Rolling forecast vs annual budget: how to run each one

A rolling forecast always looks the same distance ahead, 12 or 18 months, adding a month as each month closes. That matters because most annual budgets are out of date within a few months: a large deal slips, a hire starts late, churn ticks up, and the budget still says what it said in January. Leadership then makes decisions from a plan nobody believes, or from a bank balance and instinct.

A rolling forecast fixes that without throwing the budget away. It is one of the core tools of FP&A, and it works best alongside the budget, not in place of it.

Definitions: budget, forecast and rolling forecast

Annual budget. The company's financial plan for the fiscal year: revenue targets, headcount by team, operating spend, profit and cash. It is agreed before the year starts, approved by leadership and usually the board, and then held fixed.

Forecast (or reforecast). An updated estimate of how the year will actually end, based on results so far and what has changed. A reforecast usually still ends at fiscal year end, so by November it barely looks ahead.

Rolling forecast. A forecast that always looks the same distance ahead. Each period, actual results replace the forecast for the period just closed, the outlook for the remaining periods is updated, and a new period is added at the far end.

Rolling forecast vs budget: what each is for

Why keep both

Some advice treats the rolling forecast as a replacement for the budget. For most growing companies, keeping both works better.

The budget is a commitment. It sets the targets leadership agreed to, the headcount each team may hire, and the baseline for bonuses and commissions. If it moved every month, nobody could be held to it.

The forecast is a best estimate. Its job is to be accurate, not to be met. Hiring, spending and fundraising decisions should be made from the forecast, because it reflects what is actually happening.

The useful report puts them side by side: actual, budget, latest forecast. The gap between budget and forecast is the honest answer to "will the company hit its plan?", and it tells leadership how much needs to change to close it.

Choosing a horizon and cadence

There is no single correct design. Each choice trades effort for visibility.

Horizon

• 12 months. The simplest option and enough for most decisions about hiring and spend. Its weakness is that by mid-year it barely reaches into next year's plan.
• 18 months. Always shows at least part of the next fiscal year, which helps with annual planning and fundraising timing. It takes more effort, and the outer months are less reliable.
• Current year plus next year. Some companies extend to the end of the following fiscal year instead of a fixed window. The horizon varies from 13 to 24 months, but every forecast lines up with fiscal years.

A company with tight cash should also run a weekly 13-week cash flow forecast alongside, whatever the main horizon.

Cadence and granularity

• Monthly updates keep the forecast current and fit naturally after the month-end close. They work when the forecast is built on drivers, so the update takes a few days and is ready shortly after the close. If it takes one to two weeks of manual work, the model or the tools need attention (see FP&A software).
• Quarterly updates take less effort and suit businesses with stable, predictable results. The forecast can be up to three months stale when a decision comes up.
• Mixed granularity is common: monthly detail for the first six months or so, quarterly beyond that. Precision in month 17 is not real, so there is little point building it.

Driver-based forecasting

Driver-based forecasting means modeling the few operating inputs that move the numbers, rather than forecasting each general ledger account. When a driver changes, revenue, cost and cash change with it. It is what makes a monthly rolling forecast practical.

Revenue drivers

For a subscription business, the core revenue drivers are new bookings, expansion, churn and price. New bookings in turn come from capacity: how many sellers, how many deals each closes, and at what price.

For example, if five ramped sellers each close two deals a month at an average of $4K in monthly recurring revenue (MRR), new MRR is 5 × 2 × $4K = $40K a month. If a seller leaves, the forecast changes through the driver, not through someone editing a revenue cell.

Headcount drivers

People are usually the largest cost. The headcount plan lists every current and planned role with a start date and a fully loaded monthly cost (salary, payroll taxes, benefits and other costs that scale with people). Two engineers starting in September at $15K each fully loaded add $30K a month from September.

Cost drivers

Most other costs tie to a driver too: hosting as a percentage of revenue or per customer, software and equipment per head, recruiting per hire, marketing tied to pipeline targets, and rent as a fixed amount that changes on known dates. A good test: can every material line be explained by a driver a manager understands? If a line can only be explained as "last year plus 10%", it is not driven yet.

How to run a monthly rolling forecast, step by step

1. Close the books. The forecast starts from final actuals, so a late close delays everything.
2. Load actuals. Replace the forecast for the month just closed with actual results, including cash and customer counts.
3. Compare drivers. Look at deals, price, churn, start dates and spend, not just totals, and find which driver caused each meaningful variance.
4. Decide whether each variance is timing or permanent. A deal that slipped a month is timing. A lost deal or a lower win rate is permanent and changes future months.
5. Collect driver updates from owners. Sales leadership updates pipeline and hiring capacity; department heads confirm hiring dates and committed spend. Ask for drivers, not dollar totals.
6. Update the remaining months and add the new month. Roll the window forward so the horizon stays the same length.
7. Check cash and runway. Run the updated profit and loss through to cash. Check covenants if the company has debt.
8. Review with leadership. Show actual, budget, prior forecast and updated forecast, with a short explanation of what changed and why, and the decisions it raises.
9. Save the version. Keep every forecast instead of overwriting it. Old versions are how accuracy gets measured.

Worked example: one month of actuals

A software company starts September with $500K of MRR. The prior forecast assumed 10 new deals a month at $4K MRR each ($40K of new MRR) and $5K of churned MRR a month, so MRR grows by $35K a month. Here is September against that forecast.

Illustrative figures, not client data. Operating expense difference shown as favorable.

Finance works through the drivers. Win rates held, but two sellers hired over the summer are ramping more slowly than planned, so sales leadership lowers new deals to 9 a month for October through December, then back to 10 from January as they reach full productivity. The $30K expense underspend came from two engineering hires who will now start in November instead of September.

The updated forecast rolls forward one month: September becomes actual, and September 2027 is added at the end.

Illustrative figures, not client data. Selected months shown.

Three things stand out. The September miss is not just $12K; slower ramp adds another $4K of shortfall in each of the next three months, so the gap settles at $24K of MRR. If the budget called for $660K of MRR in December, the honest year-end view is now $44K short, and leadership knows it in October rather than in January.

The expense side needs care too. Moving two hires from September to November saves $60K in cash this year (two months at $30K), but it is a timing difference, not a permanent saving. If those engineers were meant to ship a feature sales is counting on, the delay may affect revenue later, and the forecast should say so.

Common mistakes

• Too much detail. Forecasting hundreds of accounts makes each update slow without making it more accurate. Forecast the drivers and the material lines; let small accounts follow simple rules.
• Forecasting to the budget. A forecast that matches the budget until the last possible month is not a forecast. Budget owners need to know that an honest miss in the forecast is useful, while a surprise at year end is not.
• Sandbagging. The opposite problem: owners pad their forecasts so they can beat them. Cushions stack up across teams, and leadership holds back hiring or spend it could afford.
• No accountability for accuracy. Without comparing old forecasts with actual results, there is no feedback. Track accuracy for new bookings, ending recurring revenue, operating expenses and cash, and review it each quarter.

How an investor reads your forecast

Investors, lenders and acquirers rarely take a forecast at face value. They judge it partly by how well the company's earlier forecasts held up.

• Forecast accuracy over time. In diligence, a common request is a company's past forecasts next to its actual results. A company that landed within a reasonable range quarter after quarter earns trust in its current projections. One that missed widely, or kept moving the goalposts, gets its forecast discounted.
• Explained misses. Every company misses sometimes. What matters is whether each miss was explained by a driver and corrected in the next forecast, or quietly absorbed.
• The bridge from budget to forecast. A clear walk from budget to latest forecast, driver by driver, shows a finance function that understands its business.
• Drivers that match reality. Investors test whether assumed win rates, ramp times, churn and hiring pace match recent history. A forecast that requires sudden improvement on every driver at once is easy to spot.

The practical lesson: keep every version, measure accuracy, and record why each change was made. A history of honest, explained forecasts is worth more in a raise or a sale than an optimistic model. Decipher's FP&A Lead engagement builds and runs this process alongside an existing CFO or controller, and the finance scorecard shows where the gaps are today in about two minutes.