Turning market noise into investment conviction.

The spreadsheet rarely fails all at once. It fails one symptom at a time: the month-end close that used to take three days now takes ten, two people can't have the workbook open together without a version conflict, and nobody can say with confidence which tab is the "real" one anymore.

The trigger isn't a revenue number — plenty of $20M businesses run fine on spreadsheets, and plenty of $5M businesses have already outgrown them. The real signals are structural: you're closing books manually across more than one entity or location, an investor or lender is asking for numbers faster than your process can produce them, or a key person leaving would mean nobody else understands how the model actually works.

When two or more of those show up at once, the fix isn't a bigger spreadsheet. It's a reporting cadence, a chart of accounts that actually maps to how the business is run, and a system that survives someone taking a vacation.

Almost every deck leads with a TAM slide, and almost every TAM slide has the same flaw: it's a market research figure with the company's logo stapled on top of it, not a number that describes what this specific business can actually sell into.

The tell is a TAM that never gets broken down. If a deck can't walk from the analyst-report headline number down to a serviceable segment, and from there down to what's actually reachable given the current product and go-to-market motion, the top-line figure is doing narrative work, not analytical work.

The fix is boring but effective: rebuild it bottom-up from unit economics — number of reachable accounts, realistic win rate, realistic contract value — and treat the analyst number as a sanity check, not a foundation. If the two numbers are wildly apart, that gap is usually the most important line in the whole memo.

Not all expert calls carry the same weight, but most diligence memos quote them as if they do. A former VP who left three years ago and a current regional manager who touches the product weekly get the same one-line citation — and that's how a stale, biased, or simply wrong opinion ends up shaping a thesis.

A useful filter has three questions: how recent is this person's direct exposure to the thing they're describing, do they have something to gain or lose by how they characterize it, and does anyone else independently say the same thing without being prompted. An observation that passes all three is evidence. An observation that fails one or more is a data point to note, not a conclusion to lean on.

In practice this means weighting corroborated, recent, low-incentive observations heavily, and treating single-source, secondhand, or clearly motivated commentary as color — worth including, but flagged as such in the memo rather than presented with false confidence.

Operators tend to watch net revenue retention because expansion from existing accounts can mask a real problem underneath it. A business can post 108% NRR while quietly losing a third of its logos every year — the number looks healthy right up until the accounts big enough to expand stop being there to expand.

The investor's read starts by splitting the blended figure into its parts: gross logo churn, gross revenue churn, and expansion, tracked by cohort rather than as a single monthly blend. A cohort view answers the question a blended number can't — is churn concentrated in a specific segment, price point, or onboarding vintage, or is it broad-based across the customer base.

The practical rule: never accept a single retention percentage as the finding. Ask for the cohort table behind it, and treat any refusal or delay in producing one as a finding in itself.

Most deals that die in confirmatory diligence don't die from one dramatic discovery — they die from a pattern of small inconsistencies that add up to the same conclusion: the numbers in the room aren't the numbers in the business.

The first flag is customer concentration that wasn't disclosed at the size it actually exists — not just a top-five list, but what happens to the growth story if the largest account doesn't renew. The second is revenue recognition that shifts under examination: deals booked before delivery, contracts with side letters that change the economics, or renewal terms that were quietly renegotiated ahead of the process. The third is a gap between the reported financials and the audited or bank-facing ones that the seller can't fully explain on the first attempt.

Any one of these is a question. Two of them together are a reason to slow down. All three, in the same data room, are usually the deal telling you something the deck never did.