Aug 12, 2026Finance
It Works Every Time in Excel
Every rollup model works on the way in. It is a projection, and I have never seen one land exactly right. What closes the gap is the operator.
A manufacturing owner I heard about recently had two offers on his desk.
One was 7x from a search fund. Two MBAs, an SBA loan, a clean model, a 100-day plan.
The other is 4x from another manufacturer. An operator who has run the same kind of shop for two decades and knows what the machines cost to maintain.
He is leaning toward the 4x.
Three turns of EBITDA is real money to leave alone, and he is not being sentimental about it. He trusts the operator and he believes the operator can actually run the place. That is the trade: three turns for confidence about which buyer is still standing in year three, with his name on the building and his employees on the floor.
Let me be clear about my seat in this. I have never run a plant. I have never made payroll off a line of credit, or told a shift supervisor his line is being consolidated. My partner Vijen Patel, though, has experienced this.
After co-founding 81, 5 years ago, we have seen many teams proposing to buy and run businesses like that manufacturer’s, and after enough of those meetings the pattern in who pulls it off stops being subtle.
The Meetings
I have had a lot of meetings this year with teams building “AI rollups”. Buy fragmented services businesses at 3x, install software, cut the cost of delivery, exit the platform at 8x. (See my other article on valuations).
The people running these are impressive, mostly ex-PE, investment banking or consulting. They build the model, source the pipeline, structure the debt and negotiate the LOI. Very little about the analysis is wrong.
But most of them are missing one seat, and it is the seat that determines whether any of the analysis ever happens: the operator.
Spreadsheet Math
Every model can be inputted to say the same thing. Buy the business at 3x. Exit at 8x. Integration is seamless. Organic growth continues uninterrupted through the transition. Synergies land in month four.
It works every single time in Excel.
Every one of those models is a projection. I couldn’t name one that came in exactly right. Not directionally right, exactly right, the way it was presented in the LOI.
That is not a knock on modeling. A model is how you decide whether a deal is worth attempting at all. A projection is a claim about the future made by people who have to execute it, and the size of the gap between the model and the outcome is almost entirely a function of who is doing the executing.
The operator is the part that decides it. The general manager who has run the place for 15 years quietly deciding whether to cooperate with the new owner or slow-walk him until he goes away. Two acquired companies on different systems with overlapping customer lists. The three best drivers who are loyal to the old owner, not the logo. The founder you just bought from, now technically your employee, in a job he has never had.
None of this shows up as a line item. All of it shows up in the exit multiple.
Multiple arbitrage is the easiest part of a rollup to underwrite and the last part you actually earn. The hard part is integration, and integration is not a financial skill. It is an operating skill.
The Four Seats
Four capabilities we want in the room before we back an acquisitive business in services or manufacturing; not four people necessarily, but four capabilities. When one is missing you can usually predict where it breaks.
Operational expertise. Someone who has run the business being bought, not advised it or diligenced it. This is why a seller picks you at a lower price, and why the acquired team stays after the wire clears.
Financial steward. Someone who owns the capital structure and has discipline to walk from a deal that prices badly. In a rollup, the worst outcome is not failing to buy; it is overpaying at scale.
Growth. Someone accountable for organic growth, because the model assumes it and acquisitions actively distract from it. Rollups that stop growing organically become a portfolio of declining assets with debt on top.
Tech. Someone who can build and implement, not just buy software. Right now that is the difference between AI as a line in the deck and AI as a real change in unit cost.
Terra Freight
Our most recent Fund II investment is Terra Freight, an AI-enabled freight brokerage rollup. Part of why we invested is that all four seats were filled by people who had already done the work.
David bootstrapped the business to eight figures of revenue with no outside capital, during one of the toughest freight markets. He knows carriers, brokers, and shippers like the back of his hand.
Elie owns the technology, as a builder, previously working at Uber.
Mitch came out of PE and M&A where he actually implemented acquisitions, not just closed them.
The Test Cuts Both Ways
The reason I take the seller’s instinct seriously is that our own investment firm has to pass the same test we are applying. I am a version of the financial seat. Vijen built and sold an operating business before any of this, and he is the growth operator in our room. Jeff did M&A and integration at Uber after being COO of Tide Cleaners. Jordan was CTO at Tide Cleaners and Head of Product at Rocketmiles (Booking.com). If the room were only people with my background, we would be one of the teams described in this post.
Execution is the whole thing, and execution has a resume.
The manufacturing owner weighing 4x against 7x is pricing that directly. He is willing to give up three turns for trust and for operational competence, and from where I sit he is probably right to.
Subscribe nowAlex Kirshenbaum is an investor and co-founder at The 81 Collection, a high-growth equity fund that invests minority positions in service and manufacturing businesses at the intersection of cash flow and technology. Please reach out to alex@81collection.com for further inquiries.
