Six dead months

Clockwork makes forecasting software for accounting firms. By July 2026 their outbound email had produced nothing for six months, and the CEO opened our call by saying he’d lost tens of millions in valuation and was thinking about walking.

I didn’t argue. I asked for access to what they already had: their CRM and their recorded sales calls. Nobody had read them as a set. There were 1,089 deals and 617 calls.

Two weeks later I had a different picture of their business than the one they’d been selling into. About 80 percent of revenue came from one kind of customer. Deals closed far more often when the buyer was an owner or partner. Big customers churned at more than twice the rate of small ones, the opposite of what everyone assumed. I also found that an earlier analysis, including one run independently, had overstated churn by a third.

Before I handed it over, I gave the whole thing to two different AI models with no knowledge of each other and published where they disagreed with me: twelve corrections, numbered.

The CEO’s last question on the follow-up call was “what do you guys need from me?”