4 – Demand Planning and Forecasting

Introduction

Demand planning and forecasting is a business process that involves projecting future demand for products and services and allocating production and distribution resources accordingly. It entails a variety of various business operations and necessitates the sharing of timely data, correct data processing, and agreement on collaborative business plans along the supply chain. Forecast and demand planning is the process of identifying and forecasting recipient needs to ensure that the end client receives the appropriate quantities of commodities consistently, at the right time and location, and at the lowest possible cost. Demand forecasting and planning is both an art and a science. It necessitates sound judgment, business knowledge, and technological abilities. When done correctly, it can provide a true competitive advantage and increase sales while managing inventory and providing best-in-class customer service. Forecasting is dependent on an organised approach as well as modelling. Both are equally significant. We will cover the principles of forecasting and planning, why forecasting is vital, and its function in demand planning.

4.1 Demand Forecasting

Our future vision assists us in selecting what product to offer, what procedure to apply, and what values to supply to customers. We need to be able to see around the corner to prevent things from getting out of hand. To accomplish this, we will need a variety of tools. Forecasting tools aid in analyzing the environment and provide input on how the organisation can best leverage its resources. This unit will look at some of these forecasting methods.

Analyzing the factors that affect future values determines the method for estimating future values. One approach to distinguishing different types of forecasting is to consider how far into the future they look. Individual item detailed predictions are typically used for short-term forecasting. Such forecasts are used to plan short-run decisions such as inventory control, order sizing, and transportation scheduling, among other things.

Demand levels are forecasted as part of medium-term projections. This is critical to the company as a whole since it offers fundamental inputs for the planning and controlling all functional areas, including the supply chain. Demand predictions are required throughout the planning and control process. Demand planning attempts to address the issues posed by these considerations. The following are some examples of broad fundamental questions:

How do you know which new products or services to launch or discontinue, which markets to enter or quit, and which products to promote?

What sales plans do you have, given that sales targets are typically based on future revenue forecasts?

How will demand fluctuation be absorbed over the next 6 to 18 months? How will production, procurement, and logistics strategies be made?

What should our financial goals be? How can demand changes be absorbed through inventories, manpower, work hours, supplier activity, and so on? What impact do they have on profit expectations?

Will the organisation lose orders if all demands are not met? What policy should the company implement?

Each option determines the organization’s tactical manoeuvres (medium-term policy). Once decided, the policy directs the organization’s actions. A good policy must be founded on an understanding of what customers value. For example, if a policy of not meeting all needs leads to customers switching to a competitor’s goods, the company may find it difficult to wean them back when demand drops.

Demand levels and their timing significantly impact capacity levels, financial needs, and the organisation’s overall structure. Each functional area has its own set of forecasting issues.

Supply chain forecasting is concerned with demand’s spatial and temporal fluctuation, the magnitude of its variability, and the degree of randomness. Accurate predictions of the product and service volumes to be handled by the supply chain are required for planning and controlling supply chain activities. These estimates are often forecasts and predictions. The supply chain professional is frequently required to produce estimates for short-term planning, such as inventory control, order sizing, or transportation scheduling. Demand planning becomes required for longer-term decisions.

The time dimension is reflected in supply and demand. It is critical to remember that management activities can influence both supply and demand. Forecasting has several connotations in business and economics. Forecasting involves two separate quantities: a forecast and a prediction. A prediction is a more general concept. It is a prediction of a future event based on subjective factors other than historical facts; this subjective factor does not have to occur in any particular order.

We use a pretty specific definition of a forecast in supply chain management, which is presented below:

A forecast is an estimate of a future occurrence obtained by systematically combining and casting forward facts from the past in a predetermined manner.

The supply chain spans both space and time. The supply chain professional must understand where and when the demand volume will occur. The demand must be spatially located to establish warehouse sites, balance inventory levels across the supply chain network, and distribute transportation resources regionally.

Demand might vary substantially depending on the firm’s activities and the activity for which the prediction is necessary. Demand is classified into two sorts.

Independent demand is forecasted using statistical techniques. These theories are based on demand independence and randomization. In the case of dependent demand, however, the market is known.

4.1.1 Forecasting Methods

Various forecasting approaches can be applied to generate the forecast. The forecasted item type and availability of historical data will determine the best strategy. These factors frequently influence the method you use to generate the forecast.

Forecasting can be done using one of four methods, which are listed below:

Qualitative: Qualitative forecasts are used when there is little or no historical performance data to determine demand. They are usually based on an expert’s knowledge of the product, the industry, and customer preferences. When new items are brought to the market, an expert’s viewpoint is frequently practical.

Time Series: Time series forecasts use historical demand to estimate future demand. Numerous computational methods are available. This strategy is typically best suited for items with a well-defined historical pattern that does not fluctuate dramatically from one year to the next, such as “staple stock” items in a retail store.

Causal forecasting: Causal forecasting is employed when there is a clear association between one or more variables and product demand. For example, disposable income, lifestyle indicators, etc, may be used to determine the demand for several durable consumer items. The strategy, however, necessitates a high level of modelling skill.

Simulation: This highly sophisticated method is typically utilised when an organisation wants to produce many “what-if” scenarios. For example, such a model could predict the impact on product demand if prices were raised or disposable income was reduced. Organizations must assess these sensitivities in many circumstances to produce a more accurate projection.

The method utilised should effectively meet the forecasting model’s aims. More than one approach may be employed to provide the types of outputs requested. For example, the method used for short-term forecasting may differ from that used for long-term forecasting.

4.1.2  Accuracy and Validation Assessments

Validation and verification are required for all models. The validation question is, “Are we developing the proper system?” In contrast, verification aims to answer the question, “Are we designing the system correctly?”

Because validation is used to establish the credibility of a model, the method used for validation must also be trustworthy. Time series characteristics that can be disclosed by observing its graph, including anticipated values and residual behaviour, condition forecasting modelling

Holding out a specified number of data points for estimation validation (i.e., estimation period) and a specific number for forecasting accuracy is an effective technique for modelling forecasting validation (i.e., validation period). The non-held-out data are used to estimate the model’s parameters; the model is then tested using data in the validation period, and if the findings are satisfactory, forecasts are created beyond the end of the estimation and validation periods.

A successful model should have modest error measures in both the estimate and validation periods, and its validation period statistics should be similar to those in the estimation period.

Holding data for validation is perhaps a model’s most significant diagnostic test; it provides the best indicator of the accuracy that can be expected when projecting the future. It is a general rule that at least 20% of data should be retained for validation purposes.

4.2 Collaborative Forecasting

Companies will gain from such forecasting models as technology develops faster and more innovatively and supply chains’ willingness to exchange information grows. Inventory will gradually be replaced with information. This optimism can be shown in Collaborative Planning, Forecasting and Replenishment (CPFR), widely recognised as an extension of supply chain management and a component of supply chain philosophy.

Wal-Mart and Warner-Lambert conducted the first CPFR audit for Listerine products. To trade forecasts, they employed specialised CPFR software. Iteratively, supporting data such as previous sales trends, promotion strategies, and weather were transferred. This enabled them to create a single forecast based on their original projections. The outcomes were encouraging. Listerine sales climbed, fill rates improved, and inventory investment was significantly reduced.

CPFR forecasting is based on supply chain management. It is a business concept that holistically approaches supply chain management and information exchange among trading partners. Utilising standard metrics, consistent terminology, and firm commitments improves supply chain efficiencies for all partners.

In other words, collaborative forecasting is predicated on treating the entire supply chain or partnership as a single entity and sharing information among the network’s members. The goal is for supply chain members to work together to meet the final consumer’s needs. This is accomplished by providing the right product to the customer at the right place, time, and price.

The “CPFR Overview Committee” defined the target objectives and business advantages of employing CPFR, according to the Round Table conducted at the University of Denver in May 2002. These are as follows:

• Increased in-stock at the shelf by 5-8%; decreased average network inventory by 10%; and increased revenues by 8-10%.
• 1-2 percent reduction in operational expenses
• 3-4 percent reduction in the cost of goods
• Reduced lead time/cycle time by 25-30%
• Account receivables were reduced by 8-10%.
• Forecast error was reduced by +/-20% (six weeks out) and +/-30% (six months out) (twelve weeks out)

Forecasting and demand planning are critical for successfully implementing a supply chain management strategy. Reduced inventory investments and improved customer service levels are strongly related to the accuracy and efficiency with which demand is forecasted and communicated up and down the supply chain.

Though precise and effective forecasting remains elusive, many businesses employ a collaborative method. Collaborative forecasting involves the whole supply chain that participates in demand decisions. This requirement entails obtaining internal and external forecasting information, which is utilised to drive supply chain actions.

Collaborative forecasting overcomes some of the fundamental issues with traditional forecasting. It is a method for combining enterprise-wide knowledge into a forecast that is more accurate than a conventional forecast and has the backing of the entire supply chain. The goal is to provide the most accurate and timely demand predictions.

Rising competition and the requirement that producers in a supply chain synchronize operations to benefit from collaboration are the driving forces behind the demand for collaborative forecasting. Inventory understocking and overstocking both cause problems and reduce a manufacturer’s competitiveness. Collaborative forecasting can aid in eliminating surplus inventory while supporting the participating enterprises’ supply chain management initiatives.

4.3 Collaborative Planning, Forecasting, and Replenishment (CPFR)

Collaborative Planning, Forecasting, and Replenishment is a nine-step approach to enhancing supply chain management that integrates demand and supply planning. The CPFR process is divided into three key sub-processes: planning, forecasting, and replenishment.

It is customarily started by identifying a ‘forecasting champion.’ The forecasting champion might be an individual, a department, or a company. Selecting a “forecasting champion” is essential to the collaborative forecasting technique. The champion’s role is to effectively communicate with and lead the organisations involved in the sharing and agreement on information sharing, forecasting methods, and technologies. There are several methods for collaborative forecasting. These forecasting processes are typically custom-built and developed to match the unique requirements of different businesses. To complete the assignment successfully, the champion must comprehend and underline the crucial aspect of the process. The champion must also enable the cross-functional initiatives needed for better forecasting.

The forecast collaboration group will be formed next. Each organisation should select a representative for this group. The group’s makeup, however, should be such that its members represent a wide range of functional areas, such as sales, marketing, logistics/operations, finance, and information systems. Members from external partners, such as suppliers and customers, are included in this definition. The work must be directed toward two goals:

(a) including the most recent and best available information in the final projection and

(b) addressing the changing demands and circumstances confronting the organisation.

The group determines the collaborative forecast process’s goals, objectives, and immediate needs based on all forecast users’ informational needs. The group will identify the factors, methods, and technologies that influence the forecast and the relevant sources of information. Internal or external sources could be used. The group’s ability to ensure that all users have access to information at all necessary levels determines the outcome.

Companies frequently hold at least two meetings per month. The first meeting gathers information and prepares the base forecast. The second conference brings together alternative forecasts and works through issues to reach a consensus.

Once the relevant information has been decided upon and is available, the next step is at the firm level. The supply chain participants determine how the various bits of information will be combined. After necessary approvals, the consensus forecast is used for the company’s sales and planning systems.

For instance, Gillette discovered that senior managers connect, distributors link with Gillette, and so on.

The aligned teams are advantageous because they support the company’s mission of strengthening key customer relationships through an effective, collaborative, and improvement-oriented process. Gillette discovered various opportunities to address issues such as shrinkage, shelf replenishment, packaging, and display design by doing it this way. It also makes sense because, by worrying about their customers’ performance issues, they reduce the retailer’s loss of sales while increasing their sales.

However, metrics and incentives must be included for this type of synergy to occur. These ensure that forecast accuracy and related supply chain performance increase due to the collaborative approach. Measurements should illustrate the success of joint efforts not only at a specific point in time but at the pace of improvement over time.

Measurements can vary, but they should contain a comparison of the reality to the predicted. These allow organisations to compare forecasts for consistency and comparability. The absolute error for each item is a standard method (the actual minus the estimates, divided by the actual without the sign).

Another critical metric is a bias indicator. The percentage of things that were either over- or under-anticipated is shown here. The bias indicator identifies trends and tendencies that cause certain items to be over- or under-forecasted. Compensating for this bias can be critical to improving the affected forecasts.

The collaborative approach is a departure from tradition, requiring participants to change how they previously worked. Changes in working procedures can lead to resistance concerns. If participants do not change their behaviour, the work put into developing a better forecasting process will not yield the most significant outcomes. The collaborative process necessitates extra work for many of the participants, in addition to adjusting previous work patterns. Participants who have never participated in the forecasting process may see it as extra effort.

However, results do not appear right away. It takes time to implement the system and see its outcomes. Before results can be obtained, participants must go through a learning curve, systems and sub-systems must be built, and process decisions must be made. All of these difficulties combine to make the transition to collaborative forecasting a challenge for all businesses in the supply chain.

In collaborative forecasting, the information is more current and accurate because companies supplement statistics with information gathered directly from customers, markets, and other sources. This supplemental information reduces the forecast’s uncertainty and, as a result, the inventory carried because the need for inventory to cover uncertainty is reduced.

The fundamental tenet of CPFR is that all supply chain players create a coordinated forecast. A corporation can work with many other supply network members, both upstream and downstream. Every stakeholder in a CPFR process—supplier, manufacturer, distributor, and retailer—can examine and modify forecast data to optimise the entire process. CPFR effectively eliminates predicting and guessing. This means that producers and retailers discuss their ideas, with each other’s assumptions and limits, in a well-understood manner.

However, using such systems necessitates a significant investment as well as a high level of skill. Gillette discovered that not everyone in the supply chain could join the integrated supply chain. Finally, it decided to differentiate its customer strategy based on customer size. Based on Gillette’s value chain structure, more complicated, sophisticated retail chains received more unique and integrated services. Smaller, independent businesses are provided with a standardised set of supply chain services. The cost to serve and the level of customer sophistication are the driving forces behind this distinction. As a result, Gillette only performs CPFR on its largest accounts.

Numerous practical examples of CPFR exist. Heineken USA uses CPFR and has successfully reduced the order cycle time. Its programme is being expanded and gives its top 100 distributors collaborative planning and replenishment tools.

4.4 Quantitative Techniques

Most businesses, especially tiny and medium-sized businesses, do not consider implementing such IT-based models cost-effective, even if they have the expertise and competence to do so. They can create their projections using classic quantitative approaches, the most widely employed of which is the ‘time series’ method.

4.4.1 Time Series

A time series is a characterization of change that occurs throughout time. It is a quantitative model that, using historical data, shows the shift in demand for products and services and the pattern in the sequence of occurrence. A ‘times series’ refers to specific patterns or characteristics of the change process.

One key theme in the ongoing development of inventory theory is the creation of realistic inventory models that represent product demand. In reality, demand is erratic and difficult to foresee. Furthermore, when product life cycles shorten, the randomness and unpredictability of these demand processes increase.

Inventory managers rely on estimates based on a historical demand time series, such as a weighted moving average. These estimates are typically based on the premise that the most recent demand data are the best indicators of future demand. Time series are often used for inventory decisions to generate and maintain predictions at multiple product levels, provide appropriate forecasts for product and location planning and replenishment, and optimise demand history through demand cleansing and seasonal profiling.

The strategy will be determined by the forecast’s accuracy and the pattern of prior demand. Below are some of the approaches utilised to overcome such challenges.

4.4.2 Moving Average Method

The ‘moving average approach’ is the most basic type of time series analysis. In this methodology, raw data is turned into a moving average that depicts the trend in demand change. A moving average is an arithmetic average of data across time. By averaging previous data, an attempt is made to reduce random variations.

The data is updated regularly by replacing the item in the average with the new item. This model type is very beneficial when no discernible pattern or seasonal influence on demand exists. It is commonly used to analyse this type of data, which is superior to raw data since it eliminates or smoothens out irregularities in the time series.

The general formula for the moving average is:

F_{t + 1} = (A_t + A_{t – 1} + A_{t – 2} + A_{t – 3} + ……+ A_{t – n + 1})/n

Where: Ft + 1 is the moving average for the period t + 1,

A_t, A_{t – 1}, A_{t – 2}, A_{t – 3} etc. are actual values for the corresponding period, and ‘n’ is the total number of periods in the average.

For example, suppose the prices for a product are given for 12 months and a five month average is to be computed. Each month is sequentially designated as A₁, A₂, A₃, A₄, A₅ etc.

Then, the first 5-month moving average would be; F₅ = [(A₁ + A₂ + A₃ + A₄ + A₅)/5]

The second moving average of the next five months would be;

F₆ = [(A₂ + A₃ + A₄ + A₅ + A₆)/5]

And so on.

The last item would be the average

F₁₂ = [(A₈ + A₉ + A₁₀ + A₁₁ + A₁₂/5]

The series’ stability frequently determines the number of periods to include in the moving average. When the series is very stable, large values of ‘n’ should be employed, integrating more previous data in the average results in a forecast less sensitive to random changes. On the other hand, small values of ‘n’ are advised where individual values are prone to vary.

Simple Moving Averages (MA) are an effective and efficient method for forecasting the future if the time series is stationary in terms of mean and variance. This is critical and must be determined. Furthermore, if there is a pattern in the data, the moving average has the disadvantage of lagging behind the trend.

WMAs (Weighted Moving Averages)

Each period in a simple moving average has the same weight. However, it is frequently advantageous to accentuate specific components more than others. For example, you may select that recent demand should be prioritised over earlier demand. In this scenario, weights can be assigned to each element as desired, requiring that the total of the weights equals ‘1’.

The weighted moving average’s general formula therefore becomes:

F_{t + 1} = [(w_tA_t + w_{t – 1}A_{t – 1} + w_{t – 2}A_{t – 2} + w_{t – 3}A_{t – 3} + ……+ w_{t – n + 1}A_{t – n + 1})/n]

Where: F_{t + 1} is the weighted moving average for the period t+1,

And, wt is the weighing factor, and wt = 1

For example, if ‘n’ is 5, we could weight the moving average as follows:

w1 = 5/(1 + 2 + 3 + 4 + 5) = 5/15 = 1/3;

w2 = 4/15;

w3 = 3/15 = 1/5;

w4 = 2/15 and w5 = 1/15.

∑w = 1/3 + 4/15 + 1/5 + 2/15 + 1/15 = 1

In this example, the most recent era is given the most weight compared to the earlier periods. As the period lengthens, the weight gradually decreases.

This model allows one to compensate for seasonality or other unexpected occurrence by carefully fitting the coefficients, wA_t. However, it should be noted that management must choose the coefficient, which is important to the model’s applicability.

Exponential-weighted moving average

An exponential-weighted moving average prediction is based on the idea that the most recent demand observations can best anticipate future demand. As a result, exponential smoothing models are widely used in Supply Chain Management. When the forecasting horizon is relatively short and little knowledge about the cause-and-effect relationship between an item’s demand and the independent factors that influence it, a smoothed time series is created.

In contrast to regression models, explained in the following section, exponential smoothing does not employ information from series other than forecasted ones. These models are also easily accessible in regular computer software and require little data storage and computational power.

The Exponential Smoothing approach is simple to compensate for past errors, simple to prepare follow-up forecasts from, and ideal for scenarios requiring several forecasts.

We must define an initial value because exponential smoothing is an iterative procedure.

Smoothing on a single axis: The values for a smoothed series are calculated using the Single Exponential Smoothing method. You select a damping coefficient, referred to as the weighting factor. This factor is employed to smooth the data. It ranges from ‘1’ to ‘0’ and defines the smoothing effect’s sensitivity. The previously demonstrated exponential relationship can now be represented as follows in standard notation:

F_{t + 1} = αDt + (1 – α) F_t

Where:  Dt is the actual value

Ft is the forecasted value

α is the weighting factor, which ranges from 0 to 1 t is the current time period.

Since      F_{t + 1} =    Dt + (1 – α) F_t

F_t = αD_{t – 1} + α(1 – α) F_{t – 1} and so on

Therefore F_{t + 1} = D_t + α(1 – α) + α( D_{t – 1} + (1 – α) F_{t – 1})…….

F_{t + 1} =   αD_t + α(1 – α)F_{t – 1} + α(1 – α)² F_{t – 2} + (1 – α)³ F_{t – 3}…….

Thus, the forecast for the following period is the algebraic sum of the forecast for the previous period plus ‘?’ times the forecast error from the previous period.

As the observation ages, the weights are assigned in an exponentially decreasing order. This means that recent observations are weighted more heavily in forecasting than older observations.

A small “smoothens the values by giving lower weightage to recent changes,” whereas a big “responds quickly to recent changes in the time series but provides less smoothing.”

When the data is exponentially smoothed, the smoothed value is the forecast for period ‘t + 1’. In addition, unlike moving averages, which require the first value for the fifth week, the analysis requires only three pieces of data. It’s worth noting how the moving average, weighted moving average, and basic exponential smoothing smooth out the seasonality in this specific series. As the number of readings increases, the difference between the various weighting components becomes more apparent.

The selection of the smoothing constant is the most crucial decision the manager must make. How should it be interpreted? The constant must be equal to or between ‘0’ and ‘1’ in value. As it grows, so does the variance of the mistake. To minimise the inaccuracy, we want to make the forecast as small as feasible (0), but this renders the forecast insensitive to changes in the underlying time series. We want the prediction to be as large as possible (1) to make it sensitive to changes, which raises the error variance.

There are no set rules for determining the value of α. If more weight must be given to recent data, the value should be closer to ‘1’, with values between 0.1 and 0.3 being the most typically used.

However, one technique for selecting the best fit is to choose the value of α so that the error variance is as tiny as possible. This is demonstrated in the following example:

Problems:

Saluja Brothers manufactures simple lathes for the international market. The factory manager employs the exponential smoothing technique to arrive at his forecasts, using a smoothing constant of 0.2.

The sales manager also forecasted using the exponential smoothing method but with a smoothing constant of 0.5.

Compare the forecasts for the series data under two scenarios, and decide which forecast you will accept and why.

Period	1	2	3	4	5	6	7	8	9	10
Observations	30	32	35	34	31	30	33	36	36	34

Solution:

The basic exponential smoothing model is F_t = αD_t + (1 – α) αF_t–1

Where: D_t is the actual value

F_t is the forecasted value

α is the smoothing constant or weighting factor, and

F_t–1 is the current period.

Assume that the smoothed value of the time series for the first period is equal to the actual first value of the time series. You can calculate the values as shown in the table below. The calculations are simple.

The table shows the forecasts under the two specified conditions, i.e., α = 0.2 and α = 0.5:

	ά = 0.2					ά = 0.5
Period	Dt	Dt– Ft-1	ά* Dt– Ft-1	Ft	(Dt– Ft-1)²	Dt– Ft-1	ά* Dt– Ft-1	Ft	(Dt– Ft-1)²
1	30	0.00	0.00	30.00	0.00	0.00	0.00	30.00	0.00
2	32	2.00	0.40	30.40	4.00	2.00	1.00	31.00	4.00
3	35	4.60	0.92	31.30	21.16	3.00	1.50	33.00	9.00
4	34	2.70	0.54	31.85	7.29	1.00	0.50	33.5	1.00
5	31	-0.85	-0.17	31.68	0.73	-2.50	-1.25	32.25	6.25
6	30	-1.68	-0.34	31.35	2.83	-2.25	1.13	31.13	5.06
7	33	1.65	0.33	31.68	2.73	1.88	0.94	32.06	3.54
8	36	4.32	0.86	32.54	18.67	2.94	1.97	34.03	8.65
9	36	3.46	0.69	33.23	11.97	1.97	0.99	35.00	3.88
10	34	0.77	0.15	33.38	0.60	-1.00	-0.50	34.50	1.00
Total					69.98				42.38

 

∑(D_t – F_t–1)² i.e. Total Variance when α = 0.2 is 69.98

Therefore,

Error Variance of the series = ∑(D_t – F_t–1)²/(n – 1) = 69.98/9 = 7.75

Similarly, ∑(Dt – Ft–1)² i.e. Total Variance with α = 0.5 is 42.38

Therefore,

Error Variance of the series = ∑(D_t – F_t–1)²/(n – 1) = 42.38/9 = 4.70

Where ‘n’ is the number of observations.

One measure of the accuracy of the forecast is the error variance, which is the mean squared error between the forecast and the actual data in the next period [∑(D_t – F_t–1)²/(n – 1)], which has been calculated above. You have to pick the α that gives you the most smallest mean squared error or error variance.

Since the error variance for the case of α = 0.2 is greater than for α = 0.5, the forecast with α = 0.5 is the correct choice as it is more accurate.

Simple Moving Average and Exponentially Weighted Moving Average: An exponentially weighted moving average with a smoothing constant ‘α’, roughly corresponds to a simple moving average period of length ‘n’, where ‘ ‘ and ‘n’ are related by the following equation:

α = 2/(n + 1)  OR  n = (2 – α)/α.

As a result, an exponentially weighted moving average with a smoothing constant of 0.1 corresponds to a 19-day moving average. A 40-day simple moving average would roughly correspond to an exponentially weighted moving average with a smoothing constant of 0.04878. These values are derived from the equations mentioned above.

This demonstrates that ‘simple moving average’ is a subset of exponential smoothing. The average age of forecasts obtained by exponential smoothing is the same as that of a moving average of order ‘n’ with an integer portion of (2 – α)/α.

Double Exponential Smoothing: Double exponential smoothing is applied to a smoothed time series. Data with trends can be forecasted using double exponential smoothing. While the single exponential approach solves problems with stationary trends, the double exponential method solves problems with nonstationary trends.

One can create a linear trend in the anticipated value by exponentially smoothing a smoothed series once more. After the data period, the extrapolated series grows constantly, equal to the growth of the smoothed series.

Triple Exponential Smoothing: When the trends are non-linear, double-exponential smoothing or even triple-exponential smoothing may be required. Triple Exponential Smoothing is more effective at dealing with parabolic trends and is commonly employed for such data.

While basic exponential smoothing requires constant conditions in the demand parameters, double-exponential smoothing can capture trends when demand changes linearly. Triple-exponential smoothing can handle almost all other business time series.

The benefits of exponential smoothing include not imposing any deterministic model other than what is inherent in the time series to suit it. It can be customised to capture seasonal patterns in a time series. Moving averages assign identical weights to previous observations, whereas exponential smoothing assigns decreasing weights as the observation ages.

REVIEW QUESTIONS:

1. Explain the concept of Demand Forecasting.
2. Outline various methods used for forecasting.
3. Explore the importance of Accuracy and Validation Assessments in predicting.
4. Analyze Collaborative Planning Forecasting and Replenishment (CPFR) in detail.
5. Evaluate whether CPFR aligns with the principles of supply chain management, providing reasons for your stance.
6. Elaborate on how Collaborative Forecasting addresses shortcomings of traditional forecasting methods.
7. Provide a brief overview of the CPFR Model.
8. Enumerate the activities involved in the CPFR Process.
9. Define the collaborative approach in the context of forecasting.
10. Discuss how the time dimension influences the concepts of supply and demand.