sixty percent sit two layers below the CEO and thirty percent three or more layers down. A retention package aimed at the C-suite is aimed at the wrong heads. The directors who carry the margin recovery and the controller who is the only one who can close the books hold the value and the truth at once, and they are precisely the people the deal process is structured never to interview.

The operators who run the value plan from the inside say it in plainer language. Adam Coffey, three times a PE-backed CEO across fifty-eight acquisitions^[6], insists that in a service business “your product is people,” so his real diligence on an add-on is two levels down: who runs dispatch, which technicians customers ask for by name, what walks out the door the day the owner leaves. Jeff Sands, the turnaround practitioner^[7] puts it as a rule: find the truth on the shop floor, not the conference room, because the floor knew the diagnosis years before the board did. The information that prices the deal is real, specific, and held by a person with every structural reason not to volunteer it to the people writing the check.

Now price what it costs to be wrong

This is not an academic complaint, because the base rates on being wrong are brutal and public. Take the searcher first, where one bad read on one asset is the entire fund. Stanford's 2024 Search Fund Study^[8], across six hundred and eighty-one funds, reports headline returns that look terrific, a 35 percent IRR and 4.5 times invested capital, and in the same breath reports that thirty-one percent of acquisitions lost money, a third of those a total loss. A buyer with one shot who underwrote the owner's narrative and skipped the long-tenured employee who could have said what dissolves the day the owner stops answering the phone has roughly a one-in-three chance of destroying capital. Ruback and Yudkoff, who teach this at HBS^[9], are explicit that confirmatory diligence has to go past the quality-of-earnings to how the business actually works: talk to the employees and the customers, map what only the owner does, and treat a seller's refusal to grant that access as information.

In the larger funds the cost of a bad management read shows up not as a total loss but as a CEO change in year two, the same failure wearing a suit. The AlixPartners PE Leadership Survey^[10], the standard citation on portfolio-company churn, finds that sixty-five percent of PE firms replace the portfolio-company CEO during the hold, eighty-six percent of that turnover sponsor-driven, spiking around year two “as expectations collide with performance reality.” That is a confession that the management bet was wrong when it was made and discovered only after two years on the meter. The literature on both ends, the ninety-percent structured hit rate going in and the sixty-five-percent replacement rate coming out, is telling the same story from opposite directions.

The strategy houses reach the same conclusion from the returns data, and this is the part that makes it urgent rather than merely true. Bain's Global Private Equity Report^[11] documents that multiple expansion drove roughly half of PE returns over the last decade, that margins often declined when teams had underwritten expansion, and that the era of buying low and selling high on the multiple is over. When the cheap lever is gone, the return has to come from operational truth found before close, the real margin mechanism with a name and an owner attached, which is exactly what the CIM cannot give you and the floor can. EY-Parthenon makes the structural version^[12]: diligence has shifted from confirming risk to being the source of the value plan itself. Diligence that reads only the documents is reading the seller's pitch back to itself and calling it confirmation.

Why it stays buried

Pull the threads together and they converge on one mechanism, which is the thing we actually believe. The information that prices a deal rarely lives in the data systems. It lives in the heads of the people doing the work, and it stays there because every channel built to collect it punishes honesty and gets quieter at every level it climbs. The skip-level who says the named executive is carried by two deputies is handing the buyer a reason to replace the boss who controls his review. The controller who admits the close is one person's spreadsheet is admitting the person is himself, and that his leverage is the secret. Each silence is locally rational. The first manager rounds the floor's diagnosis up into a cleaner number, the cleaner number averages into a dashboard, and the dashboard reaches the committee as a fact. The sum is a decision-maker who is structurally the last to know the one thing the floor knew first.

The grim joke is that the truth is not missing. Every expert in this literature is, underneath the framework, describing the same recovery move: get a neutral party to ask a specific person about a specific recent moment, somewhere the answer cannot be used against them. Smart's named-reference Who interview, Topgrading's candidate-arranged reference calls, Sands's floor-first walk, Ruback and Yudkoff's talk to the employees and treat refusal as information. The reason a fund does not have the answer is not that it is unknowable. It is that the answer is unasked, because the only channels built for asking are the two places the people who know have the most reason to keep quiet.

Which points, finally, at the kind of instrument this calls for. Not another data room, which records outputs by design, nor another management presentation, which is advocacy by design. Something closer to what the canon has described for years and few have run at the scale and speed a deal clock allows: a neutral, confidential conversation with the specific people two and three levels down who hold the value and the truth, anchored to a real recent episode rather than an opinion on request, run for enough voices that no answer traces back to one person. That is the instrument we are building at Latent Variables. The literature already told us where the equity case sits. The open problem was only ever reaching it before the wire goes out.