Your forecast is a grid, not a model

Almost every startup forecast we are handed in a finance audit has the same defect. It looks like a model — tabs, formulas, a summary page — but the numbers are typed in. Revenue for month fourteen is a value somebody chose, not an output of anything.

The test is simple and slightly cruel. Ask the owner to raise sales headcount by two people starting in March and show you the effect on cash. If the answer takes more than about ninety seconds, the model is a grid.

A driver-based model inverts the structure. You do not forecast revenue; you forecast the things that produce revenue. Reps hired, ramp time, quota, attainment, win rate, average contract value, churn. Revenue falls out. Cost of sales falls out of headcount. Cash falls out of both, plus collection timing.

The immediate payoff is speed of iteration, and it changes the character of planning meetings. When a scenario takes ninety seconds instead of an afternoon, people ask more questions, and the questions get better. Nobody asks the fourth follow-up when each one costs someone a day.

The second payoff is diagnostic. When a driver-based forecast misses, you can see which driver missed. Revenue came in eleven percent light because attainment was sixty-two percent against a plan of seventy-five, not because of some undifferentiated shortfall. That is an actionable finding; "we missed revenue" is not.

The third payoff shows up in diligence. An analyst who can trace your revenue line back to a hiring plan and a conversion rate believes the model. One who finds a hardcoded number in month fourteen starts checking everything else, and that is an expensive mood to create.

Building one is less work than it sounds — usually two to three weeks for a company under fifty people. The hard part is not the mechanics. It is agreeing on which eight or ten drivers actually run the business, which is a management conversation wearing a spreadsheet costume.