Quantitative Concepts and Methods in Supply Chain - Coursework Example

Quantitative Concepts and Methods in Supply Chain Name: Tutor: Course: Date: Table of contents Introduction 3 Quantitative Concepts and Methods 3 Method Selection 5 Exponential Smoothing Method and Application 5 Exponential Smoothing in Solving Sales Demand and Industrial Production Problems 7 Decision making using Quantitative Techniques (Exponential Smoothing) 8 Limitations of Exponential Smoothing Method 9 Mitigating the negative effects of the exponential smoothing limitations 10 Conclusion 11 References list 12 Quantitative Techniques in Supply Chain and Logistics Management Introduction Supply chain relies on accurate forecasts to ensure constant supply and satisfaction to the consumer market of its products or services. When forecasts become inaccurate it can lead to excesses and shortages across the supply chain. Chances of poor customer service, work disruption and missed deliveries arise due to shortages of services, parts and materials. On the other hand, overly optimistic forecasts which increase costs arise from excesses of capacity or materials. In the supply chain, both excesses and shortages have a negative impact not only on profits but also on customer service. According to Wilson, Holton and Barry (2004: 44), organizations lower the chance of occurrences in various ways as they strive to build up the best feasible forecasts. Forecasting, information sharing and collaborative planning increases supply chain visibility when they engage major supply chain partners. These supply chain partners to have access to inventory and sales information in real-time (Chandra & Grabis, 2005: 343). Owing to rapid communication about unplanned events and poor forecasts such as work stoppages causes changes in plans. When choosing among different techniques, decision makers will want to include accuracy along with cost as a factor. Every business organization for the success of its daily activities requires accurate forecasts which are the basis for schedules in the organization (Zhang, 2004:22). Schedules will be generated only if the forecasts are accurate using optimal resources, output and correct timing of output. These lead to headaches for managers, dissatisfied customers and additional costs if they remain unfulfilled. Use of demand forecasting applications use time series techniques involving parameters such as weekly revenues while associative techniques for a one-time decision involve a single forecast such the size of a production plant (Delurgio, 2008:42). Quantitative Concepts and Methods Quantitative methods involve the development of associative models or projection of historical data that attempts to use explanatory (causal) variables to develop a forecast. For example, the time series consist of observations in time-ordered sequence that are taken at regular intervals such as annually, quarterly, monthly, weekly, daily and hourly. The data on the other hand assume measurements of productivity, precipitation, shipments, consumer price index, demand, accidents, earnings, output and profits (Delurgio, 2008:43). Such forecasting techniques take the assumption that past values of the series can help generate future values. Despite the methods being used widely, there has been failed attempt to identify variables influencing the series that sometimes produce satisfactory results. Any underlying behavior of the series is brought about by analyzing time-series data by visually examining the plot and merely plotting the data. Some parameters that are observable in this circumstance are variations around average, cycles, seasonal variations and trends which provide irregular or random variations (Dejonckheere et al. 2003: 575). Trend is the long-term downward or upward movement in the data such as cultural changes, changing incomes and population shifts. Seasonality is fairly regular and short-term variations associated with factors like time of day or the calendar experienced by theaters, supermarkets and restaurants depicting daily ‘seasonal’ and weekly variations. Zhao and Leung (2002:328) note that cycles are essentially wavelike variations spanning more than one year and includes agricultural, political and economic conditions. Irregular variations arise from unique situations like a major product or service change, strikes and severe weather conditions though they do not reflect typical behavior since when they distort the overall picture when included in the series. These should be removed from the data when identified whenever possible. After all other behaviors have been factored, random or residual variations always remain (Sanders & Graman, 2009:121). In an industrial production case, demand forecast is based not on unit sales but time series of the past demand since one or more stock-outs happened, sales would shallowly reflect demand. Time-series include; naïve methods, simple exponential smoothing and moving averages methods based on historical data patterns. They take measurements at successive points and utilize time as independent variable to generate demand over successive periods (Gardner, 2006:652). These measurements can be at regular or irregular interval such as annually, monthly, daily or hourly based on historical data which easily represent expected future conditions. For demand to show consistent patterns, the time-series models should forecast the past and show future recurrence. Method Selection The preferred method is exponential smoothing which gives more weight to new data while old data receives increasingly less weight. It might lag behind the trend in an event of definitive trend just like the moving averages and is widely used class of procedures. In order to forecast the immediate future, exponential smoothing of discrete time series is paramount (Sanders & Graman, 2009:122). The reason for choosing this method is that it is simple, computationally efficiency and easy to adjust hence responsive to changes in the forecasted processes with reasonable accuracy. In exponential smoothing just like the moving average smoothens the original series and uses the smoothed series to forecast future values of the variable concern. More distant observations of the series are allowed to exert less influence on the forecast of future values than the more recent values (Holt, 2004:6). Drawn from the weighted average of past observations, exponential smoothing is a pragmatic and simple approach to forecasting where the immediately preceding observation has less weight compared to the present observation (Sanders & Graman, 2009:118). This shows that any previous observations show exponential decay owing to the influence of past data. Exponential Smoothing Method and Application Exponential smoothing is relatively easy to understand and use hence a sophisticated weighted averaging method. It considers each new forecast based on the previous forecast and a proportion of the difference between the actual value of the series and the forecast at that point (Tiacci & Saetta, 2009:68). This means; Next forecast = α (Actual - Previous forecast) + Previous forecast Where; Actual - Previous forecast is the forecast error while α is the percentage of the error. In a more concisely way, Ft = Ft-1 + α (At-1 – Ft-1) Where; Ft = Period t forecast Ft-1 = the period t-1 or previous period forecast Α = Percentage or smoothing constant At-1 = Sales for the previous period or actual demand A percentage of the forecast error is shown by α which represents the smoothing constant. The percentage of the previous error plus the previous forecast equals a new forecast. For example, if the previous forecast in a mobile phone sales outlet was 42,000 units per month and the actual demand were 40,000 units given α = .10, the new forecast will be calculated as follows: Ft = 42,000 + .10(40,000-42,000) = 41,800 Now, if the actual demand of mobile phones goes to 43,000 then the next forecast becomes Ft = 41,800 + .10(43,000 - 41,800) = 41,920 Alternatively, the weighting of the latest actual demand and the previous forecast becomes Ft = (1 – α) Ft-1 + αAt-1 For instance, if α = .10, this would turn out to be Ft =.10At-1 + .90Ft-1 The smoothing constant, α determines the quickness of forecast adjustment to error. The smoothing becomes greater when its value gets closer to zero, since the forecast slow in adjusting to forecast errors. In the converse, when the values of α approaches 1.00, there is less the smoothing and greater the responsiveness (Snyder et al. 2002:13). Using forecast errors to guide the managerial decision is through a selection of a smoothing constant which is trial and error or a matter of judgment. The main aim is to select a smoothing constant that balances the benefits of responding to real changes and smoothing random variations if and when they happen. Usually, the values of α range from 0.05 to 0.50 where low values of α is used if the averages are to be stable. When the underlying average is vulnerable to changes then higher values are used (Gardner, 2006:650). In forecasting, exponential smoothing becomes one of the most commonly used techniques, owing to ease of calculation and alteration of the weighting scheme simply by changing α values. In order for the forecasts to adjust to the data, exponential smoothing begins several periods back instead of starting a single period back (Sanders & Graman, 2009:119). To obtain a starting forecast, a number of different approaches are taken such as the subjective estimate, the naive approach (forecast for second period based on actual values) and the average of the first several periods. In practice, moving averages is obtained by getting the forecast of say the fourth period by taking the average of the first three values. Exponential smoothing provides a better starting forecast as it tends to be more representative given as α approaches zero, smoothing increases (Gardner & McKenzie, 2005:1231). Exponential Smoothing in Solving Sales Demand and Industrial Production Problems To understand exponential smoothing and its implication on managerial decision making, a case study is considered. The case of Nokia mobile phones sold at an outlet in Sweden is forecasted to respond to market demand. Demand forecasting was applied especially using simple exponential smoothing from the year 2001 to 2012. The results of the forecast based on alpha value 0.1 and 0.4 are represented graphically. The table below shows the forecast in the 12 year period. Year Actual demand α = .10 forecast α = .40 forecast 2001 42,000 - - 2002 40,000 42,000 42,000 2003 43,000 41,800 41,200 2004 40,000 41,920 41,920 2005 41,000 41,730 41,150 2006 39,000 41,660 41,090 2007 46,000 41,390 40,250 2008 44,000 42,070 42,550 2009 45,000 42,350 43,130 2010 38,000 42,350 43,880 2011 40,000 41,920 41,530 2012 41,370 40,920 Figure 1: Exponential smoothing for demand (‘000’) of Nokia mobile phones in a Swedish outlet From the graph above, it shows that α are critical in smoothing the random values of actual demand to provide a more ‘smooth’ pattern of demand. This helps in relating dynamic decision technologies to a broad range of managerial decisions involving time and money. Under uncertainty, strategic decisions are made based on forecasts given that during forecasting, choices are a result of anticipation of results of inactions or actions (Gardner, 2000:490). Delays and indecision engender failure hence use of quantitative methods in sales and production help administrators and industrial managers to anticipate and manage uncertainty (Gardner & McKenzie, 2005:1233). Decision making using Quantitative Techniques (Exponential Smoothing) Exponential smoothing has been used to provide ‘smooth’ patterns of demand based on the use of α-values to affect the forecasted outcome or result (Hyndman & Koehler, 2006:684). According to the graph provided earlier in this report, it can be learned that lower α-values of 0.6 (1-0.4) takes more information from historical values and produces a forecast result that is similar in shape to the actual values. On the contrary, higher α-value of 0.9 (1-0.1) due to its reliance on current values produced a much smoother demand forecast result. In Nokia sales, decision-makers in-charge of key functional areas use such information to make critical decisions related to material purchasing, distribution and logistics, management inventory holding and production capacity of mobile phones. They predict their demand forecast for the next period based on the two sets of results give in a range as well as an idea on company performance in coming year with respect to sales volume. Managers can easily make strategic and tactical decisions related to managing their logistics and supply chain network (Sanders & Graman, 2009:120). For example, the knowledge of how much demand of mobile phones in following year (2012) was taken from the previous production capacity hence allowing the buyers to have an idea on how much accessories and parts are needed in the next few months. Supply chain partners upstream can receive and use this information to undertake their own production planning and forecasting. When downstream demand is fulfilled excess inventory and dead stock will be avoided if this is well managed. Limitations of Exponential Smoothing Method It is evident that when smaller value of α values are used there will be slower responses while larger α values cause quick reaction of the smoothed value to random fluctuations and real changes. This limits exponential smoothing model to non-seasonal patterns for short-term forecasting owing to approximately zero-trend. If one is extending past the next period, any forecast past the next period is a forecast of values for that period used as a surrogate for the actual demand (Holt, 2004:8). Consequently, any error grows exponentially since there is inability to add actual demand or corrective information. Regardless of how accurate and suitable, all forecasting methods have limitations with exponential smoothing model taking the assumption that the forecasted variable behave similarly both in the past and the future. It assumes that demand is dependent on only one variable while in reality, due to the existence of other factors influencing demand, it is believed to utilize multivariate forecasting methods (Gardner, 2000:494). Taking the example of Nokia, the demand of mobile phone on the whole is highly influenced by other factors such as technological advancement, sales and advertising and government policies. Using more than one variable, demand forecasting is required at this stage. Nevertheless, it is a process to collect data belonging to several variables for analysis and take up a substantial amount of time. In order to ensure that the forecast result applies to the exponential smoothing formula, the variable is checked for suitability and reliability. This means more effort and time will be needed as companies would not want to make costly investments for short-term forecasts (Dejonckheere et al. 2004:745). Though the demonstration was done, unfortunately it was not able to identify the seasonal component of Nokia Mobile phone sales despite time-series forecasting methods being excellent tools to determine seasonal and growth effects in a firm’s demand. Mitigating the negative effects of the exponential smoothing limitations Mitigating the negative effects of exponential smoothing model is through removal of other factors likely to affect demand. Demand pattern does not change over time since the exponential smoothing model is based on the assumption that removal of irregular and seasonal components can aid in more accurate forecasting (Delurgio, 2008:44). In the removal of both seasonal and irregular components, the usage of seasonal indices to de-seasonalizing the historical information is important. Improving forecast accuracy has undertaken a broad research spectrum based on exponential smoothing model with multiple variables. Tactical managers need this information qualitative forecasting methods besides the forecast result data set to obtain of individual’s expertise on performances and market trends. It also arrives at more accurate forecast results and corrects the forecast figures (Dejonckheere et al. 2003: 571). Alternatively, Nokia firm as part of risk control can set up specialised groups to observe and provide feedback on potential factors likely to pre-emptively affect the market demand. The company can adjust their forecast figures accordingly and anticipate the possible changes in demand instead of reacting to demand based on actual changes. Conclusion The exponential smoothing model as demonstrated in this report has shown that time-series forecasting set of methods provides a simplistic and robust means based on historical information way to prepare future forecasts (Gardner & McKenzie, 2005:1232). Demand as it is acknowledged is not only dependent on time only but also on other factors such as seasonality and irregularity of the components that allows the final forecast figures to be reduced. Qualitative methods that depend on the value judgement of experienced individuals while quantitative methods like de-seasonalising forecast result helps to mitigate the negative effects of the exponential smoothing model’s limitations (Chandra & Grabis, 2005: 340). In order to rapidly respond to any significant changes, further studies are encouraged if Nokia Company is aware of factors affecting its demand it can easily communicate forecasted information upstream hence reducing the chain effect of these changes. References list Chandra C & Grabis J 2005, Application of multiple-steps forecasting for restraining the bullwhip effect and improving inventory performance under autoregressive demand, European Journal of Operational Research, 166, 337-350. Dejonckheere J Disney S M Lambrecht M R & Towill D R 2003, Measuring and avoiding the bullwhip effect: A control theoretic approach, European Journal of Operational Research, 147, 567-590. Dejonckheere J Disney S M Lambrecht M R & Towill D R 2004, The impact of information enrichment on the Bullwhip effect in supply chains: A control engineering perspective, European Journal of Operational Research, 153, 727-750. Delurgio S 2008, Forecasting Principles and Applications. New York: Irwin/ McGraw-Hill. Gardner E S 2000, Evaluating forecast performance in an inventory control system, Management Science, 36, 490-499. Gardner E S 2006, Exponential smoothing: The state of the art-Part II, International Journal of Forecasting, 22, 637-666 Gardner ES & McKenzie E 2005, Forecasting trends in time series, Management Science, 31, 1237-1246. Holt CC 2004, Forecasting seasonals and trends by exponentially weighted moving averages, International Journal of Forecasting, 20, 5-10. Hyndman RJ & Koehler AB 2006, Another look at measures of forecast accuracy, International Journal of Forecasting, 22, 679-688. Sanders NR & Graman GA 2009, Quantifying costs of forecast errors: A case study of the warehouse environment, Omega: The International Journal of Management Science, 37, 116-125. Snyder RD Koehler AB & Ord JK 2002, Forecasting for inventory control with exponential smoothing, International Journal of Forecasting,18, 5-18. Tiacci L & Saetta S 2009, An approach to evaluate the impact of interaction between demand forecasting method and stock control policy on the inventory system performances, International Journal of Production Economics, 118, 63-71. Wilson J Holton R & Barry K 2004, Business Forecasting. New York: McGraw-Hill. Zhao X Xie J & Leung J 2002, The impact of forecasting model selection on the value of information sharing in a supply chain, European Journal of Operational Research, 142, 321-344. Zhang X 2004, The impact of forecasting methods on the bullwhip effect, International Journal of Production Economics, 88, 15-27. Read More