Example: Rebuilding a Model Three Weeks Before a Series B Term Sheet

A single flagged discrepancy, three weeks before an expected term sheet, is close to the worst possible moment to discover that a financial model no longer reflects the business. This company got that discovery anyway, and had five weeks to fix it before the round could close.

Background

The company builds supply chain visibility software for mid-market manufacturers and distributors, the kind of business that grows in fits and starts as individual enterprise contracts land rather than in a smooth monthly curve. It had raised a modest Series A about eighteen months prior, grown revenue several times over, and moved through two pricing changes and a fundamental shift from seat-based to usage-based billing along the way, each one a reasonable response to what customers actually wanted to pay for. None of those changes had been fully reflected in the financial model that the team kept presenting to the board and, eventually, to Series B investors.

A model that drifted quietly, then loudly

For a while, the drift didn’t matter. The Series A model had been built to a reasonable standard, and the business was growing quickly enough that small inconsistencies stayed buried under bigger, more obviously good numbers. But by the time Series B diligence began in earnest, the gap had compounded: the logo-count metric in the pitch deck, carried forward from an earlier internal tracker, no longer matched the customer list sitting in the data room, off by four logos, roughly 13% of the total. A sharp-eyed investment associate found the mismatch within the first week of formal diligence, and the tone of the conversation shifted immediately from enthusiastic to skeptical.

Rebuilding from the general ledger, not the old spreadsheet

Patching the existing model to fix the specific number the associate had flagged would have been the fast option, and it was tempting given the clock was already running. But it would have left the same structural problem in place: a model built on assumptions nobody could point to a source document for. Instead, the model was rebuilt from the general ledger up, reconciling every revenue line to actual invoiced usage rather than the sales team’s pipeline estimates, and rebuilding the customer cohort and retention analysis entirely on the current usage-based pricing structure instead of the legacy seat-based one. Every number in the new model could be traced, on request, back to an invoice or a signed contract addendum.

Rehearsing the hard questions before they were asked live

Before the next scheduled investor call, two mock diligence sessions were run with the CEO and the Head of Sales, structured around the kind of pointed, slightly adversarial questions a real investment associate asks when they are actively testing whether a founder’s answer holds up under a follow-up. Two of the specific questions rehearsed in those sessions came up almost word for word in the actual call that followed, giving the CEO a calm, specific, already-rehearsed answer instead of an improvised one delivered live under pressure.

What this meant beyond the immediate fix

The deeper lesson for the company’s finance function wasn’t just that the model needed a rebuild, it was that the underlying process for updating the model after a pricing or revenue-model change needed to become a standing discipline rather than a one-time cleanup. A quarterly reconciliation step, checking every headline metric used externally against its underlying source data, was put in place alongside the rebuild specifically so the same kind of drift couldn’t quietly reaccumulate before the next round.

Results

The term sheet was signed five weeks later, at the top of the valuation range the company had been modeling internally before the discrepancy surfaced, with no further metric inconsistencies raised during the remainder of diligence.