Business Planning Baseline Generation

Business Planning Baseline generation is used to create a starting point for planning based on historical ledger data. The generated baseline can either serve as a reference point for planned values but can also be incorporated directly into the plan as a manual planning transaction.

Analyzing Historical Data

Before generating the baseline, users can analyze historical balances of combinations of planning dimensions. This analysis (Business Planning Baseline KPI Analysis) helps determine which combinations are most suitable for creating an accurate baseline.

Generating a Baseline

A baseline can be generated from either:

Planning Unit page – Select one or multiple planning transaction(s) and access the baseline command.
Planning Units page – Select one or multiple planning unit(s) and access the baseline command.

When the Generate Baseline assistant completes, historical balances from the selected ledger and business plan are sent to The Forecasting Service to produce the baseline.

Forecast Models

The Forecasting Service supports seven model variants for calculating the baseline:

Adaptive exponential smoothing with level
Bayesian
Brown’s smoothing with level and trend
Exponential smoothing with level
Exponential smoothing with level and trend
Naive
Regression line – Least squares

For each planning dimension combination, the forecast service evaluates which model best fits the data. The evaluation compares how well each model forecasts the last 12 periods of available ledger data, using MAE (Mean Absolute Error) as the accuracy metric.

The winning model is then used to generate the baseline for the selected output periods. Results are displayed as a separate planning transaction with source Business Planning Baseline. Alongside generated balances, the system also presents:

The selected model
The WMAPE (Weighted Mean Absolute Percentage Error) score. The WMAPE metric measures how well the selected model forecasted the last 12 periods of available data. Unlike MAE, which was used for selecting the winning model, WMAPE allows for comparing forecasting quality across different time series, not just within a single series.

Best Practices and Limitations

Use at least 24 periods of input data (preferably more) to ensure realistic accuracy.
Forecasts are only reliable if historical data reflects future developments.
If known changes (e.g., trend shifts, volume jumps) will occur, users must manually adjust the forecast.

Multi-Company Planning

For multi-company plans:

All companies must use the same accounting calendar.
Companies must be connected to a planning entity in the plan-holding company.

High-Level Planning

Users can plan at a higher aggregation level by setting selected code parts as “not used” in the planning unit. In this case, ledger balances for unused code parts are aggregated before being sent to the forecast service.

Baseline Processing and Cost

Baseline is processed in batches of 200 combinations of planning dimensions.
Each batch of 200 combinations incurs a fixed cost.