Audit adjustments with FVA
Use the FVA (Forecast Value Added) module to determine whether each phase's manual adjustments improved or worsened accuracy relative to the AI-generated Baseline.
To understand what FVA measures and the logic behind it, read What is FVA and the 50/50 Rule? first.
Before you begin
- FVA is calculated only on closed versions. Make sure the cycle you want to audit is already closed.
- You need access to the FVA module (check your role in Roles and permissions).
Steps
1. Open the FVA module
From the main menu, navigate to the FVA module. The view shows a comparative table by phase and SKU (or family, depending on the active filter).
2. Read the side-by-side metrics table
For each audited phase (Commercial, Marketing, Consensus…), the system shows in parallel:
| Column | Description |
|---|---|
| Baseline (AI) | Accuracy metrics for the forecast without human intervention |
| Adjusted phase | Accuracy metrics after that phase's adjustment |
| BIAS | Average bias (positive = overestimation, negative = underestimation) |
| MAE | Mean Absolute Error |
| MAPE | Mean Absolute Percentage Error |
| RMSE | Root Mean Square Error |
| Accuracy | Forecast accuracy as a percentage |
| FVA (difference) | Absolute and percentage difference between the phase and the Baseline |
| Color | Green if the phase improved; red if it worsened |
3. Interpret the color and the FVA sign
| Result | Meaning | Suggested action |
|---|---|---|
| Green / positive FVA | Adjustments improved accuracy vs. the Baseline | Keep the adjustment process in that phase |
| Red / negative FVA | Adjustments worsened accuracy | Review adjustment criteria; consider reducing intervention |
| FVA ≈ 0 | Adjustments had no significant impact | Evaluate whether the adjustment effort justifies the operational cost |
If a phase shows consistently negative FVA across several consecutive cycles, SofIA generates an automatic alert and recommends reducing human intervention in that phase. Review alerts in the SofIA panel.
4. Analyze by SKU or by family
Use the table filters to go deeper:
- By SKU: identify which specific products worsen with adjustments.
- By family: detect whether the issue is systemic across a category.
- By phase: compare which team (Commercial vs. Marketing vs. Consensus) adds the most value.
5. Document and act
If you find phases or SKUs with recurring negative FVA:
- Share the report with the team responsible for that phase.
- Apply the 50/50 Rule in the next cycle to require justification before adjusting.
- Monitor the FVA of the next closed cycle to verify whether it improved.
See also
- What is FVA and the 50/50 Rule? — conceptual foundation
- Reference: Forecast Value Added — formal metric definitions
- Apply the 50/50 Rule — how to require justification before adjusting
- Compare versions — see forecast evolution across versions