Demand Forecasting in Supply Chain
Introduction to Forecasting in Supply Chain
In this section, the speaker introduces the importance of forecasting in supply chain management and its impact on strategic and planning decisions.
Forecasting as a Basis for Decision Making
- Forecasting is a crucial tool that forms the basis for all strategic and planning decisions in a supply chain.
- It is used in both the push and pull phases of the supply chain process.
- Forecasting plays a role in various activities across the value chain, including production, marketing, finance, and human resources.
The Role of Coordination
- Close coordination between different functions such as production, marketing, finance, and HR is essential for achieving desired outcomes.
- Forecasting provides input for decision-making in these areas.
Implied Demand Uncertainty
- Forecasts are never exact and always have some level of error.
- Low implied demand uncertainty results in lower forecasting errors (around 10%).
- High implied demand uncertainty leads to higher forecasting errors (ranging from 40% to 100%).
Expected Value and Error Calculation
- When developing forecasts, two key factors are considered: expected value (forecasted demand) and the level of error associated with that forecasted demand.
- Knowing both values helps make better decisions based on forecast accuracy.
Time Horizon Considerations
This section discusses the importance of time horizons when it comes to forecasting. Different time horizons require different forecasting techniques.
Short-Term vs. Long-Term Forecasts
- Forecasts can be done for different time horizons.
- Short-term forecasts (e.g., immediate requirements or weather predictions) tend to be more accurate than long-term forecasts.
- Long-term forecasts (over 3-4 years) are used for strategic planning decisions like facility development.
- Intermediate time horizons (around 6 months to 1 year) involve strategic decision-making.
- Short time horizons (one day to 3 months) focus on operational decisions.
Accuracy and Precision
- Short-term forecasts benefit from available tools and data analytics techniques, resulting in higher accuracy.
- Long-term forecasts may be less accurate due to changing conditions and limited data availability.
Types of Forecasting Techniques
This section explores different types of forecasting techniques based on the time horizon and the nature of decision-making.
Time Horizon Classification
- Forecasting techniques can be classified based on the time horizon they cover.
- Long-term forecasting (over 3-4 years) involves planning approaches for decisions like new facility development.
- Intermediate time horizon (around 6 months to 1 year) focuses on strategic decision-making.
- Short time horizon (one day to 3 months) is associated with operational decision-making.
Disaggregated Demand
- As the number of SKUs (Stock Keeping Units) and distribution channels increase, demand gets distributed across these channels and SKUs.
- Each variant or SKU may have its own level of forecast inaccuracy.
Importance of Forecasting Time Horizon
This section emphasizes the significance of considering the time horizon when selecting a forecasting technique.
Accuracy Based on Time Horizon
- The accuracy of forecasts varies depending on the chosen time horizon.
- Short-term forecasts tend to be more accurate due to available tools and data analytics techniques.
- Long-term forecasts may have lower accuracy but can be improved with more accurate data over time.
Managing Immediate Issues
- Accurate short-term forecasts help manage immediate issues effectively, as seen in the example of Bandra Worli Sea Link's car traffic forecast.
Aggregate Forecast
- With an increasing number of SKUs and distribution channels, demand is distributed across these different elements.
- Each variant or SKU may have its own level of forecast inaccuracy.
New Section
In this section, the speaker discusses the characteristics and methods of forecasting.
Forecasting Characteristics
- Forecasting helps in predicting future demand and reducing forecasting errors.
- Demand and forecast are closely related, and accurate forecasting is crucial for effective decision-making.
Forecasting Methods
- There are two broad categories of forecasting methods: qualitative and quantitative.
- Qualitative methods rely on subjective opinions and interviews when data is not available. They are used for long-term forecasting.
- Quantitative methods are data-based and include time series analysis, regression analysis (causal analysis), and simulation. Time series analysis is widely used for immediate and intermediate requirements, while regression analysis focuses on identifying independent factors affecting demand.
Time Series Analysis
- Time series analysis is a popular method for forecasting that utilizes historical data to analyze patterns and make predictions. It can be used for both immediate and intermediate requirements.
- Adaptive time series methods are particularly useful as they allow the model to evolve over time based on new data, resulting in more accurate forecasts.
Regression Analysis (Causal Analysis)
- Regression analysis involves identifying independent factors that influence demand and developing mathematical relationships between these factors and demand. Future values of these independent factors can then be substituted into the equation to predict demand accurately.
- The relationship between demand and independent factors can be linear or nonlinear, depending on the complexity of the model being used.
Simulation
- Simulation is an alternative method when a proper mathematical model cannot be developed for demand forecasting purposes. It involves using random future data to make projections about future demand trends.
New Section
In this section, the speaker discusses the use of time series methods, causal methods, and simulation methods for forecasting. The focus is on quantitative forecasting methods.
Quantitative Forecasting Methods
- Time series methods, causal analysis (regression analysis), and simulation are commonly used quantitative forecasting techniques.
- These methods utilize historical data, mathematical relationships between variables, and random future data to make accurate predictions about future demand trends.
New Section
This section discusses the systematic and random components in forecasting models. The random component is unpredictable, while the systematic component can be determined using mathematical formulas.
Determining Systematic and Random Components
- The systematic component can be determined using mathematical formulas.
- The random component is unpredictable and attributed to forecasting errors.
- It is important to use appropriate methods to determine the systematic component accurately.
- Errors can magnify if the wrong method is used to determine the systematic component.
New Section
This section explains the concept of systematic and random components in demand data. Demand fluctuates around a straight line, representing the systematic component, with variations attributed to the random component.
Systematic and Random Components in Demand Data
- Demand fluctuates around a straight line representing the systematic component.
- Fluctuations above or below the straight line are attributed to the random component.
- The amount of randomness in demand cannot be determined precisely.
- Seasonality, trend, and level are three characteristics of the systematic component.
New Section
This section further explores different types of characteristics in demand data's systematic components: level, trend, and seasonality.
Characteristics of Systematic Components
Level Data
- Represents almost constant demand over time with minor fluctuations.
- Desesonalized demand removes seasonal effects for analysis purposes.
Trend Data
- Shows continuous increase (positive trend) or decrease (negative trend) in demand over time.
- Actual data may have fluctuations around the trend line due to random components.
Seasonality
- Some products exhibit seasonal behavior with peak demand during specific periods.
- Demand increases significantly during these periods while remaining low during other times.
New Section
This section discusses the relationship between seasonal and level/trend data. It also explains how product life cycles can be related to level and trend components.
Relationship Between Seasonal, Level, and Trend Data
- Desesonalized demand removes the seasonal effect from the data.
- Level data represents constant demand after removing seasonality.
- Trend data shows continuous increase (positive trend) or decrease (negative trend) in demand over time.
- Product life cycles can be related to level and trend components.
New Section
This section concludes the discussion on systematic components by relating them to different stages of a product's life cycle.
Systematic Components in Product Life Cycle
- Introduction stage: Growth stage with increasing trend in demand.
- Maturity stage: Level data with stable demand.
- Decline stage: Negative trend with decreasing demand over time.
The transcript does not provide further sections or timestamps beyond this point.
New Section Understanding Product Life Cycle and Forecasting Methods
In this section, the speaker discusses the importance of understanding the product life cycle and how it relates to selecting appropriate forecasting methods. They also explain the components of time series data and the concept of adaptive forecasting.
Selecting Forecasting Methods Based on Product Life Cycle
- When a product is in the growth stage, marketers should select methods that deal with positive trends. Negative trend or level data methods should be avoided.
- During the maturity period of demand data, methods suitable for handling level data should be chosen.
Components of Time Series Data
- Time series data consists of systematic components (such as level, trend, and seasonality) and random components (deviations from systematic components).
- Fluctuations in demand are results of random components.
- Forecasting error refers to the difference between forecasted value and actual demand. It is used to update forecasting models in adaptive forecasting systems.
Importance of Forecasting and Method Selection
- Forecasting helps in predicting future demand accurately.
- Various methods exist for forecasting, each with its own relationship to time horizon.
- Understanding different components present in time series data aids in selecting appropriate forecasting methods.
New Section Summary
In this section, the speaker concludes their discussion on time series analysis by summarizing key points related to forecasting methods and their relationship with time horizon.
Recap: Time Series Analysis
- The systematic components include level, trend, and seasonality.
- The example of product life cycle was used to illustrate how these components relate to each other.
- Different forecasting methods are suited for different types of components.
Timestamps were not provided for all sections.