It's late October, which means you're probably somewhere in the middle of next year's budget. If the AFP's numbers hold for you, you'll spend about two months on it; its 2026 FP&A benchmarking survey put the average budget build at 8.7 weeks, up slightly from 8.5 in 2022. And by March, a rate decision or a big customer's guidance cut will have made a chunk of it wrong. That's usually when the rolling forecast vs annual budget question comes up, often from whoever just rebuilt the same model for the third time.
Why the annual budget isn't going anywhere
The famous case for getting rid of the budget is Handelsbanken. Jan Wallander took over the Swedish bank during a crisis in 1970, and in the early 1970s it walked away from central budgeting. Years later he wrote a paper about it called "Budgeting, an unnecessary evil." I like the story a lot. What sticks with me is Wallander deciding what the bank's numbers were for and refusing to let a yearly ritual answer that question for him, which is a separate thing from whether budgets are good or bad.
Most of us don't run a Swedish bank. The lender wants an annual plan, the comp plan pays out against one, and the board that approved it is going to ask in May how you're tracking. So for a mid market company the practical answer is to keep the budget as the commitment and run a rolling forecast as your best current view, which is what most companies with a rolling forecast already do. The AFP found 43% of organizations use rolling forecasts, and the 2025 FP&A Trends Survey found that while 46% had adopted them, only 5% had gone as far as a rolling budget.
One more AFP figure stuck with me: 88% of CFOs said the budget was useful, compared with 62% of business unit leaders. The people being held to the number trust it least, and a rolling forecast is partly a way of giving them one they believe.
How far out a rolling forecast should go
For most mid market companies, five or six quarters is enough, with monthly detail for the next two quarters and quarterly after that. That's close to the hybrid Anaplan suggests. Going much further mostly adds precision you don't have; the same FP&A Trends survey found 61% of companies could forecast with confidence only about six months ahead. An eight quarter forecast can look great in a board deck, right up until someone asks what you assumed for the back half of 2028.
The rule I'd care about more than the horizon is that the forecast never overwrites the budget. Report actuals against both every month, so you can tell the difference between being behind plan and having known since June that you would be.
Driver-based rolling forecasts, with a team of two
A two person team can't reforecast 400 lines a month and shouldn't try. Pick ten to fifteen drivers that move most of the P&L, things like volume by product family, price, a few input costs, headcount, freight, the interest rate and DSO. Forecast those, let the model do the rest, and hold the small lines at run rate. The survey data backs this up fairly strongly: among companies with fully or dynamically driver-based models, 77% rated their forecast accuracy good or better, compared with 27% of those running basic models.
Some of those drivers should come from outside the building. Your revolver rate changed in September. Your customers are telling investors about next year's volumes this month, and steel, diesel and freight move on their own schedule whatever your model says. When we look at what changes our read on a company from one month to the next, it's mostly outside drivers like these, and a forecast built purely from internal run rates ends up being last month's actuals extended forward.
How to implement a rolling forecast in a month
A routine a two or three person team can keep up with:
- Close plus three days: actuals in and locked.
- Days four and five: every driver has an owner outside finance. Sales gives you volume and price, operations gives you throughput and headcount, purchasing gives you input costs. Send each of them last month's number and ask what changed. It should take about fifteen minutes each.
- Day six: roll the forecast forward a month and add a month at the far end, so the horizon stays the same length.
- Day seven: write one page covering actuals against budget, the latest forecast against budget, and the three drivers that moved most since last month, in plain language.
- Once a quarter: save the forecast versions and compare them with what happened. In my experience most forecasts lean the same way quarter after quarter, a bit optimistic on volume and a bit hopeful on price, and you can only correct for that if the old versions are saved somewhere.
A spreadsheet is fine for this at most mid market companies. Software starts to pay for itself once the driver list settles down and more than three people are working in the model. If you're still deciding who owns it, the post on FP&A team structure covers the seat a lot of companies skip, and this one lists the 2027 assumptions most likely to need reforecasting first. A free briefing on your company can supply the outside drivers each month so the forecast doesn't depend on somebody remembering to look.
CFOmarketIQ reads the outside events that move your forecast drivers, from rate decisions to customers' guidance, and the first briefing is free.
Questions CFOs ask about rolling forecasts
Does a rolling forecast replace the annual budget?
Usually not, and for most mid market companies it shouldn't. Lenders, comp plans and boards still want an annual commitment. Run the rolling forecast alongside the budget as your current best view, report actuals against both each month, and keep the forecast from overwriting the budget.
How far out should a rolling forecast go?
Five or six quarters works for most companies, with monthly detail for the next two quarters and quarterly detail after that. Survey data suggests most companies can only forecast with confidence about six months out, so a much longer horizon mostly adds precision that isn't there.
How often should a rolling forecast be updated?
Monthly is the common cadence, done right after the close and rolled forward one month each time so the horizon stays the same length. Some companies update quarterly instead. If your drivers move fast, input costs or interest rates for example, monthly is worth the extra effort.
Can a small finance team run a rolling forecast without FP&A software?
Yes. A spreadsheet built on ten to fifteen drivers, each with an owner outside finance, covers most mid market companies. Software starts to earn its cost once the driver list stops changing and several people need to work in the model at the same time.
A note on all of this: it is general information, not financial, legal or tax advice. Run anything you plan to act on past your own advisors first.