# Demand Planning

Demand planning is the process of estimating future customer demand and translating those estimates into actionable plans that drive supply, inventory, and financial decisions across the business. 🎯

It sits at the heart of every IBP process — get demand wrong, and everything downstream suffers.


# Demand Planning vs. Demand Forecasting

These terms get used interchangeably, but they shouldn't be.

Demand Forecasting Demand Planning
Scope Generating a numerical estimate of future demand End-to-end process: forecast, consensus, shaping, communication
Inputs Historical data, statistical models Forecasts + market intelligence + business strategy
Output A baseline forecast An agreed demand plan aligned with business objectives
Who owns it Analytics / data science Demand planning team with cross-functional input

Forecasting is a component of demand planning — an important one, but not the whole story. A demand plan also incorporates demand shaping decisions, new product launches, and consensus from sales and marketing.


# The Demand Planning Process

At a high level, the demand planning cycle follows four steps:

  1. Data cleansing — Remove outliers, correct errors, handle promotions and one-off events so history reflects true underlying demand.
  2. Statistical baseline — Apply forecasting models to generate an unbiased starting point (see methods below).
  3. Enrichment — Layer in market intelligence, sales input, promotional plans, and known events.
  4. Consensus — Cross-functional review to agree on one demand number the organization will execute against.

This cycle typically repeats monthly in an IBP cadence, though short-term adjustments may happen weekly via demand sensing.


# Statistical Forecasting Methods

Averages the last n periods of demand. Simple and transparent, but lags behind trends and ignores seasonality. Best suited for stable, mature products with low variability.

\displaystyle F_t = \frac{1}{n} \sum_{i=1}^{n} D_{t-i}

Applies exponentially decreasing weights to older observations. More responsive than moving average, with a smoothing parameter alpha controlling how fast old data is discounted. Variants (Holt, Holt-Winters) handle trend and seasonality.

\displaystyle F_t = \alpha \cdot D_{t-1} + (1 - \alpha) \cdot F_{t-1}

Models demand as a function of one or more independent variables (price, GDP, weather, marketing spend). Useful when causal drivers are well understood, but requires clean driver data and careful variable selection.

\displaystyle D = \beta_0 + \beta_1 X_1 + \beta_2 X_2 + \cdots + \epsilon

# Judgmental Forecasting

Statistical models can't capture everything. Judgmental forecasting fills the gap:

  • Sales input — Account-level intelligence on pipeline, customer plans, and competitive dynamics.
  • Market intelligence — Category trends, regulatory changes, macroeconomic shifts.
  • Event-driven adjustments — Promotions, product launches, supply disruptions, weather events.

The key is to apply judgment on top of a statistical baseline, not instead of one. Unstructured overrides tend to introduce bias rather than accuracy — this is why tracking Forecast Value Added matters.


# Demand Segmentation: ABC-XYZ Analysis

Not all items deserve the same forecasting approach. ABC-XYZ analysis classifies products on two dimensions:

  • ABC (volume / revenue contribution): A = top 80%, B = next 15%, C = remaining 5%
  • XYZ (demand variability): X = low variability, Y = moderate, Z = highly erratic
X (Stable) Y (Variable) Z (Erratic)
A (High volume) Statistical forecast; high automation Statistical + judgmental overlay Judgmental; event-driven
B (Medium volume) Statistical forecast; periodic review Blend of statistical and judgmental Judgmental; safety stock focus
C (Low volume) Simple model or rule-based Rule-based; minimal effort Make-to-order or carry safety stock

This matrix helps demand planners allocate their time where it matters most — spend energy on high-value, hard-to-forecast items (AY, AZ) rather than spreading attention evenly.


# Planning Horizons 📆

Horizon Timeframe Primary Use Typical Granularity
Short-range 0–3 months Execution, replenishment, scheduling SKU-location, daily/weekly
Medium-range 3–18 months S&OP/IBP, capacity planning, budgeting Product family, monthly
Long-range 18–36+ months Strategic planning, network design, capital investment Category / business unit, quarterly

Short-range forecasts benefit most from demand sensing. Medium-range is the sweet spot for the IBP process. Long-range planning is less about precision and more about directional alignment with strategy.


# Further Reading