02 / Adaptive Demand Forecasting · Study 07 of 07

Demand Score & Decision Support

Translating predicted demand into an interpretable 0–100 decision layer

Decision-support layer

My Role: I developed and evaluated this component within the demand-forecasting project, including benchmark comparison and interpretation.
Business / Analytical Question: How can a weekly point forecast become an intuitive score without introducing subjective weights?
Evidence: Map predicted customers to a frozen development-period percentile scale. 2025 holdout observed demand rose across score quintiles; the score adds no separate forecast lift.
Decision / Finding: Use for pacing and performance review alongside predicted customer counts.
Outcome / Status: Decision-support layer: translates the provisionally validated integrated forecast into a shared business language without changing its underlying predictions. Analytical validation; no operational deployment is claimed.

Business Question

How can a weekly point forecast become an intuitive score without introducing subjective weights?

Why It Matters

A common scale helps marketing and operations compare expected demand across weeks. Internal planning should retain the underlying customer forecast so the score does not obscure business meaning; this public case study withholds customer-volume values.

Data / Method

Locate predicted New Customers within the empirical distribution of usable 2023–2024 development predictions, then scale the percentile to 0–100. Freeze the reference distribution before the holdout.

Analysis

Predicted New Customers
Percentile relative to development-period predictions
0–100 Demand Score

0–20

Unusually weak expected demand

20–40

Below-normal expected demand

40–60

Typical expected demand

60–80

Above-normal expected demand

80–100

Unusually strong expected demand

Customer-volume thresholds are withheld; the 0–100 percentile interpretation and frozen reference-distribution method are preserved.

Results

In the 2025 holdout, observed demand increased monotonically across score quintiles: higher score bands corresponded to higher observed demand. Use the score to compare expected demand alongside the underlying customer forecast.

Method detail

Supporting diagnostic: 2025 holdout Spearman rank correlation ≈ 0.905. Quintile-level observations are not supplied, so no numerical quintile chart is reconstructed.

Interpretation

This is not a manually weighted index. It is a relative demand-intensity layer, not a probability, causal measure, or guarantee of realized acquisition. Use it for pacing and performance review alongside the customer forecast.

Limitations & Considerations

The score inherits the forecast’s limitations and its development-period scale. Any recalibration must be versioned and validated; the source band boundaries are approximate.

How this fits into the forecasting system

Decision-support layer: translates the provisionally validated integrated forecast into a shared business language without changing its underlying predictions.