Ask a mid-market M&A team where the model for a given deal lives, and you’ll get a folder path. Project Atlas / 04_Financials / Model_v17_JC_final_FINAL.xlsx. The model is a file. It sits next to the deal, but it is not part of the deal. The data room is one thing, the CIM is another, the model is a third, and a junior analyst is the only connective tissue holding the three in sync.
That separation is the source of most of the modeling pain we hear about. Not the formulas. The disconnection.
A model is a snapshot of documents that keep moving
When an analyst builds a deal model, they are doing one thing at the core: turning the financial documents in the data room into a structured, calculable view of the business. Historical P&L from the management accounts. Working capital from the balance sheets. Adjustments from the QoE. Revenue build from the contract schedule.
The moment that extraction is finished, it begins to rot. The seller uploads a revised set of management accounts. A new monthly drops into the data room. The QoE team reclassifies three add-backs. Each of those events should ripple into the model. In practice, someone has to notice the change, find the cell, re-key the number, and re-check every formula downstream of it.
On a live deal, with documents arriving weekly, that reconciliation is a part-time job no one is formally assigned. So it doesn’t happen consistently. The model drifts from the data room, and nobody knows by how much until a buyer’s diligence team finds the gap.
The version problem is really a source problem
Everyone blames version control. v17_final_FINAL is the industry’s running joke. But versioning is a symptom. The real problem is that a spreadsheet cell has no memory of where its number came from.
When a partner looks at the EBITDA line and asks “where does the €4.2m come from?”, the honest answer is often a chain of recollection: that came from the FY24 management accounts, page 12, the adjusted figure, before the rent normalization we agreed on the call. That chain lives in the analyst’s head. It is not written down anywhere. When the analyst is on holiday, or has rolled off the deal, the chain is gone, and the number becomes unauditable.
This is why models can’t be safely reused, why every new analyst rebuilds rather than inherits, and why partner review is so slow: the reviewer can’t trust a number they can’t trace, so they re-derive it themselves.
What changes when the model is attached to the deal
The fix is conceptually simple and operationally significant: stop treating the model as a file that lives near the deal, and start treating it as a structured object that lives inside the deal, derived from the same documents everything else is derived from.
Three things follow from that.
Every figure traces to a source. A cell is not just a number; it carries the document, the page, and the line it was read from. “Where does the €4.2m come from?” becomes a click, not a conversation. The audit trail is the model, not the analyst’s memory.
The model is downstream of the data room, not parallel to it. When a new set of accounts lands, the figures that depend on it are flagged, not silently stale. The question shifts from “is the model current?” — which no one can answer — to “here are the four lines affected by the document that just arrived.” Reconciliation becomes a review task, not an archaeology project.
The architecture is consistent across deals. When every mandate produces a model with the same backbone — same historical build, same normalization logic, same working-capital treatment — a partner can review the tenth deal as fast as the first, because the structure is familiar. The firm’s modeling standard stops living in one senior analyst’s habits and becomes a property of the platform.
The objection: “our models are bespoke”
They are, and they should be. The valuation logic, the operating assumptions, the sensitivity cases, the way you bridge to a buyer’s likely synthetic — that is judgment, and it is yours. Nobody wants that automated, and it shouldn’t be.
But the foundation of every deal model is not bespoke. Pulling three years of P&L out of a PDF, structuring it, normalizing it, tying it to the balance sheet, and keeping it current with the data room — that is the same mechanical work on every mandate. It is where most of the hours go and almost none of the judgment lives. Standardizing the foundation is what frees the analyst to spend their time on the part that is actually bespoke.
Where this is heading
The teams that win competitive mid-market mandates are not the ones with the cleverest formulas. They are the ones who can respond to a buyer question in an hour instead of two days, because the model and the documents are the same source of truth rather than two artifacts a junior is racing to reconcile.
That is the shift: from a model that is a file you maintain, to a model that is a view of the deal that maintains itself.
NaS_OS builds a structured, source-traced financial model on every deal directly from the documents in the data room — so the model is part of the deal, not a file beside it. If you run an M&A team and want to see what that looks like on your own mandates, apply for access.