Where the Truth Leaks

The CIM said on-time delivery ran at ninety-eight percent, and I believed it right up until I found the freight file. Two hundred pages of carrier invoices, and the phrase that kept repeating was next-day air. A precision fabricator does not ship steel by air because things are going well. Somebody was paying a fortune to make late parts arrive on time, and the somebody was about to be me.

Nobody had lied. On-time delivery, measured against the promise dates in the system, genuinely ran at ninety-eight percent, because the scheduler moved the promise dates. The freight bills were simply the only document in the building that had never been asked to hold an opinion. They recorded what happened, invoice by invoice, because the carrier needed to get paid. That is the property I have learned to prize above everything else in diligence: paper that exists for an operating reason, generated continuously, too boring and too granular for anyone to bother dressing up.

Last week I argued that the CIM is advocacy and that being surprised by this is the amateur move. The professional move is knowing where advocacy cannot reach. Every company generates a layer of records with no author and no audience, and that layer is where the truth leaks.

Documents With No Author

Everything in a data room belongs to one of two families. The first was prepared for you: the CIM, the management presentation, the adjusted bridge, the projections. It has an author, and the author has a goal. The second family was prepared for nobody. The AR aging exists so the controller knows who to call on Thursday. The maintenance log exists so second shift knows what first shift did to the press. The quote log exists because no estimator alive can remember four hundred open quotes. These documents have users rather than readers, and that is what keeps them honest.

They are also hard to fake for a mechanical reason: volume and time. A freight file is thousands of entries accumulated over years by people who had no idea a sale was coming. Reshaping it would mean rewriting history across systems, vendors, and carrier records, which is fraud, and sellers in the lower middle market are optimists, frequently, but they are very rarely felons. So the operating layer stays true even when every prepared page above it has been flexed to its limit.

The work is knowing which unprepared files to pull and what each one cannot help telling you. Money first, then the floor.

The Money Moves First

The AR aging is the first thing I read on any deal, before the CIM if I can get it. Customers vote with their payment behavior. A top account drifting from thirty days to sixty over four quarters is a relationship souring in real time, whatever the customer chart claims, and a cluster of short pays and credit memos around one part number is a quality problem wearing an accounting costume. The aging also reads forward: a customer who suddenly starts paying early, after years at forty-five days, may be cleaning up its own balance sheet for a sale, and two exits colliding is worth knowing about before you underwrite the relationship.

Monthly working capital tells you what the annual statements smooth away. Inventory that balloons every fourth quarter is either seasonality or a plant building ahead to make the revenue line, and the receivables curve will tell you which, because shipped-and-billed shows up there within weeks. On one deal the December inventory spike repeated three years running while December shipments stayed flat. The plant was building parts nobody had ordered so the year would end well. That habit had a cost, and it was sitting on the balance sheet pretending to be an asset.

The payables side leaks just as steadily, and almost nobody reads it. A shop stretching its own vendors to seventy days is financing itself on the backs of the suppliers it depends on, and the suppliers respond the way suppliers always respond, with longer lead times, stricter terms, and the quiet deprioritization that shows up later as missed deliveries the company will blame on the market. Vendor concentration hides here too. A fabricator with one heat treater within economic shipping distance has a single point of failure that appears nowhere in the CIM, and the AP file will show you exactly how that relationship has been treated.

Then rerun the revenue yourself, by invoice, by customer, by month. The pie chart in the book was drawn once, at whatever altitude the story required. The invoice file was drawn ten thousand times by a billing clerk, and it does not know what story it is in.

You can stage a plant tour. You cannot stage the maintenance log.

Paul W. Swaney III

The Floor Keeps Its Own Books

On a manufacturing deal the floor documents outrank everything. The maintenance log is a diary of how the owner really felt about the future. A press that got its rebuilds on schedule until two years ago, then went quiet, gives you the exact date the seller started managing for the exit, and it pairs with the walk-through: grease and fresh paint hide from photographs, and they do not hide from a person standing next to the machine with the log in his hand.

Scrap and rework reports carry the margin story the bridge never mentions. A shop quoting at thirty points and scrapping eight is a twenty-two point shop with good intentions, and the overtime file tells you whether throughput is capability or heroics. Sixty-hour weeks landing in the same three work centers, month after month, is a bottleneck being paid time and a half to stay invisible.

The most underrated file in the building is the quote log. Win rate by customer and by margin band is the pricing power section the CIM never includes. A shop winning ninety percent of its quotes is leaving money on every job it ships. One winning fifteen percent is a bid mill burning estimating hours. And the loyal account that quietly starts sending everything out for requote in the founder's final year is telling you exactly what it thinks of the transition, in a file nobody thought to clean before the process started.

The payroll register deserves an hour of its own, read against the org chart. Hire dates tell you whether the workforce is a pipeline or a cohort, and in the trades it is usually a cohort: five of the six machinists who can run the hard parts hired within four years of each other, all now within sight of retirement, with nobody behind them because the apprenticeship program listed in the CIM consists of a plan to have one. Turnover in the skilled positions over the trailing three years tells you what the shop is like to work in, which no plant tour conducted in front of the owner will ever reveal.

Asking for the Ugly Files

None of this takes forensic tools, but it does take asking, because the operating layer rarely appears in a data room uninvited. The request is a moment worth handling well. Framed as ordinary, which it is, it reads as a buyer who knows manufacturing: the AR aging, maintenance logs on the five biggest machines, the quote log, twelve months of freight bills, scrap by month. Any decently run shop can produce that list in a day, and the banker has seen serious buyers ask for less. Timing matters as much as tone: the list goes in with the first diligence request, before the LOI hardens, while the answers can still shape the price instead of arguing with it.

The response is data all by itself. A controller who ships everything by Friday is telling you the house is in order. Three weeks of silence followed by a summarized version tells you the records are either ugly or absent, and a shop with no maintenance log and no scrap reporting has described its own management system without sending a single number. Outright refusal is rarer than sellers realize, because the banker knows exactly what it signals. I have had one hard no in all my years of asking, and walking away from that deal turned out to be the cheapest diligence I ever ran.

What This Means for a Fundless Sponsor

An institutional process routes this work through a QoE firm, which does it competently and from the outside, testing the bridge because the bridge is what the memo will cite. Freight bills and quote logs rarely make the workpapers. My model lets me sit in the plant, pull the ugly files myself, and read them next to the people who fill them in, and years spent running operations means I know what a healthy scrap number looks like in that process before anyone explains it to me. The committee buyer pays a firm to test the seller's document. I go find the documents nobody prepared.

The operating layer will not price the deal for you. It tells you what is true, and the next step is deciding what true is worth, which means setting the banker's book aside entirely and writing your own. That is the last piece in this series. For now the discipline fits in one line: read the paper that never met the banker.

Paul Swaney is the founder of Swaney Group Capital, a fundless sponsor focused in the lower middle market. LeverUp® is published weekly. More if I have something else to say.