Synergies Were Priced Before Anyone Asked the People Who Owe Them

The number arrives before the people do

A deal model is a promise made on behalf of strangers. The banker builds it from a data room, the board approves a premium against it, and a synergy line gets allocated to a department head who will not learn the deal exists until the morning it is announced. By then the number is load-bearing: it has justified a price, sized a retention pool, and set a date. The person who has to deliver it hears a figure attached to their area for the first time and nods, then goes back to their desk and quietly works out what is real. Nobody asks them, and the structure is built so that nobody can. That gap between the model and the floor is where most deals go to die, and the people who write about M&A for a living have said so plainly since the early 1990s.

The failure is structural, not a matter of competence. Synergy estimates are built top-down, under confidentiality, before the close, by a deal team forbidden from talking to the middle of the target. The people who hold the ground truth, which accounts buy from a person rather than a logo, which cost line two workstreams have each quietly claimed, which billing run lives on one spreadsheet, are excluded from the model by design and then made accountable to it by surprise. The information that would falsify the number is sealed off from the number until it is too expensive to change. This is not a soft people issue. It is an information architecture that manufactures wrong prices.

What the canon actually found

The cleanest indictment of the premium is the oldest. Mark Sirower's synergy-trap work, carried forward with Jeff Weirens^[1] into a study of 1,267 deals from 1995 to 2018, makes an argument most acquirers still refuse to internalize: the premium is not a price, it is a performance bar. Pay 30 to 50 percent over market and you have committed to a level of operating improvement that has to be earned back before the deal creates a dollar of value. The data is unsentimental. The market's first reaction tends to persist: deals it liked at announcement averaged plus 8.4 percent a year out, deals it disliked minus 9.1 percent. And 60 to 90 percent of acquisitions fail to deliver their original objectives. The trap is that the premium gets justified by synergies nobody has tested against a named owner, a baseline, or a one-time cost, which is to say synergies that exist only in the model.

If Sirower priced the bar, McKinsey's merger practice^[2] measured how reliably acquirers clear it, and the split is the most useful fact in the field. Cost synergies are roughly knowable: about 60 percent get achieved, a quarter overestimated by at least 25 percent. Revenue synergies are where the model is fiction. Roughly 70 percent of deals miss their revenue targets, because a revenue synergy is an assertion about how customers and salespeople will behave, made by people who have never met either. And 42 percent of the time, diligence produces no adequate roadmap for capturing the value the deal was priced on. In four cases out of ten there is no map to the top. The dis-synergies, the customers who leave and the salesforce that goes dark during the distraction, are simply omitted, which makes the number not only optimistic but structurally one-sided.

Where Sirower and McKinsey diagnose the model, Bain's integration practice^[3] names where the model goes to break. In Bain's review, 83 percent of failed acquisitions trace a primary cause to integration, and about 75 percent of integrations struggle with culture even though 80 percent of acquirers claim to address it before diligence closes. That last pairing is the tell. Culture is not under-attended out of ignorance; it is attended to in a register that never reaches the friction. You can run the culture survey, write the integration thesis, and still discover at month nine that the floor has been quietly working around the harmonized process the whole time. Bain's own rule, integrate where the value is and protect Day 1, is correct and rarely enough, because the data on where the value actually sits and what Day 1 will actually break is held by people the integration team has no clean channel to.

Money aimed at titles, not at the people who hold the deal up

Nowhere is the structural failure more legible than in the retention pool, because there the mis-targeting has a dollar figure. WTW's recurring M&A retention study^[4] puts these pools at 1 to 2 percent of purchase price, so a billion-dollar deal carries a ten-to-twenty-million-dollar pool, and its central warning is that money buys time, not loyalty, and is routinely aimed at title rather than at criticality. The pool gets allocated off the org chart, because the org chart is what the buyer has, and it is exactly the document that cannot show who the team actually calls at 2 a.m. Rob Cross's organizational network analysis^[5] puts numbers on the blind spot: 3 to 5 percent of people drive 20 to 35 percent of the value-adding collaboration, and formal talent systems miss about half of them. So the dollars flow to the credited names while the actual linchpin, junior, unpromoted, invisible to the HRIS, takes a recruiter's call the week of the announcement. The pool was real money spent on the wrong people, because the instrument used to target it was a chart.

And the people the pool is meant to hold leave anyway, on a clock longer than the pool's. Jeffrey Krug's study of more than 12,000 executives^[6] across 473 firms over fifteen years found that acquired companies lose a third to 40 percent of their top managers within two years, elevated for as long as nine. The mechanism matters: departures track perceived loss of status and autonomy more than money, precisely the thing a thirteen-to-eighteen-month cash bonus cannot touch and a believable role story might. A retention design that pays a leader to stay while stripping the autonomy they valued has bought a countdown, not a commitment. The signal that predicts the exit, how the person talks about their changed role, lives in a conversation the buyer never has.

The merger syndrome, and why the survey is too slow to catch it

The human side has its own canon, and Mitchell Lee Marks and Philip Mirvis^[7] named the pathology in the 1980s and have advised more than a hundred combinations since. Their merger syndrome is a predictable degradation: heightened self-interest as everyone computes what the deal means for them, crisis reflexes, decisions centralizing upward while rumor runs sideways, distrust, a measurable productivity loss. The part that indicts the standard toolkit is their method. They do not run a survey and read the average; they sit with people below the executive layer and ask them to recount specific recent events since the announcement, what they heard, from whom, and what they did next. The rumor content is the data, because it reveals the operative fear. Mercer's people-risk practice^[8], which runs diligence on thousands of deals, reaches the same place from the actuarial side: most failed deals trace to unaddressed people issues, and retention is the number-one perceived people risk among dealmakers. Both describe information that decays. The us-versus-them flashpoint, the manager gone quiet, the hallway promise the 100-day plan is about to break, all are visible on the floor weeks before an engagement survey lands, and people do not write the truth into a survey during a merger anyway.

Why it stays in their heads

Pull the threads together and they converge on a single mechanism, which is the thing I actually believe. The information that would correct the deal model 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 during exactly the window the deal depends on. Tell the integration meeting your number is fiction and you have marked yourself disloyal in the week your new owners are deciding who to keep. Write the broken founder promise into a memo and you have created the document that ends the earnout. Flag the single-person dependency in payroll and you have volunteered a problem nobody thanked you for. So the manager nods and recalculates in private, the supervisor who can see who went quiet is on no distribution list, the truth gets rounded up, averaged into a workstream color, and handed to the steering committee as green. Each step is locally rational. The sum is a board structurally the last to know the one thing the floor knew first.

The grim joke is that the truth is not missing. Every named expert above is describing the same recovery move under a different framework. Sirower wants each synergy tested against the named owner who would deliver it. McKinsey wants the estimate rebuilt bottom-up with the line owners. Cross wants the real network mapped before the pool is set. Marks and Mirvis want people below the executive layer walked through specific recent events. None of it is exotic. Acquirers do not have the answer not because it is unknowable, but because it is unasked, since the only channels they built for asking are the ones the floor learned, fast, to go quiet in.

Which points, finally, at the kind of instrument this calls for. Not another data room, which by construction cannot show the workaround, the side promise, or the linchpin. Not another engagement survey, which arrives a quarter late and reads the safe answer back. Something closer to what the canon has described for thirty years and few have run at scale before the price locks: a neutral, confidential conversation, anchored to a real recent moment rather than an opinion on request, run with enough of the people who would actually deliver the model that no answer traces to one person. That is the kind of instrument we are building at Latent Variables. The literature already told us where the deal model is wrong. The open problem was only ever reaching the people who knew it, before the premium became the thing the integration team is quietly failing against.