Rolling forecast: how a 12-month plan stays current

A rolling forecast keeps a constant-length view of the future. When 1 month closes, the forecast replaces that month with actual results, updates the assumptions that changed, and adds another month at the far end.

A 12-month rolling forecast therefore continues to show the next 12 months. It does not count down toward December and become less useful as the year passes.

Each update requires finance judgment about which forward assumptions the new actuals should change. The recurring mechanics are a natural job for a finance agent: load the evidence, update the Model, explain what moved, and bring the judgment call to a person.

How a rolling forecast works

Assume the current forecast covers January through December. After January closes, the team imports January actuals, explains the material differences, revises the drivers affected by those differences, and adds the following January. The updated horizon runs from February through January.

New forecast horizon = 11 remaining periods + 1 new future period

The same method works with 13 weeks, 18 months, or another horizon. Choose a period long enough to see the decisions that require lead time. A company planning hiring and cash may use 12 or 18 months. A near-term liquidity process may use 13 weeks.

The horizon and update cadence are separate choices. A 12-month forecast can update monthly. A 13-week cash forecast may update every week.

A connected source can make new actuals available throughout the month. Finance advances the rolling forecast after reconciling the closed period and deciding which forward assumptions the new evidence supports.

A rolling forecast and an annual budget have different jobs

The annual budget records the plan the company approved. It provides a stable baseline for goals, resource allocation, and accountability. A rolling forecast records what the company currently expects.

Keep both. If every new result rewrites the budget, the team loses the original commitment. If the forecast never changes, it stops helping people decide what to do next.

This distinction makes variance analysis more useful. Compare actuals with the budget to understand performance against the commitment. Compare actuals with the prior forecast to understand forecast quality. Then update the rolling forecast to reflect the information the business has now.

A worked monthly example

A subscription company begins January with $1,000,000 in monthly recurring revenue. Its January forecast assumes $100,000 of new MRR and $40,000 of churned MRR:

January forecast ending MRR = $1,000,000 + $100,000 − $40,000 = $1,060,000

January closes with $80,000 of new MRR and $50,000 of churned MRR:

January actual ending MRR = $1,000,000 + $80,000 − $50,000 = $1,030,000

The company starts February $30,000 below the level assumed in the old forecast. If the February assumptions remain $100,000 of new MRR and $40,000 of churned MRR, the updated February result becomes:

February rolling forecast ending MRR = $1,030,000 + $100,000 − $40,000 = $1,090,000

The old February forecast would have ended at $1,120,000. The $30,000 gap carries forward because February starts from the January actual. This example treats churn as a fixed dollar amount. A model using a churn rate would recalculate churn against the updated customer or revenue base.

If the company keeps the same fixed monthly assumptions for the full horizon, the rolling forecast becomes:

This table extends the arithmetic. It is not a claim that the assumptions should remain unchanged for 12 months.

That is the mechanical update. Finance still needs to decide whether January changed the forward assumptions. If lower new business came from 1 delayed contract, the February sales assumption may remain reasonable. If pipeline conversion has weakened across the board, keeping $100,000 would ignore the new evidence. The forecast should preserve that judgment as a visible assumption.

What changes when a month closes

Use a repeatable close-to-forecast process:

1. Load and reconcile actuals. Replace the closed period's forecast values with actual results, then confirm that important totals agree with the accounting records.
2. Explain material differences. Separate timing, volume, price, mix, headcount, and one-time items where those distinctions change the next decision.
3. Update affected drivers. Change the assumptions supported by the new evidence. Avoid carrying every one-time variance through the remaining horizon.
4. Advance the horizon. Add a new period with explicit assumptions. Check each carried-forward driver before using it again.
5. Compare the new forecast. Show the difference from the prior forecast and the original budget, with an owner for unresolved questions.

The process should focus attention on the assumptions that can change the plan. Re-entering every line each month creates work without improving the forecast.

Where a finance agent changes the process

The monthly cycle begins when new actuals arrive. A finance agent can reconcile the closed period, extend the horizon, trace the material variances, and prepare the affected forecast drivers for review.

In the example above, the important result is the $30,000 MRR gap. The agent should trace $20,000 to lower new business and $10,000 to higher churn, then show how the gap flows through the remaining forecast. If the sales miss came from 1 delayed contract, the current forward assumption may still hold. If conversion weakened across the pipeline, the forecast needs a different assumption. The agent should expose that choice and the evidence behind it.

The work should continue after the monthly update. A useful trigger can bring the team back when a material variance appears, a hiring plan no longer fits the cash forecast, or new data changes the expected outcome. The message should identify what moved, what it changed, and which model needs review.

Which assumptions deserve monthly review

Review the drivers that have new evidence or a meaningful decision attached to them. Common examples include:

• sales pipeline, conversion, contract timing, and expansion
• churn, retention, and pricing changes
• hiring dates, compensation, and contractor plans
• vendor commitments and payment timing
• gross-margin drivers and usage-sensitive costs
• financing timing, covenant limits, and the company's cash floor

Some assumptions change quarterly or only after a formal decision. Name their review cadence. A forecast becomes easier to operate when the team knows which inputs are live, who owns them, and what evidence permits a change.

Common rolling-forecast failures

Common failures include:

• Replacing the approved budget with the latest forecast. Preserve the original baseline and show the rolling forecast beside it.
• Extending a one-time variance through every future period. A delayed payment can move cash between months without changing annual revenue, and an unexpected legal bill may have no recurring effect.
• Adding detail without ownership. Keep the detail that changes decisions and assign an owner to each material input.
• Updating a number without recording the reason. Identify the changed driver, the supporting evidence, and the person responsible.
• Presenting the forecast as certainty. Record the current expectation under named assumptions that the team can revise.

What forecasting software should support

The mechanics should be boring. Good forecasting software should preserve the budget, import actuals into the closed period, extend the horizon, and let the team change a driver once without rebuilding every output.

During a software evaluation, ask the vendor to demonstrate 1 complete cycle:

1. Close a sample month.
2. Replace that period with actuals.
3. Explain a material variance.
4. Change 1 forward driver.
5. Add a future month.
6. Compare the updated forecast with both the prior forecast and budget.

The demonstration should show where the values came from and which formulas changed. A dashboard that refreshes without exposing the model gives the team a new number, but little additional understanding.

Rolling forecasts in CFO.ai

In CFO.ai, `Last close` separates closed Actuals from the Forecast. A Variable can use imported data in Actuals and a planning formula in future periods while keeping the same business meaning across the Model.

Integrations bring accounting, HR, CRM, spreadsheet, database, and uploaded-file data into the Model. A connected source must finish loading before its data is available, and a saved query may intentionally include only part of the source. Those boundaries should be checked during the close.

Ari can do the recurring work against that shared Model: inspect the new actuals, trace a variance, update the forecast, and leave the formulas and sources available for review. When the team wants to test a change without altering Main, Ari can create a Scenario and compare the alternative with the baseline side by side.

Ari is designed to keep the work current and speak first when something material moves. A proactive check-in should start from the changed number and the decision it affects. The Model remains the place where a person verifies the evidence, changes an assumption, or corrects Ari's judgment.

Use the forecast to record changed expectations

Each update should connect new evidence to the assumptions that shape hiring, spending, revenue, and cash.

Keep the budget fixed, let actuals replace the closed period, and record why each material forward assumption changed. The next review can compare the new evidence with the reasoning already in the Model, while Ari carries the update cycle forward.