Forecasting: Methods, Models & Applications in OM

What is Forecasting?

Forecasting is the process of predicting future events or trends based on historical data, market trends, and expert judgment. In operations management, it helps businesses plan resources, manage inventory, and optimize production.

Why is Forecasting Important?

• Reduces uncertainty in decision-making.
• Helps allocate resources efficiently.
• Improves customer satisfaction by meeting demand.
• Minimizes costs (e.g., overstocking or stockouts).

Types of Forecasts

Forecasts can be classified based on time horizon and data used:

mindmap
  root((Forecasting Types))
    Time Horizon
      Short-term (<3 months)
      Medium-term (3 months - 2 years)
      Long-term (>2 years)
    Data Used
      Qualitative (Expert Judgment)
      Quantitative (Statistical Models)

Qualitative Forecasting Methods

Used when historical data is unavailable or when expert opinion is critical.

1. Jury of Executive Opinion

• A group of experts (managers, sales teams) discuss and agree on a forecast.
• Example: A bank predicting loan demand based on economic trends.

2. Delphi Method

• Experts provide anonymous forecasts, which are refined in multiple rounds.
• Example: Tech companies forecasting AI adoption trends.

3. Market Research

• Surveys, focus groups, or consumer polls to gauge demand.
• Example: Daraz using customer surveys to predict holiday sales.

Quantitative Forecasting Methods

Based on historical data and statistical models.

1. Time-Series Forecasting

Assumes future trends follow past patterns.

Components of Time-Series Data

mindmap
  root((Time-Series Components))
    Trend (Long-term movement)
    Seasonality (Repeating patterns)
    Cyclical (Economic cycles)
    Random (Unpredictable fluctuations)

Common Time-Series Models

Worked Example: Daraz’s Holiday Sales Forecast

• Data: Past 12 months of sales (in ₹ crore): 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160
• Method: 3-month moving average
  • Forecast for Month 4: (50+60+70)/3 = 60
  • Forecast for Month 5: (60+70+80)/3 = 70
  • Forecast for Month 13: (140+150+160)/3 = 150

2. Causal Forecasting

Assumes future values depend on external factors (independent variables).

Common Models

Worked Example: NTC’s Electricity Demand Forecast

• Variables:
  • Y = Electricity demand (in MW)
  • X₁ = Temperature (°C)
  • X₂ = Number of households
• Regression Equation: Y = 50 + 2X₁ + 0.5X₂
  • If X₁ = 30°C and X₂ = 1,000,000 households: Y = 50 + 2(30) + 0.5(1,000,000) = 500,110 MW

Forecasting Accuracy & Error Measurement

No forecast is perfect—we measure errors to improve models.

Common Error Metrics

Example: Nabil Bank’s Loan Forecast Error

• Forecasted Loans: 500 crore
• Actual Loans: 550 crore
• MAD: |500 - 550| = 50 crore (if only one data point)

In the Real World

1. eSewa & Khalti (Digital Payments)

   • Use time-series forecasting to predict transaction volumes during festivals (Dashain, Tihar).
   • Helps allocate servers and prevent crashes.

2. Daraz (E-commerce)

   • Uses demand forecasting to stock inventory before major sales (e.g., Black Friday).
   • Reduces stockouts and improves customer satisfaction.

3. NTC (Electricity Supply)

   • Applies causal forecasting (temperature + household data) to manage power generation.
   • Prevents blackouts during peak demand (e.g., summer heatwaves).

Case Study: Toyota’s Production Forecasting

Toyota uses JIT (Just-in-Time) inventory with short-term forecasting to minimize waste.

• Method: Combines exponential smoothing (for demand trends) and supplier lead-time data.
• Result: Reduces excess inventory by 30% while maintaining production efficiency.

Exam Tip

• Short-answer questions: Know the difference between qualitative and quantitative methods.
• Problem-solving: Practice moving averages, exponential smoothing, and regression with given data.
• Case studies: Be ready to explain how real companies (Daraz, NTC, banks) use forecasting.
• Diagrams: Draw time-series components and regression lines clearly in exams.