The argument is usually framed as a choice. It is not. They do different jobs, and the businesses that run well run both, with one of them taking about ninety minutes a month.
What each one is for
An annual budget is a commitment. It is set once, approved, and then held still deliberately, because its purpose is accountability. Variance against it is the measurement.
A rolling forecast is an estimate. It is updated monthly, always looks the same distance forward, usually twelve or eighteen months, and its purpose is decision support. It has no memory and holds nobody to anything.
Confusing the two is what produces the familiar dysfunction: a budget quietly re-forecast 4 times until it agrees with reality, at which point it has stopped measuring anything.
| Annual budget | Rolling forecast | |
|---|---|---|
| Purpose | Accountability | Decision support |
| Set | Once, before the year | Every month |
| Horizon | To year end, shrinking | Constant, 12 to 18 months |
| Changes | No | Yes, by design |
| Answers | Are we on plan? | What happens if we do this? |
| Effort | Weeks, annually | Around 90 minutes, monthly |
What a board actually reads
In our experience a board reads 3 things in a pack and skims the rest: cash, the forward view, and the variance on the 2 or 3 lines that are moving. The budget matters to the third of those. The forecast matters to the first two.
What a board ignores is a full-year budget variance table in month two, which contains almost no information, and a re-forecast that arrives without an explanation of what changed since the last one. The change is the content. The numbers are the evidence for it.
A forecast that arrives without a note saying what moved and why is a spreadsheet, not a forecast.
How to run both without doing the work twice
The trick is that both come out of the same driver-based model. If your model is built on volume, price, headcount and ramp, then the budget is one locked scenario and the forecast is the live one. You are not maintaining 2 files. You are maintaining one file with a frozen column.
- Build the drivers once. Whatever actually causes revenue and cost in your business, and nothing else.
- Freeze the budget scenario at approval and never touch it again, including when it becomes embarrassing. Especially then.
- Update actuals monthly, then roll the forecast horizon forward one month so it always looks the same distance ahead.
- Report 3 things: actual versus budget on the lines that moved, the new forward view, and one paragraph on what changed.
Ninety minutes a month is realistic once the model exists. It is not realistic if the forecast is a separate spreadsheet somebody rebuilds each time, which is how most of them are run.
When a budget stops being useful
A budget earns its keep when the business is predictable enough that variance means something. In a company doubling every year, a full-year budget set in January is a work of fiction by April and holding people to it damages judgement rather than sharpening it.
Those companies should still set one, because investors and lenders ask for it, but they should treat the rolling forecast as the operating document and say so out loud in the board pack. The failure mode is pretending the budget is still live while everybody privately runs on the forecast.
The cash version, which matters more than both
Neither instrument is a cash forecast. A P&L forecast can look comfortable while the bank account does not, because revenue recognised in March is frequently cash in June.
A 13-week cashflow, built on collection and payment dates rather than months, is the thing that actually prevents surprises. It sits underneath both the budget and the forecast and takes far less time to maintain than either.
One thing to try this month
Take your last board pack and delete the full-year budget variance table. Replace it with 3 lines: cash today, cash in 12 months on the current forecast, and the single largest thing that changed since last month. If nobody asks for the table back, you have your answer.