Demand Forecasting & Demand Planning Master Notes (Ch. 1–5)

CHAPTER 1 — Forecasting: What & Why

• Purpose and Motivation for Forecasting

  • Forecasts serve as the primary driver for production scheduling, sales targets, marketing initiatives, and financial planning.
  • They provide a lens through which an organization can identify market opportunities, recognize potential threats, track the growth or decline of specific products, and observe broader shifts in the market.
  • The Value of Timeliness over Perfection: A critical principle is that ‐Forecasts are more useful if we get them early enough to react‐. Timing is prioritized because even an imperfect forecast allows for proactive adjustments, whereas a perfect forecast delivered too late is useless.

• Taxonomy of Forecast Types

  • Micro Forecasting: This occurs at the company or organizational level. It involves predicting sales, shipments, and cash flow.
  • Macro Forecasting: This occurs at the national or international level. Key variables include Gross Domestic Product ($GDP$), inflation rates, and unemployment rates.
  • Operational Forecasting: Defined as forecasts with a horizon of $\leq 6\,\text{months}$. These are used by functional teams for production scheduling, procurement of raw materials, and logistics management.
  • Strategic Forecasting: Defined as forecasts with a horizon of $> 6\,\text{months}$. These guide long-term capacity planning (e.g., building new plants), marketing strategy development, and Mergers & Acquisitions (M&A).

• The Nature and Failures of Forecasting

  • Forecasting is described as a blend of ‐art + science‐.
  • Common Sources of Error:
    • Data Integrity Issues: Utilizing wrong data types, such as using shipment data as a proxy for actual demand.
    • Data Quality Issues: The presence of missing values, statistical outliers, or structural breaks in time-series data.
    • Faulty Assumptions: If the foundational assumptions about market conditions or consumer behavior are incorrect, the forecast will inevitably be incorrect.
    • Model Selection: Choosing an inappropriate statistical model for the specific data pattern.

CHAPTER 2 — Evolution in Forecasting

• Historical and Terminological Shifts

  • The field has transitioned from ancient methods involving ‐prophets & astrologers‐ to modern, rigorous statistical forecasters.
  • The Nomenclature Evolution: The discipline has been renamed multiple times as its scope broadened: Forecasting $\rightarrow$ Business Forecasting $\rightarrow$ Sales Forecasting $\rightarrow$ Demand Forecasting $\rightarrow$ Supply Chain Forecasting.

• Changing Perceptions and Market Dynamics

  • Formerly dismissed as ‐voodoo science,‐ forecasting is now recognized as a critical tool for reducing inventory carrying costs, shrinkage, obsolescence, and for significantly improving customer service levels.
  • Consumer and Market Trends:
    • Consumers have become less loyal and increasingly demanding.
    • Product variety has exploded, while product life cycles have shortened drastically.
    • Globalization has lengthened supply chain lead times.
    • Retailers now hold less inventory, leading to a shift toward more frequent, smaller orders.

• Shifting Forecasting Strategies

  • Demand Push to Demand Pull: Moving from pushing products into the market to pulling products based on actual demand signals.
  • SKU Rationalization: Reducing the number of distinct items ($SKUs$) to improve overall accuracy.
  • Frequency: Modern forecasts are updated monthly or even weekly.
  • Lean Forecasting: A specialized focus on high-value items or items that are notoriously difficult to forecast.

• Process and Data Evolution

  • Stages of Process Maturity:
    1. Silo Forecasting: Every department (Sales, Finance, Production) creates their own forecast independently, leading to organizational chaos.
    2. Consensus Forecasting: All departments agree upon and work from a single shared forecast.
    3. Sales & Operations Planning (S&OP): Executive leadership joins the decision-making process in a structured monthly cycle.
    4. Collaborative Planning, Forecasting, and Replenishment ($CPFR^\circledR$): High-level collaboration with customers using real Point of Sale ($POS$) data.
  • Data Hierarchy: Shipment Data $\rightarrow$ Demand/Order Data $\rightarrow$ Point of Sale ($POS$) Data. $POS$ data is considered the gold standard as it is more stable and reflects genuine consumer behavior.

• Model and Scope Evolution

  • Moving averages and exponential smoothing have been adapted for modern business contexts.
  • Advanced Models: Incorporation of Box-Jenkins (ARIMA), Neural Networks, Regression analysis, and Croston’s method for intermittent demand.
  • Strategic Shift: There is a declining relevance for macro forecasts in daily business operations; micro-level operational forecasts now drive the core functional planning of the firm.

CHAPTER 3 — Fundamentals of Demand Forecasting & Supply Planning

• Core Principles

  • Forecasts are inherently ‐always wrong‐; the objective is not to achieve perfect accuracy but to minimize the margin of error.
  • Error Tolerance: The acceptable level of error is dictated by the lead time and the financial cost associated with a forecasting error.
  • Demand vs. Shipment: Demand forecasts are categorically better and more accurate than shipment forecasts.

• Consumer vs. Customer Demand

  • Consumer Demand: Defined as the end-user's purchase. It is generally more stable and is the preferred metric for forecasting.
  • Customer Demand: Defined as orders from retailers or distributors. This fluctuates significantly due to internal inventory policies, trade promotions, and financial incentives.

• Conceptual Distinctions

  • Forecast: An expectation of what will happen.
  • Goal: An aspiration of what the company wants to achieve.
  • Budget: A fixed financial target.
  • Plan: The set of assumptions and actions used to construct the forecast.

• Forecastability and Factors affecting Accuracy

  • Hard-to-Forecast Categories: New products, fashion items, high-technology goods, highly promoted items, and products with intermittent demand.
  • Functional Biases:
    • Sales: Tends to under-forecast to ensure they exceed their quotas.
    • Finance: Tends to be overly conservative.
    • Production: Bias varies based on Key Performance Indicators ($KPIs$), such as whether they are judged on inventory levels versus service levels.
  • Forecast Horizon: This is strictly driven by total lead time, which equals $\text{Production Lead Time} + \text{Supplier Lead Time}$.

• Supply Planning Fundamentals

  • Forecasts are the foundation for production, procurement, and logistics operations.
  • Focusing on consumer-based forecasts helps mitigate the ‐bullwhip effect‐ (the amplification of demand volatility up the supply chain).
  • Longer forecasting horizons lead to greater uncertainty, which necessitates higher levels of safety stock.

CHAPTER 4 — Demand Planning

• Purpose and Components

  • Demand planning aims to manage and grow demand by improving forecast accuracy, reducing operational costs, and minimizing business risk.
  • The 8 Components of Demand Management:
    1. Volume Management.
    2. Supply‐Demand Alignment.
    3. Product Portfolio Management.
    4. Product Mix Management.
    5. New Product Introductions ($NPI$).
    6. Management of Hard-to-forecast products.
    7. Communication.
    8. Performance Monitoring.

• Management Strategies

  • Volume Management: Aligning short-term operational forecasts with long-term strategic goals; tracking progress via monthly targets.
  • Supply-Demand Alignment Scenarios:
    • If Demand > Supply: Increase production shifts, outsource manufacturing, allocate limited stock to top customers, or reduce promotional activity.
    • If Supply > Demand: Increase promotions, adjust pricing downward, or slow down production rates.
  • Portfolio and SKU Rationalization: Pruning the product line to focus on profitable items reduces costs and improves the accuracy of the remaining forecasts.

• New and Difficult Products

  • New Products: Require two distinct forecasts: the total lifecycle sales volume and the specific shape of the demand curve. Methods include using analogs (similar products), pre-order data, and early $POS$ signals.
  • Difficult Products: Effort should be made to reduce volatility caused by promotions and focus on high-value/high-variability items. Utilizing ‐aggregate forecasting‐ at the category level can improve results.

• Communication and Performance

  • Communication: Requires continuous dialogue between Sales, Marketing, and Production.
  • Example (Volvo): A miscommunication regarding a promotion for ‐green cars‐ serves as a cautionary tale of functional silos.
  • Key Performance Indicators ($KPIs$):
    • $MAPE$ (Mean Absolute Percentage Error).
    • $WMAPE$ (Weighted Mean Absolute Percentage Error).
    • Rate of forecast changes.
    • Customer service levels.
    • Inventory turnover/levels.
    • Performance metrics for new products.

• Business Policies for Demand Planning

  • Organizations must establish formal policies for: Profit objectives, limited supply allocation, handling small orders/Less-Than-Truckload ($LTL$) shipments, customer service tiers, SKU pruning criteria, cannibalization rules, managing volatile products, exception management, disaster response, and price discrimination strategies.

CHAPTER 5 — POS-Based Demand Planning

• Data Types

  • Point of Sale ($POS$): Direct data showing sales to the end consumer.
  • Syndicated Data: Third-party aggregated data that combines $POS$ with competitor market data.

• The Superiority of POS Data

  • It is the most stable and least distorted data source.
  • It reflects the ‐true‐ consumer demand and effectively reduces the bullwhip effect.
  • It enables high-frequency (weekly) forecasting and supports assortment planning, promotion analysis, and measuring the effectiveness of in-store displays.

• Technical Details

  • Transmission Methods:
    • Private Exchanges: Such as Wal-Mart's Retail Link or Partners Online.
    • Electronic Data Interchange ($EDI 852$): A standard business protocol.
    • Internet: Considered the least reliable method for transmission.
  • Market Intelligence Extraction: $POS$ data reveals SKU/Regional/Store seasonality, promotion effectiveness, display impact, inventory balance, and can compare consumer appetite against store expansion.

• Operational Challenges and Takeaways

  • Challenges: Lack of a standard format across retailers, missing data segments, converting multi-packs into single units, disparate calendar systems, self-scanner promotion noise, and technical transmission failures.
  • Key Takeaways:
    • $POS$ forecasts act as leading indicators.
    • It takes approximately $3-4\,\text{weeks}$ of $POS$ data to determine the success of a new product introduction.
    • Processing must be automated due to the sheer volume of data.
    • Verification Rule: Shipments are usually $5-10\,\%$ higher than $POS$ values.
    • $POS$ data is the foundation for advanced collaboration models like $CPFR^\circledR$ and Vendor Managed Inventory ($VMI$).