CHAPTER 6: DEMAND FORECASTING – PREDICTING THE FUTURE

  

CHAPTER 6: DEMAND FORECASTING – PREDICTING THE FUTURE

Objective

This chapter introduces practical and analytical tools for predicting future demand. It focuses on how businesses, institutions, and governments can use forecasting to make decisions about resources, inventory, and services. The chapter simplifies complex models and offers real-world projects and case studies, such as hostel mess planning and metro ticket estimation.

 

Understanding Demand Forecasting

Forecasting demand involves estimating the future demand for a product or service. It helps avoid waste, reduce costs, and serve customers better. Forecasting combines historical data, patterns, and economic behavior.

 

Popular Forecasting Methods

1. Naïve Forecasting

This is the simplest method. It assumes that what happened last time will happen again next time. For example, if 400 students ate lunch yesterday, you forecast that 400 will eat today too.

Formula in words:
Forecast for the next period = Actual demand in the current period

 

2. Moving Average Method

This method takes the average of demand over a fixed number of past periods. It smooths out short-term ups and downs.

Formula in words:
Forecast = Sum of demands over the last few periods ÷ Number of periods

Example:
If the last 3 days had 390, 420, and 410 meals, then the forecast is (390 + 420 + 410) ÷ 3 = 406.7 meals.

 

3. Weighted Moving Average

Here, more recent days are given higher importance (weight) than older ones.

Formula in words:
Forecast = (Weight1 × Demand1) + (Weight2 × Demand2) + ... and so on.
Make sure all weights add up to 1.

Example:
If today’s weight is 0.6, yesterday’s is 0.3, and the day before is 0.1, then
Forecast = (0.6 × 420) + (0.3 × 390) + (0.1 × 370)

 

4. Exponential Smoothing

This method also gives more importance to recent data, but in a more continuous way. It uses a smoothing constant (between 0 and 1) to update the forecast based on actual demand.

Formula in words:
New forecast = (Smoothing constant × Actual demand) + (1 - Smoothing constant) × Old forecast

If the smoothing constant is 0.3 and the actual demand was 430 and the previous forecast was 410, then
New forecast = (0.3 × 430) + (0.7 × 410)

 

5. Regression Models

Regression models use relationships between variables. For example, metro ticket sales may depend on fare, income, or population.

Simple Regression Formula in words:
Demand = Constant + (Change in demand per unit price × Price)

Multiple Regression Formula in words:
Demand = Constant + (Effect of price × Price) + (Effect of income × Income) + (Effect of season × Season variable)

These models are useful for public transport, retail stores, or utilities where demand depends on many factors.

 

6. Seasonal Adjustment

Sometimes demand varies due to days of the week, seasons, or events. For this, we calculate a seasonal index.

Seasonal Index Formula in words:
Seasonal index = Average demand in a specific period ÷ Overall average demand

Example:
If average demand on Sunday is 550 and overall average is 500, then
Seasonal Index = 550 ÷ 500 = 1.10

Adjusted forecast in words:
Adjusted forecast = Unadjusted forecast × Seasonal Index

This helps in adjusting demand based on regular patterns like weekends, festivals, or climate.

 

Forecasting Tools: Excel and Python

Using Excel

You can use Excel for forecasting with built-in functions:

·         AVERAGE() for moving average

·         Manually apply weights for weighted average

·         Regression via "Data Analysis" → "Regression"

·         Seasonal adjustment via formulas

Using Python (for advanced students)

Here's a simplified view:

·         Use the panda's library to read data.

·         Use Stats models for time-series forecasting.

·         Use Holt-Winters model for smoothing and seasonality.

Example structure:

·         Load your demand data from a CSV file.

·         Apply exponential smoothing.

·         Forecast demand for the next 7 days.

·         Plot actual vs. forecasted demand.

(Note: Actual code files can be shared via the instructor.)

Case Study 1: Forecasting Hostel Mess Demand

Context:
A hostel has 500 students. The mess manager struggles with food wastage and shortages.

Action Taken:
Used a 7-day moving average and exponential smoothing (constant = 0.3).

Result:

·         Wastage dropped from 15% to 5%

·         Cooks planned better based on forecasted demand

·         Event days were adjusted using seasonal indices

Learning Points:

·         Moving averages work for stable daily patterns

·         Smoothing helps in unexpected drops or spikes

·         A forecast doesn’t replace decisions—it supports them

Questions for Students:

1.      What could be added to improve the forecast model?

2.      How can we track absenteeism or guest students? 

k

Here's the graph showing Actual vs. Forecasted Hostel Food Demand over 14 days. It visually represents how well the forecasting model aligns with real consumption, helping identify over- or under-estimation on specific days.

 

Case Study 2: Estimating Metro Ticket Demand

Context:
Indore Metro wants to predict ticket demand to optimize ticket counters and train frequency.

Method:
Used regression with variables: fare price, income level, and day of the week. Also used a seasonal index for weekends.

Result:

·         Peak hours were predicted more accurately

·         Metro management used the data to reduce wait times

·         Fare discounts were timed for low-demand hours

Learning Points:

·         Regression captures external factors

·         Combining seasonal patterns with regression improves accuracy

·         Public services benefit from demand forecasting for better citizen experience

Questions for Students:

1.      What other factors might influence metro ridership?

2.      Can mobile app usage data be added?

 

Practical Project for Students

Task: Forecast demand for canteen food or metro tickets using real or assumed data.

Steps:

1.      Collect past 30-day data (real or hypothetical)

2.      Apply any two methods (e.g., moving average and exponential smoothing)

3.      Use Excel or any software you prefer

4.      Adjust for events or holidays

5.      Present:

o    Daily forecast vs. actual (in graph or table)

o    Interpretation of any error or variation

o    Suggestions for improvement

Learning Outcome:
Students will learn how to use forecasting for planning, reduce resource misuse, and understand customer behavior patterns.

 

Conclusion

Forecasting is not just a number-crunching exercise—it is a smart decision-making tool. When used well, it helps avoid problems, improve efficiency, and even create competitive advantage. Whether it's a mess manager, metro planner, retailer, or policymaker, forecasting enables all to look ahead with data-driven confidence.