The forecast got worse when it got bigger

A CFO I worked with had a forecast that started as one tab and four assumptions. Three years later it had eleven tabs and fifty-seven assumptions, and the variance to plan was wider than it had ever been.

The model didn't get smarter. It got more confident. Each new line — vendor cohort, headcount step-up, churn-by-segment — was another guess wearing the costume of a calculation. Variance compounds. Eleven tabs of compounded guesses produce a forecast that is precisely wrong instead of approximately right.

The one-tab forecast was honest. Four numbers, and the CFO carried them in her head. When something moved, she felt it within a week. The eleven-tab forecast was no longer felt; it was operated. The board got a deck instead of a person who could answer in real time.

The point of a forecast is not to be comprehensive. It is to be re-buildable from memory in five minutes. The moment the forecast can no longer be defended without opening the file, it has become accounting, not steering.

When she went back to one tab — same business, fewer lines — variance to plan dropped by half the next quarter. Not because the new model was smarter. Because she had stopped fooling herself.

The test: if the forecast can't be sketched on a napkin in front of an investor and survive without laughter, the model is too big.

A forecast that needs a second monitor to read isn't a forecast. It's bookkeeping with ambition.