The same business, two numbers
Finance forecasts revenue. Operations forecasts units. They are produced by different people on different cycles, and they disagree. Each month someone reconciles them by hand and the difference is explained away.
Before assuming a modelling problem, check whether they are even measuring the same thing. In our experience the gap is definitional far more often than analytical.
The four places definitions diverge
| Dimension | Finance usually means | Operations usually means |
|---|---|---|
| Event | Invoice raised | Goods despatched or produced |
| Timing | Accounting period, recognition rules | Calendar or trading week |
| Unit | Net value after discount | Gross units or cases |
| Scope | Includes services, excludes intercompany | Includes samples and replacements |
Any one of these produces a persistent gap that no amount of model tuning will close. Two of them together produce a gap that moves month to month and looks like forecast error.
Build a bridge, not a single number
The realistic goal is not one forecast serving both audiences. It is one underlying forecast plus a documented, repeatable bridge between the two views.
- Agree the base event - usually the operational one, because it happens first and is physically observable.
- Write each adjustment as an explicit step: units to gross value, less expected discount, less expected returns, timing shift to recognition.
- Make each step's assumption visible and owned by someone, with its own history so it can be checked.
- Reconcile automatically every cycle, so a growing gap raises a flag rather than being absorbed by whoever does the spreadsheet.
Once the bridge exists, the argument changes from 'whose forecast is right' to 'has the returns assumption drifted', which is answerable.
Who should own the forecast
A single owner for the base forecast, with both functions contributing, avoids the two-forecast problem returning within a quarter. That owner needs the authority to say no to adjustments made for comfort rather than evidence.
Forecasts adjusted upward because a target requires it are no longer forecasts, and everyone downstream learns to discount them. If a stretch target exists, hold it separately from the forecast rather than inside it.
If the forecast moves to match the target, you have two targets and no forecast.