A forecast that updates when the actuals do
The rolling forecast was adopted to replace an annual budget. It became four annual budgets, each produced under the same time pressure.
- Rolling forecasts fail operationally rather than conceptually, because each cycle requires the same manual assembly.
- The assembly work is loading actuals, reconciling to the ledger and rebuilding schedules, not the judgement about what happens next.
- Where actuals load automatically, the forecast can update on close rather than on a quarterly project cycle.
- The value is in the reforecast being cheap enough to run when something changes rather than when the calendar says so.
- Version control and a record of what changed between versions are what make the forecast reviewable.
Most companies that adopted rolling forecasts describe them the same way afterwards. The concept is right and the execution consumes the team.
The reason is rarely the forecasting itself. It is the assembly: extracting actuals, reconciling them to the ledger, rebuilding schedules that broke when the chart of accounts changed and chasing inputs from departments.
By the time the forecast is ready to discuss, the period it started from is six weeks old and the assumptions have moved. The meeting then spends its first half establishing what has changed since the forecast was built, which is time taken from the decisions the forecast was produced to support.
The fix is not a better planning tool in the first instance. It is removing the assembly step so the forecast updates when the actuals do. That also changes what the FP&A team is for, since a team that spends most of its time assembling is not doing the work it was hired to do, and the capability erodes as people leave for roles where the analysis is the job. Retention consequences of that kind are rarely modelled, and they are real, as the reviewer shortage and the loss of training work both describe.
Where the cycle time actually goes
In most FP&A teams the majority of a forecast cycle is spent on data preparation rather than on analysis or judgement.
Actuals have to be extracted and mapped, prior versions reconciled, departmental inputs gathered and schedules repaired where something changed in the underlying structure.
Only after all of that does anyone consider what the numbers mean or what should be assumed about the periods ahead.
That ordering is why forecasts arrive late and why the analysis is the part that gets compressed, which is the same pattern described in preparing the first draft mechanically.
Four things that break every cycle
These are the recurring failures that turn a rolling forecast into four annual budgets.
What continuous actually means
A rolling forecast is continuous when a reforecast can be produced because something changed, rather than because the quarter ended.
That requires the marginal cost of a reforecast to be low, which in turn requires actuals to load without manual work and schedules to survive structural change.
Once that holds, the forecast becomes a management tool rather than a reporting deliverable, because it can answer a question at the time the question is asked.
The judgement layer, which is where the value is, does not become cheaper. It becomes the only part anyone spends time on.
Continuous does not mean frequent. It means cheap enough to run when something changes.
Three levels of forecast maturity
Most companies describing themselves as rolling are on the second level.
| Level | How it works | Cycle time |
|---|---|---|
| Periodic rebuild | Each forecast assembled largely from scratch on a quarterly cycle. | Weeks, and the analysis is compressed at the end. |
| Templated rebuild | Stable templates with manual actuals loading and input chasing. | Days to weeks, dependent on the slowest submitter. |
| Actuals driven | Actuals load on close, schedules update, inputs are exceptions. | Hours, and reforecasting on demand becomes possible. |
Forecast methodology and any external commitments based on it should follow your own governance. Changing forecast cadence can affect covenant reporting and investor communication.
The forecast is late because the assembly is manual, not because the judgement is hard.
ChatFinRemoving the assembly
Five steps, in the order that produces the largest reduction in cycle time first.
Automate the actuals load
Directly from the ledger on a defined mapping, which removes both delay and transcription risk.
Make the mapping resilient
So a new account defaults somewhere sensible and is flagged, rather than silently disappearing from a report.
Convert inputs to exceptions
Departments confirm or amend a prepared baseline rather than submitting from scratch, which removes most of the chasing.
Version every forecast
With a record of what changed and why, so the review can be about assumptions rather than about reconciliation.
Measure cycle time
From period close to forecast available, because that is the number that determines whether it is a management tool, in the way close metrics work.
Two traps in forecast redesign
Both are common in projects that start with a tool selection.
Buying a tool first
A planning tool with manual actuals loading and unresolved mapping problems reproduces the existing cycle time in a new interface, at additional cost.
More detail as an improvement
Forecasting at greater granularity increases assembly work and rarely improves accuracy, since the error is usually in assumptions rather than in arithmetic, as driver based planning shows.
How ChatFin supports the forecast cycle
We work on the assembly side, which is where the cycle time is, rather than on the forecasting judgement.
Assumptions, judgement and the forecast itself remain with your FP&A team, and nothing about the numbers is asserted without a person.
AI Cost Estimator for the CFO Office
Four inputs. A first year range, a timeline, and three delivery models compared. No login required.
Frequently Asked Questions
Why do rolling forecasts revert to periodic ones?
Because each cycle costs almost as much as a budget, so the organisation runs them as infrequently as it can. The cost is in assembly rather than in judgement, which means the fix is removing the assembly rather than simplifying the forecast.
Do we need a planning tool?
Eventually, perhaps, and buying one before the actuals loading and mapping problems are resolved usually reproduces the existing cycle time in a new interface. Fix the data flow first and the tool decision becomes clearer and cheaper.
How often should we reforecast?
When something material changes rather than on a fixed calendar. That is only possible when a reforecast is cheap, which is the point of removing the assembly work. Fixed quarterly cycles are a symptom of expensive reforecasting.
What should we measure?
Cycle time from period close to forecast available, and forecast accuracy by horizon. The first determines whether the forecast can inform decisions, and the second determines whether it should.
The judgement is not what makes the forecast late.
Automating the actuals load, making the mapping resilient and converting departmental inputs to exceptions is what turns a quarterly project into a continuous capability.
If you want a first year range for automated actuals preparation and variance analysis, the estimator will produce one before any scoping.
Open the estimatorBook a Demo
See what this looks like in your ERP.
We will walk through the workflow behind this article on the system you already run, using your own data, and show what a first phase would actually cover.
This article is general information for finance and accounting teams. It is not legal, tax, accounting or investment advice. Dates and thresholds are given as published at the time of writing and several of these timelines have moved before, so confirm the current position with your advisers and with the relevant authority before acting.