Skip to main content

Apply the 50/50 Rule

Before making a manual adjustment to the forecast, apply the 50/50 Rule: empirical evidence shows that of every 100 manual adjustments, approximately 50 worsen the forecast. For this reason, SofIA requires you to justify the adjustment with concrete qualitative information before modifying the Baseline.

For the statistical foundation of this rule, see What is FVA and the 50/50 Rule?.

When this guide applies​

Every time you attempt to edit a forecast cell or initiate a manual adjustment, SofIA may present the 50/50 Rule questionnaire. Answer it honestly before confirming the change.

Steps​

1. Review SofIA's 5 questions​

Before adjusting, answer these questions. If none applies to your situation, do not adjust:

#QuestionExample of a valid situation
1Are there new markets or sales channels the model does not know about?You opened distribution in a new region this month
2Have you lost customers or markets the model has not yet reflected?A major customer canceled their contract
3Are there new products with no sales history?Launch of a new SKU (phase-in)
4Are there discontinued products whose demand is declining?SKU in the process of being retired (phase-out)
5Is there a product transition (phase-in / phase-out)?One SKU replaces another; demand migrates gradually

2. Evaluate your answer​

  • If you answered "Yes" to at least one question and can describe the concrete event: the adjustment is justified. Proceed to edit the value and add a note explaining the reason (if a comment field is available).

  • If you answered "No" to all questions: the AI Baseline is likely better than your intuition in this case. Do not adjust. Leave the forecast unchanged.

Doubt is a signal not to adjust

If you cannot articulate in one sentence why the model is wrong, that is a clear signal the adjustment is not justified. Trust the Baseline.

3. Document the reason for the adjustment​

When an adjustment is justified, record the reason. This allows you to:

  • Verify later (with FVA) whether the adjustment actually improved accuracy.
  • Build institutional knowledge about which events justify intervention.

4. Verify the impact in the next cycle​

When the version closes, review the FVA for the phase where you adjusted. Check whether the adjustment improved (positive FVA) or worsened (negative FVA) accuracy relative to the Baseline.

See Audit adjustments with FVA for the full review process.

Decision summary​

Can I answer "Yes" to at least one question?
  └── Yes → adjust, document the reason, verify with FVA
  └── No  → do not adjust, keep the Baseline

See also​