Lesson 2 Forecasting

Summary

Highlights

• Forecasting combines art and science to predict future demands. 🔮
• Tupperware uses a range of forecasting methods to maintain its global edge. 🌍
• Time series and associative models are key to accurate business forecasts. ⏳
• Product lifecycle stages heavily influence forecasting strategies. 📊
• Quantitative techniques like regression provide deeper forecasting insights. 📈

Key Takeaways

• Forecasting is the magical crystal ball of business, guiding decisions in HR, supply chains, and beyond! 🔮
• Tupperware's global success is partly due to their savvy forecasting techniques, proving numbers matter! 💹
• From exponential smoothing to regression analysis, mastering these methods can transform business forecasting. 🌟
• An accurate forecast can make or break a business—it’s vital for long-term success! 🚀
• Got forecasting woes? Remember, no method fits all; it's about finding what clicks for your biz! 😉

Overview

Chapters

• 00:00 - 03:00: Introduction and Lesson Overview The chapter titled 'Introduction and Lesson Overview' begins with a cordial welcome to the production and operations management class, specifically lesson two which focuses on forecasting. It outlines the topics covered, including a global company profile of Topperwork Corporation, and delves into the concept of forecasting. This includes discussions on the forecasting time horizon, implementation within the product life cycle, and various types of forecasts. The chapter also highlights the strategic importance of forecasting in areas like human resources, capacity planning, and supply chain management.
• 03:00 - 06:30: Forecasting Definition and Importance This chapter introduces the concept of forecasting and its significance in management. It outlines the seven steps involved in a forecasting system and contrasts different approaches, including qualitative and quantitative methods. Key techniques discussed are time series forecasting, decomposition of time series, naive approach, moving averages, exponential smoothing, exponential smoothing with trend adjustment, and trend projection. It also covers how to account for seasonal variations in data.
• 06:30 - 07:30: Tupperware Case Study This chapter delves into various statistical methods used in forecasting and analyzing data trends. Key topics covered include associative forecasting methods, regression and correlation analysis, and multiple regression analysis. The use of regression analysis to forecast, measure standard error of the estimate, and calculate correlation coefficients for regression lines is thoroughly explored. Additionally, the chapter discusses techniques for monitoring and controlling forecasts, specifically focusing on adaptive smoothing and focus methods. The central theme revolves around leveraging these statistical tools to gain insights and make informed predictions based on data variation.
• 07:30 - 09:30: Forecasting Time Horizons This chapter, titled 'Forecasting Time Horizons,' focuses on the concept of forecasting within the service sector. By the end of the chapter, learners should be able to identify and define different types of forecasts, understand various forecasting time horizons, and apply different approaches to forecasting. These approaches include moving averages, exponential smoothing, trend projections, as well as regression and correlation analysis.
• 09:30 - 12:00: Types of Forecasts The chapter titled 'Types of Forecasts' begins by referencing the various measures of forecast accuracy. It delves into the common perception of 'Tupperware' as primarily an organizer of home parties. Contrary to this belief, Tupperware is highlighted as a successful global manufacturer, generating more than 85% of its $1.1 billion in sales outside the U.S. The brand is a household name in nearly 100 countries, boasting 13 plants worldwide, including one in South Carolina and three in Latin America. This example illustrates the global scale and operation diversity of certain companies, providing insight into the broader context of international business and manufacturing.
• 12:00 - 31:00: Forecasting Methods and Techniques Tupperware is a global brand with a strong presence across multiple continents, recognized for its quality products. They offer a lifetime warranty on their 400 plastic products, ensuring durability against chipping, cracking, or breaking.
• 31:00 - 37:00: Measuring Forecast Accuracy This chapter explores the various statistical models for forecasting used at Top Errors, including moving averages, exponential smoothing, and regression. The focus is on measuring the accuracy of forecasts, which are aggregated by region and globally at Perverse World Headquarters in Orlando, Florida.
• 37:00 - 53:00: Trend Projection and Regression Analysis The chapter titled 'Trend Projection and Regression Analysis' discusses a logistical process involving the delivery and movement of plastic pellets. The process begins with a rail car delivering 2,000 pounds of pea-sized plastic pellets. These pellets are then transported through a vacuum line, transferring them from the trailer into a rail car. Subsequently, the pellets are moved to a silo. The chapter appears to highlight the sequential steps involved in this logistical operation, possibly tying it to the concepts of trend projection and regression analysis, although the specific analytical methods are not detailed in the excerpt provided.
• 53:00 - 74:00: Seasonal Variations and Associative Forecasting The chapter titled 'Seasonal Variations and Associative Forecasting' describes a process involving the handling and preparation of pellets. Initially, pellets from a silo are transferred to an implant hopper, where a mixing process takes place with color concentrate pellets. If the hopper includes color, the pellets are subsequently heated, melted, and then injected into molds to form specific products. After the injection process, items are extracted from the molds, inspected for quality assurance, and then undergo finishing or printing in the packaging area. The chapter ends at this point, indicating further processes might follow.
• 74:00 - 83:00: Monitoring and Controlling Forecasts The chapter discusses key factors involved in the sales forecasting process. It identifies three primary elements crucial for generating precise forecasts: the number of registered consultants or sales representatives, the percentage of dealers currently active, and sales per active dealer on a weekly basis. The forecasting method integrates historical data to enhance accuracy.
• 83:00 - 90:00: Forecasting in the Service Sector The chapter describes how Tupperware stays competitive against strong rivals like Rubbermaid. The company's success is attributed to its group process of refining statistical forecasts. This process involves inputs from various departments such as sales, marketing, finance, and production. Ultimately, the refined forecasts are based on a consensus among all participating managers, a process akin to the 'jury of executive opinion,' which will be elaborated on in the lesson.
• 90:00 - 98:00: Summary and Conclusion The chapter titled 'Summary and Conclusion' discusses the concept of forecasting, highlighting it as both an art and a science used for predicting future events. It explains that forecasting can involve using historical data projected into the future through mathematical models, rely on subjective judgments, or be a combination of both. The chapter introduces various forecasting techniques, emphasizing that no single method is always superior and that often a manager's judgment is crucial in refining forecasting models.

Lesson 2 Forecasting Transcription

• 00:00 - 00:30 good day welcome to our production and operations management class lesson two forecasting outline global company profile of topperwork corporation what is forecasting that includes the forecasting time horizon the implements of product life cycle and the types of forecasts the strategic importance of forecasting human resources capacity supply chain
• 00:30 - 01:00 management seven steps in forecasting system forecasting approaches that includes the qualitative and quantitative methods time series forecasting the composition of time series naive approach moving averages exponential smoothing exponential smoothing with trend adjustment trend projection seasonal variations in data
• 01:00 - 01:30 and statistical variation in data associative forecasting method regression and correlation analysis using regression analysis to forecast standard error of the estimate correlation coefficients for regression lines and multiple regression analysis monitoring controlling forecasts using the adaptive smoothing and focus
• 01:30 - 02:00 forecasting forecasting in the service sector learning objectives when you complete this chapter you should be able to identify or define forecasting types of forecast time horizon approaches to forecast when you complete this chapter you should be able to describe or explain moving averages exponential smoothing trend projections regression and correlation analysis
• 02:00 - 02:30 measures of forecast accuracy when most people think of paperwork the ambition plastic food storage container sold through home parties however topperware happens to be successful global manufacturer with more than 85 percent of its 1.1 billion dollars in sales outside the u.s a household name in nearly 100 countries the firm has 13 plans located around the world one in south carolina three in latin
• 02:30 - 03:00 america one in africa and four in europe and four in asia throughout the world topperware stands for quality providing a lifetime warranty that each of its 400 plastic products will not cheap track break or build forecasting demand a tupperware is critical never ending process each of its 50 profit centers around the world is responsible for computerized monthly quarterly and 12-month sales
• 03:00 - 03:30 projections these are aggregated by region and then globally at the perverse world headquarters in orlando florida this forecasts dry production of each plant the variety of statistical precasting models used at top errors includes every technique discussed in this chapter or in this lesson including moving averages exponential smoothing and regression
• 03:30 - 04:00 top reverse process so first rail car delivers for 2 000 pounds of pea-sized plastic pellets then clear pellets flow through vacuum line so first from the trailer um the the vacuum will get the pellets inside the rail car going to the silo and they will peel up the silo and the silo will be going the
• 04:00 - 04:30 pellets from the silo will go to the implant hopper here and there will be like a color mixing and uh color concentrate pellets if the topper will will be having a color then pellets are heated melted and forced into mold to shape products this is the injection process then items are removed from mold and ex and it's inspected for qa purposes and then finishing or printing is done in the packaging area after that it will be
• 04:30 - 05:00 transported to the customer three key factors in top perverse sales forecast includes the number of registered consultants or sales representatives second the percentage are currently active dealers this number changes each week and month and third sales per active dealer on a weekly basis forecasts incorporate historical data
• 05:00 - 05:30 recent events and promotional events topperware maintains its edge over strong competitors like rubbermaid by using a group process to refine its statistical forecast although inputs come from sales marketing finance and production final workers are the consensus of all participating managers this final step is established version of jury of executive opinion that will be described in this lesson
• 05:30 - 06:00 forecasting is the art and science of predicting future events it may involve taking historical data and projecting them into the future with some sort of mathematical model it may be a subjective or intuitive prediction or it may involve a combination of this that is a mathematical model adjusted by a manager's good judgment as i introduce different forecasting techniques in this lesson you will see that there is a seldom one superior
• 06:00 - 06:30 method what works best in one firm under one set of conditions may be complete disaster in another organization or even different department of the same firm in addition you will see that there are there are limits to what can be performed what or what expected forecasts you can get they are sell them if ever they might not be like the optimal one they're also costly and time consuming to repair and monitor
• 06:30 - 07:00 few businesses however can afford to avoid the process of forecasting by just waiting to see what happens and then taking their chances through effective planning in both of their short-term and long-term forecasting that they must under company products so next forecasting time horizon a forecast is usually classified by future time horizon that is that it covers time horizons fall into three categories first
• 07:00 - 07:30 short range forecast this forecast has a time span of up to one year but is generally less than three months it is used for planning purchasing job scheduling workforce level job assignments and production levels second medium range forecast a medium range or intermediate forecast generally spans from 3 months to 3 years it is useful in sales planning production planning and budgeting cash budgeting
• 07:30 - 08:00 and analyzing various operating plans and lastly long-range forecasts generally three years or more in time span long-range forecasts are used in planning for new products capital expenditures facility location expansion and research and development medium and long range forecasts are distinguished from short-range forecast
• 08:00 - 08:30 by three features first intermediate and long-run forecasts deal with more comprehensive issues and support management decisions regarding planning and products plans and processes implementing some facility decisions such as a company's decision to open a new manufacturing plant can take 5 to 8 years from inception to completion second short-term forecasting usually
• 08:30 - 09:00 employs different methodologies than longer term forecasting mathematical techniques such as moving averages exponential smoothing and trend extrapolation are common to short run projection broader less quantitative methods are useful in predicting such issues as whether a new product like optical this recorder should be introduced into a company's product line finally as you would expect short ranges
• 09:00 - 09:30 uh forecasts tend to be more accurate than longer range forecasts of course factors that influence demand change every day thus the time horizon lanterns is likely to for that forecast will be diminished it almost goes without saying that that sales forecast must be updated regularly to maintain their value and integrity after each sales period forecasts should be reviewed and revised
• 09:30 - 10:00 the influence of products lifecycle another factor to consider when developing sales forecasts especially longer ones is product life cycles product and even services do not sell at a constant level throughout their lives most successful products pass through four stages introduction growth maturity and and decline like the product and first two stages of the life cycle such as the virtual reality and lcd tvs needs longer
• 10:00 - 10:30 forecasts than those such as like the bloopers and skateboards in the maturity and decline stages forecasts that reflect life cycles are useful in projecting different stopping levels inventory levels and factory capacity as the product process from the first stage to the last stage like here so om strategy or issues for
• 10:30 - 11:00 introduction product design and development are critical frequent product and process designs changes short production runs high production costs limited models attention to quality for the growth forecast will be critical product and process reliability competitive product improvements and options increase capacity shift towards product focus enhance distribution
• 11:00 - 11:30 for the maturity it will be the standardization less rapid product changes more minor changes optimum capacity increasing stability of process long production runs product improvement and cost cutting for decline little product differentiation differentiation cost minimization over capacity in the industry prune line to eliminate items not returning good margin and reduce
• 11:30 - 12:00 capacity so these are some of the om strategies or issues that might be experienced during the product life cycles types of forecasts organizations use three major types of forecasts in planning future operations first economic forecasts it addresses the business cycle by predicting inflation rates money supplies housing starts and other planning indicators
• 12:00 - 12:30 second technological forecasts are concerned with rates of technical technological progress which can result in the birth of exciting new products requiring new plans and equipment third demand forecasts are projections of demand for companies products or services these forecasts are also called sales forecasts drive a company's production capacity and scheduling system and serve as inputs to financial marketing and personal planning economic
• 12:30 - 13:00 and technological forecasting are specialized techniques that may fall outside the role of the operations manager and maybe the emphasis will be therefore be on the demand forecasting the strategic importance of forecasting good forecasts are critical importance in all aspects of business the forecast is the only estimate that demands until actual demand becomes known forecasts of demands therefore drive
• 13:00 - 13:30 decisions in many areas so there are three first human resources second capacity third is the supply chain human resources hiring training and laying off of workers all depend on anticipated demand if the human resources department must hire additional workers without warning the amount of training declines and the quality of workers workforce suffers capacity when capacity is inadequate the
• 13:30 - 14:00 resulting shortages can be can mean undependable delivery loss of customers and loss of market share this can be happened when when a company underestimated a huge demand for its product even with production lines working overtime the company could not keep up with the demand and lost its customer when excess capacity is built on the other hand costs will be skyrocketed
• 14:00 - 14:30 third will be the supply chain management good supply relations relations and the ensuing price advantages for materials and parts depends on accurate forecasts for example an auto manufacturer we want to guarantee sufficient airbag capacity must provide accurate forecasts to justify their order in the global marketplace were expensive components of that product
• 14:30 - 15:00 are manufactured in dozens of countries coordination driven by porkas is critical such as the transportation scheduling the final assembly from the plant or the maker of that product to your to your endpoint so basically supply chain management will also be one key factors in strategic importance of forecasting
• 15:00 - 15:30 so seven steps in the forecasting system forecasting follows seven basic step so we will use like that topper corporation so first determining determine the use of forecasts tupperware uses demand forecasts to drive production at each of its 13 plants select items to be forecasted for tupperware there are 400 products each with its own sku or their stock giving unit topper were like other firms of this type does
• 15:30 - 16:00 that does demand forecasts by families or group of skus determine the time horizon of the forecast is it is like a short medium or long term tupperware develops forecast monthly quarterly and 12 months sales projections select forecasting models tupperware uses a variety of statistical models that we will discuss including moving averages exponential smoothing and regression analysis
• 16:00 - 16:30 they also employs judgmental and non-quantitative models gather the data needed to make the forecast top errors world uh headquarters maintains huge databases to monitor the sale of each product next make the pro the forecast and lastly validate and implement the results at top everywhere forecasts are reviewed in sales marketing finance and production departments to make sure that
• 16:30 - 17:00 the model assumptions and data are valid error measures are applied and then they will be using the forecast um the forecasted the forecasted amount to schedule materials equipment and personnel at each plant the realities regardless of the system that firms like topper were used each companies faces several realities
• 17:00 - 17:30 first forecasts are seldom perfect this means that outside factors that we cannot predict or control often impact the forecast companies needs to allow for this reality second most forecasting techniques assume that there is some underlying stability in the system consequently some firms automate their predictions using computerized forecasting software then closely monitor only the product
• 17:30 - 18:00 items will be like there will be erratic demand third both product family and aggregated forecasts are more accurate than individual product forecasts like in tupperware for example aggregates product forecasts by both family for example mixing bowls versus cups versus storage container and the region that they are selling this approach helps balance the uber and the
• 18:00 - 18:30 the over and under predictions of each of the product in every country that they're selling their product forecasting approaches there are two general approaches to forecasting just as there are two ways to tackle old decision modeling one is quantitative analysis the other is qualitative approach qualitative methods used when situation is vague and little data access like new products new technology it involves
• 18:30 - 19:00 intuition experience such as forecasting sales on internet or based on service next will be the quantitative methods used when situation is stable and historical data exists existing products or there is a current technology involves mathematical techniques such as forecasting sales of color television and more overview of qualitative methods
• 19:00 - 19:30 in this section we consider four different qualitative forecasting techniques first jury of executive opinion under this method opinions of group of high-level experts or managers often in combination with statistical model are pulled to arrive at group estimate of demand second delphi method there are three different types of participants in delphi method the decision makers staff personnel and respondents
• 19:30 - 20:00 decision makers usually consist of group of 5 to 10 experts who will be making the actual forecast stop uh staff personnel assist decision makers by preparing distributing collecting and summarizing a series of questionnaire and survey results the respondents are a group of people often located in different places whose judgments are valued this group provides inputs to the decision makers before forecast is made
• 20:00 - 20:30 next sales for composite in this approach each salesperson estimate what sales will be his or her region these forecasts are then reviewed to ensure that they are all realistics then they are combined at the district and national levels to reach an overall forecast a variation of this approach occurs like for example a car manufacturing company
• 20:30 - 21:00 where every quarter that company have to make a meeting at this meeting they will talk about what is selling in what colors and what options so the factory knows what to build fourth consumer market survey this method solicits input from customer or potential customers regarding future purchasing plans it can help not only preparing a broadcast but also improving product design and planning for new
• 21:00 - 21:30 products the consumer market service and salesforce composite methods can however suffer from overly optimistic forecasts that arise from customer input overview of quantitative methods five quantitative forecasting methods all of which use historical data are described in this ppt or in this slide for time series models it includes the naive approach moving averages exponential smoothing and trend
• 21:30 - 22:00 projection for associated model it includes the linear regression time series models is a time series models predict on the assumption that the future is a function of the past in other words they look at what has happened over a period of time and use a series of past data to make a forecast if we are predicting weekly sales of lawn mowers we will use the past weekly sales of lawn mowers when making the forecast
• 22:00 - 22:30 associated models or casual or the causal models i mean such as linear regression incorporate the variables or factors that might influence the quantity being forecast for example an associative model for lawn mower sales might include such factors as new as new housing starts advertising budget and competitors price time series forecasting a time series is
• 22:30 - 23:00 based on a sequence of evenly space like weekly monthly quarterly and so on examples include weekly sales of nike air jordans quarterly earnings reports of microsoft stock daily shipments of beers and annual consumer price indices forecasting time series data implies that future values are predicted only from past values and that other variables no matter how potentially valuable valuable
• 23:00 - 23:30 there are so the decomposition of time series analyzing time series series means breaking down past data into components and then projecting them forward a time series has four components trend seasonality cycles and random variations trend is a gradual upward or downward movement of data over time
• 23:30 - 24:00 changes in income population age distribute age distribution or cultural views may account for movement in trend second seasonality is a data pattern that repeats itself after a period of days weeks months or quarters there are six uh common seasonality patterns like week should be like that the season will be like the land the season length will be like the day month will be week
• 24:00 - 24:30 month will be day year will be quarter for year also it can be month for year it can be week third will be the cycles our path are patterns in the data that occur every several years they are usually tied into the business cycle and are of major importance in short-term business analysis and planning predicting business cycles is difficult because they may be affected by political events or by international
• 24:30 - 25:00 turmoil fourth random variations are the blips in the data caused by chance and unusual situation they follow no discernible pattern so they cannot be predicted naive approach the simplest way to forecast is to assume that the demand in the next period will be equal to the demand the most recent period like here in this example it may sales were 48 then june sales will be 48
• 25:00 - 25:30 so as easy as that sometimes cost effective and efficient moving average moving average forecast uses a number of historical actual data values to generate a forecast moving average are useful if you can assume that market demands will stay fairly steady over time a four month moving average is found by simply summing the demand during the
• 25:30 - 26:00 past four months and dividing by four with each passing month the most recent month data are added to the sum of the previous three months and the earliest month is dropped this practice tends to smooth out short-term irregularities in the data series mathematically the simple moving average which which served as an estimate of the next year's demand is expressed as moving average is equal to summation of
• 26:00 - 26:30 the demand in previous n periods over n where n is the number of periods in the moving average for example four five or six months respectively for a four five or six period moving average so we will be having one example here moving average example so as you can see there will be like one two three four five six seven seven one period and the actual sales and three month moving
• 26:30 - 27:00 averages will be indicated here so if you wanted to note a three month moving average it will be starting from january february march and you can predict the april so you just need to add this three months the 10 12 and 13 you just added and divide into three so basically the result will be 11 and two thirds so it is like 11.67 or we can round up that one into 12.
• 27:00 - 27:30 so you can just repeat it and you will just move one from january it will be now february march april so it will be 12 plus 13 plus 16 divided by 3 and this will be this will be the answer and so on so as simple as that graph of moving average so this is the the actual sales
• 27:30 - 28:00 and then the moving average forecast so there will be like a very variability on the results however it will be just based on this one it is small but however like this one we don't know if there will be like some very variability on the demand for a specific product or specific season so we don't know if the our forecast will be like this again so the forecast will be depend on the
• 28:00 - 28:30 on how the the market will behave next will be the weighted moving averages use when trend is present older data usually less important weights based on experience and intuition so the formula will be like weighted moving average is equal to summation of weight for period and multiply the demand in period n over the
• 28:30 - 29:00 summation of weights weighted moving average example first is like there will be like weights applied for last month like three two months ago will be two three months ago will be one and the sum of weights will be six so let's see our example so from here the first one will be like march it will be like last month so the weight will be
• 29:00 - 29:30 three multiplied by the actual sales multiplied by the actual sales here from here and this one last month will be here here plus two months ago will be the weights will be to multiply by the february two months from april plus 10 which is the january so it will be multiplied by one so no need to multiply divided by the total weights will be three plus two plus one will be six so
• 29:30 - 30:00 the total will be twelve point i mean twelve and one sixth so basically that's how you do the weighted moving average and so on potential problems with moving average first increasing the end smooths the forecast but makes it less sensitive to changes second do not forecast trends well so sometimes the the the actual demand to the forecasted
• 30:00 - 30:30 demand will be erratic there will be erratic um differences between them requires extensive historical data since you need to to forecast a good you need to have an accurate forecast as much as possible you need to have a huge number of historical data in order for you to do that one moving average and weighted moving average so this is the the figure shown in the slide
• 30:30 - 31:00 so this is the actual sales this is the moving average and the weighted moving average so there will be like a slightly difference between the moving average and the weighted moving average but the weighted moving average is more close to the actual sales in this figure exponential smoothing form of weighted moving average weights decline exponentially most recent data awaited most
• 31:00 - 31:30 required smoothing constant which is the alpha ranges from zero to one subjectively chosen by the experts it involves literal record keeping of past data the basic exponential smoothing formula can be shown as follows new forecast is equals to last periods forecast plus the alpha of the last year's actual demand minus the last period forecast it can be represented as this one ft is
• 31:30 - 32:00 equals to ft minus 1 plus alpha multiplied by a t minus 1 minus ft minus 1. ft will be the new forecast f sub t minus 1 is equals to previous forecast and the alpha is equals to smoothing or weighting constant that will be arranged from 0 to 1 only exponential smoothing example
• 32:00 - 32:30 predicted demand will be 142 ford mustangs the actual demand will be 153 and the smoothing constant alpha will be 0.20 so this will be the solution the new forecast is equals to 142 which is the predicted demand plus the alpha 0.2 multiplied by 153 minus 142 exponential smoothing example
• 32:30 - 33:00 in january a car dealer predicted february demand for 142 ford mustangs actual february demand was 150 autos using a smoothing constant chosen by management of alpha 0.20 we can forecast march demand using the exponential smoothing model substituting the sample data in the formula we obtain new forecast is equals to 142 plus 0.2 multiplied by 153 minus 142
• 33:00 - 33:30 is equals to 144.2 or equals to 144 cars the smoothing constant alpa is generally in the range of 0.05 to 0.50 for business application it can be changed to give more weight to recent data when alpha is high or more weight than past data when the alpha is low when the alpha reaches the extreme of
• 33:30 - 34:00 1.0 the equation that we have gotten the ft is equals to 1.0 a sub sub t minus 1 all the older values will drop out and the forecast becomes identical to the eb model mentioned earlier in this lesson that is the forecast for the next period is just the same as the periods of current demand effect of smoothing constant the
• 34:00 - 34:30 patterning table helps illustrate this concept for example when the alpha is equal equals to 0.5 we can see that new forecast is based on almost entirely on demand in the last three or four periods when the alpha is 0.1 the forecast places little weight on recent demand and takes many periods about 19 of historical values into account selecting the smoothing constant the
• 34:30 - 35:00 exponential smoothing approach is easy to use and it has been successfully applied in virtually every type of business however the appropriate value of smoothing constant alpha can make the difference between an accurate forecast and an inaccurate forecast high values on up and alpha are chosen when the underlying average is likely to change low values of alpha are are used when the underlying average is fairly stable
• 35:00 - 35:30 in picking a value for smoothing constant the objective is to obtain the most accurate forecast so measuring the forecast error we generally do this by selecting the model that gives us the lowest forecast error so forecast errors is equals to actual demand minus the forecast value or 80 minus ft several measures are used in practice to calculate the overall forecast error
• 35:30 - 36:00 these measures can be used to compare different forecasting models as well as to monitor forecasts to ensure they are performing well three of the most popular measures are the mean absolute deviation deviation or the m80 mean measured error or the mse and the mean absolute percent error or the maa mape so these are the formula for the ones
• 36:00 - 36:30 that i have mentioned earlier for mid is equals to summation of the absolute value of the actual cost minus the forecast over n or the number of sample or the number of periods mean squared error mse is equals to mse is equals to summation of the forecast error squared over n or the period mean absolute percent error is equals to 100 multiplied by the
• 36:30 - 37:00 summation of the actual minus the forecast divided by the actual or the percentage divided by n comparison of forecast error so an example during the past eight quarters the port of baltimore has unloaded large quantities of grain from ships the ports operation managers want to test the use of exponential smoothing to see how well the technique works in
• 37:00 - 37:30 predicting tonnage unloaded he guesses that the forecast of grain unloaded in the first quarter was 175 tons the values of alpha are examined at point 10 and point 15. so basically these are the actual uh damage unloaded this are the rounded forecasts exponential smoothing of 0.10 and 0.50 basically they use the exponential smoothing formula for this one and they get this value
• 37:30 - 38:00 all of this value and this one 4.5 and 0.10 so absolute deviation for alpha point 10 is just the actual minus the forecast so 180 minus 175 is equals to 5 168 minus 176 is equal to 8 since it should be absolute value 159 minus 175 is 16 and so on same with the absolute deviation value for 0.50 so 180 minus 75 168 minus 178 and so on
• 38:00 - 38:30 so now we will compute the mad med is equals to summation of absolute deviation over n for a is equals to point 10 it the total of deviation is equals to 84 divide eight since there is eight quarters the results of mad or means absolute i mean absolute deviation is 10.50
• 38:30 - 39:00 and for alpha point fifty is equals to one hundred divided by eight that is equivalent to pole point fifty on the basis of this analysis a smoothing constant of alpha point ten is prepared to alpha of 0.50 because it's mad is much smaller so mean squared error so first we just need to square all this value the mean absolute deviation
• 39:00 - 39:30 for both the alpha and then we will get the for example here 25 and get we will get the sum of this one and divided by the total number of periods so next slide so here so the mse formula is equal to summation of forecast error squared over n for point 10 is equals to 1558 divided by 8 which is 194.75 and for alpha point 50 is equals to
• 39:30 - 40:00 201.50 so basically we wanted a smaller mse so we'll again choose the alpha 0.10 mean absolute person error map the average of the absolute differences between the forecast and actual values expressed as a percent of actual values a problem with both mid and mse is that their values depend on the magnitude of the item of being forecasted if the forecast item is measured in thousand
• 40:00 - 40:30 the mad and mse values can be very large to avoid this problem we can use the mean absolute percent error map this is computed by the as the average of the absolute difference between the forecasted and actual values expressed as percentage of the actual values so that is next slide map is equals to 100 multiply the deviation over the
• 40:30 - 41:00 actual over n so for alpha point 10 is equals to 45.62 over eight so how can we get it so first we need to for this one this calculation so 100 multiplied by 5 divide 180 which is the actual
• 41:00 - 41:30 is equals to something percentage i think 2.78 and then we will add all the percentage that we will be getting here so we will get the summation of that percentage and divided by the total number of periods which is eight so it is so for 0.5 is equals to point i i mean 6.85 so based on the result
• 41:30 - 42:00 still alpha point 10 has the lower percent of error so we will choose the 0.10 so comparison of forecast errors as you can see the smoothing constant of 0.10 has the lowest of all the mid mse and maple so therefore we need to choose this one exponential smoothing with trend adjustment simple exponential smoothing
• 42:00 - 42:30 the technique we just illustrated in the previous slide is like any moving average technique it fails to respond to the trends other forecasting techniques that can deal with trends are certainly available however because exponential smoothing is such a popular modeling method or approach in business so let's take a look in this exponential smoothing with trend adjustment to improve our forecast let us illustrate a more complex
• 42:30 - 43:00 exponential smoothing model one that adjusts for trend the idea is to compute an exponentially ex exponentially smooth average of the data and then adjust for the positive negative log in the trend the new formula will be f i t or the forecast including trend is equals to exponential's exponentially smoothed forecast or f sub t plus t sub t or the exponentially smoothed trend with trend adjusted exponential
• 43:00 - 43:30 smoothing estimates for both the average and trend are smoothed these procedures requires two smoothing constant the alpha for the average and the beta for the trend when you compute the average trend each period is equals to ft is equals to alpha multiplied by the actual demand last period plus 1 minus alpha multiplied the forecast last period plus trend estimate last period so ft stands for the exponentially
• 43:30 - 44:00 rooted forecast of the data series in period t t sub t is equals to exponentially smoothed trend in period t a sub t is equals to actual demand in period t alpha is the smoothing constant for the average which is ranging from zero to one for beta smoothing constant for the trend that is ranging from zero up to one so there there are three steps to
• 44:00 - 44:30 compute the trend adjusted forecast first compute for the f sub t the exponentially smoothed forecast for period 3 using the equation this one this equation compute this mode that trend which is the t sub t using this equation calculate the forecast including trend f i t by a formula f i t is equals to f t plus t sub t so we just need to add this one together
• 44:30 - 45:00 this is the formula exponential smoothing with trend adjustment example so a large portland manufacturer uses exponential smoothing to forecast demand for pace of pollution control equipment it appears that an increasing trend is present so based on here as the month increases the demand also increases so for month one up to month nine the demand is continuously increasing so
• 45:00 - 45:30 there's a trend for this example so smoothing constant are assigned the values of alpha which is 0.2 and the beta is 0.4 assume the initial forecast for one month f1 was 11 units and the trend over the period t sub 1 was 2 units and f i t just need to add this one the f t and p sub t is equals to 13.
• 45:30 - 46:00 step 1 forecast for month 2 f 2 is equals to this is the equation or the formula is equals to alpha a 1 plus 1 minus alpha multiplied by f 1 plus t sub 1 so basically this is the computation m sub 2 is equals to 0.2 which is the alpha the a1 is 11 i mean 12 the actual demand and then plus 1 minus alpha which is 1
• 46:00 - 46:30 minus 0.2 multiplied by f t which is 11 here plus 2 which is the trend so it is is it it it is equal to 12.8 units so as easy as that you already got your ft next step 2 is the trend for month 2 so the equation of the formula is equals to beta multiplied by f sub 2 minus f sub 1 plus 1 minus beta multiplied by t sub 1.
• 46:30 - 47:00 so t2 is equals to beta which is 0.4 multiplied by 12.8 which is the f2 minus 11 from the previous plus one minus point four which is the beta multiplied by 32 this one t1 t sub one so it it is it this one is equals to 1.92 units then you just need to calculate the f i t for
• 47:00 - 47:30 1 2 so basically you just need to add this to is it is equal to 14.72 units and then you can compute it by your own so the resulting forecast with trend will be 35.16 so to see it more graphically so this is the actual demand and the forecast
• 47:30 - 48:00 including with trend which is more reliable and more nearer with the actual demand trend projection the last time series forecasting method we will be discussing is the trend rejection these techniques is like a beating trend line to a series of historical data points and then projects the line into the future of the medium to long range forecast several mathematical trend equations can be developed for example exponential and
• 48:00 - 48:30 quadratic but in this section we will look at linear or straight line trends only so this is also known as the linear regression if we decide to develop a linear trend line by precise statistical method we can apply the least square method this approach results in a straight line that minimizes the sum of squares of the vertical difference of the deviation from the line to each of the actual observation
• 48:30 - 49:00 so this is the formula y hat which is the computed value for the variable to be predicted or the dependent variable is equals to a which is the y-axis intercept plus b which is the slope of the regression line multiplied by the x or the independent variable this square method
• 49:00 - 49:30 so this is the trend line or the y hat is equals to a plus b x so this is the line and there will be like the actual observation or the y value here so this is the data points the all the stars and there will be like deviation from the line to the observed value so as you can see is the deviation so the least square method for finding the line finding the best fitting straight line where the asterisk are the location of the seven actual observation of the data points
• 49:30 - 50:00 so basically the statisticians have developed equations that we can use to find the values of a and b for any regression line the slope b is found by b is equals to summation of x y minus n multiplied by x bar multiplied by y bar over summation of x squared minus n bar x x bar squared so where b is the slope of
• 50:00 - 50:30 the regression line x known values of independent variable y is the known values of dependent variable x bar is the average of the x values and y bar is the average of the y values and n is the number of data points or observation and then we can compute the y-intercept a as follows is equals to y bar minus b x bar
• 50:30 - 51:00 so here as i mentioned previously this is the y bar i mean y hat is equal to e plus v x the b can be computed by d summation of x y minus n x bar multiplied by y bar over summation of x squared minus n multiplied by x bar squared for a you can compute this by y
• 51:00 - 51:30 bar minus b x bar so basically this is how you calculate it manually but if you have like statistical tools like the spss you can easily use it and it will automatically give you the the values for your for the ab values and also the line please square example so as you can see there will be like
• 51:30 - 52:00 a year here so i will read the question first the demand for electric power at ny edison over the period of 1999 to 2005 is shown in the following table in megawatts let's forecast 2006 demand by fitting a straight line trend to this data so this is the year sorry so these are the years this is the time period like one to seven and this is the power electrical power
• 52:00 - 52:30 demand which is in megawatts so first thing first we need to do like a time period summation which is 28 and the average will be four so just 28 divided by 7 is equals to 4 right so that's the average 28 divided seven is 4. and the electrical demand you just need to sum it up
• 52:30 - 53:00 and then you need to take the average which is 692 is the sum and the average is 98.86 and next you need to get the summation of x squared so basically you just need to square this one square this one so 1 4 9 16 25 36 and 49 and you need to sum it down so it equals to 140 and then the multiplication of x and y this and this so you have to multiply 1
• 53:00 - 53:30 multiply by 74 2 multiply by 7 is x 3 multiply by 80 something like and so on and so forth so the summation of x and y you just need to sum all the results and it it will be equivalent to 3063 so next we need to compute for the b and a so for our a which is the with the a is equals to summation of x y
• 53:30 - 54:00 minus n multiplied by x bar multiplied by y bar over summation of x squared minus and multiplied by x bar squared so we just need to substitute the values that we have computed earlier so 3063 minus 7 which is the time period or the end multiplied by 4 which is the average summation of time period multiply multiply by the 98.86 which is
• 54:00 - 54:30 the y bar here and then summation of x squared is equals to this one 140 minus n which is 7 multiplied by x bar which is 4 n squared so the value of p will be 10.54 a is equals to y bar minus b times times x bar is equals to 98.86 from here the y bar minus b which is
• 54:30 - 55:00 10.54 multiplied by x bar which is four so the a is equals to 56.70 so the trend line based from the computation of b and a is equals to y hat is equal to 56.70 plus 10.54 x so x you can put here like 8 9 10 so you can predict what's the value of your y hat
• 55:00 - 55:30 so here is the graph for the square example so this is the trend line and this one we can estimate the demand for 2007 this one and 2006 to see if our projection will be correct so to check the validity of the model you plot historical demand and the trend
• 55:30 - 56:00 line as shown in this figure so the demand in 2007 is like 152 from here megawatts so basically um the the trend or the possession of this or the equation that we have done is nearly correct or near the optimum notes on the use of the least square method using the lead square method
• 56:00 - 56:30 implies that we have met three requirements first we always plot the data because these squares data assume a linear relationship if a curve appears to be to be present curvilinear analysis is probably needed second we do not predict time periods far beyond or given database for example if you have 20 months worth of average prices of microsoft stock we can forecast only three or four months into the future forecast beyond that have
• 56:30 - 57:00 little statistical validity thus you cannot take five years worth of sales data and project 10 years into the future because it will be more are uncertain or there will be more variability into that deviation around the least square line as you can see on the figure are assumed to be random they are normally distributed with most observation close to the line and only a smaller number farther out
• 57:00 - 57:30 seasonal variations in the data seasonal variations in data are regular up and down movements in a series that relates to recurring events such as weather or holidays demand for coal and fuel oil for example peaks during cold winter months demand for gold clubs or suntan lotion may be high as in in summer seasonality may be further applied to early daily weekly monthly or other
• 57:30 - 58:00 recurring patterns fast food restaurants experience daily surges at noon and again at by vm movie theaters see higher demand on friday and saturday evenings the post office and other stores or christmas stores also exhibit seasonal variation in customer traffic and sales in what is called multiplicative seasonal model
• 58:00 - 58:30 seasonal factors are multiplied by an estimate of average demand to produce a seasonal forecast our assumption in this section is that trend has been removed from the data otherwise the magnitude of seasonal data will be distorted by the trend here are the steps we will follow for a company that has seasons for one month first find the average historical demand each season or month in this case by summing the
• 58:30 - 59:00 demand for the month in each year and by dividing the number of years of data available second compute the average demand over all months by dividing the total average annual demand by the number of seasons for example if the total average demand for a year is 120 units and there are 12 seasons each month the average monthly demand is 120 divided by 12 which is 10 units compute seasonal index for each season
• 59:00 - 59:30 by dividing the month's actual historical demand from step one by the average demand over all months from step two for example if the average historical january demand over the past three years is eight and the average demand over old man is 10 units the seasonal index for january is equals to 8 over 10 or equals to 0.8 likewise a seasonal index of 1.2 for february would mean that february
• 59:30 - 60:00 demand is 20 larger than the average demand over all months fourth estimate next year's total demand and lastly divide the estimate of total annual demand by the number of seasons then multiply it by the seasonal index for the month this provides the seasonal forecast seasonal index example monthly demand for ibm laptops computer
• 60:00 - 60:30 at this mine's distributor for 2003 to 2005 is shown in the following table first the average demand was computed by adding the 80 this one just need all of this one and get the average so it's equal to 90 and so on and then you will get the sum of this and get the average which is which is 94 and then after computing that
• 60:30 - 61:00 this one so we just need to get for example the average from 2003 to 2004 you will get the average of that and divide the average monthly and you will get the seasonal index then you can just do it and so on and so forth so here just like 80 divided by 94
• 61:00 - 61:30 which is eighty pi point one eighty pi divided by ninety four one hundred by ninety four so and so on and so forth next we will be expect so next forecast for 2006 since it's just for 2005 right so january is equals to 1200 because the expected annual demand
• 61:30 - 62:00 will be 1200 divided by 12 for uh for the month multiply by 0.96 is at 4.957 which is 96. so the expected forecast for january 2006 will be 96. for example 2006 will be 96. and for february you just also repeat it 12 1200 divided by 12 multiply by the seasonal index which is 85 and so on and
• 62:00 - 62:30 so forth now this is the figure showing the forecasted demand for each period and each year so here you can see that in the month of march there is an increase on the sales and it will be declined continuously declining after the december
• 62:30 - 63:00 okay and each year there is like a continuous trend but however for 2005 and 2006 there is like a minimum um minimum movement on the demand and the forecast associative forecasting method that includes the regression and correlation analysis unlike time series forecasting a social associative forecasting models usually
• 63:00 - 63:30 consider several variables that are related the quantity being predicted once these related variables have been found a statistical model is built and used to forecast the item of interest this approach is more powerful than the time series methods that use only the historical values for the forecasted variable many factors can be considered in an associative analysis for example sales of dell pcs may be related to
• 63:30 - 64:00 dell's advertising budget the company's prices competitors prices and promotional strategies and even the nation's economy and unemployment rates in this case pc sales would could be or would be the dependent variable and the other variables would be the independent variables the monitor's job is to develop the best statistical relationship between the pc sales and the independent variables the most common quantitative associative
• 64:00 - 64:30 forecasting model is linear regression analysis so using regression analysis of forecast we can use the same mathematical model that we employed in the least square method of trend projection to perform a linear regression analysis the dependent variables that we want to forecast will still be y hat but now the independent variable x next need no longer be time
• 64:30 - 65:00 we use the equation y hat is equals to a plus b x where y hat is equals to value of dependent variable in our example it will be sales a which is the y-axis intercept b is the slope of the regression line and x will be the the independent variable associative forecasting example so notable construction company renovates
• 65:00 - 65:30 old homes in west bloomfield michigan over time the company has found a dollar volume with an op renovation work is dependent on the best bloomfield area payroll the following table lists no dell's revenue and the amount of money earned by wage earners invest bloom field during the past six years so it includes the nodal sales and the local payroll so nodal management wants to establish a mathematical relationship to help predict sales first it needs to
• 65:30 - 66:00 determine whether there is a straight line or linear relationship between area pair and sales so it plots the known data on scatter diagram so this is the scatter diagram oops and now this is how we computed earlier right using the least really sum square so we will just be summing up the y
• 66:00 - 66:30 the x and we will do the x squared as well so just do the square root of this and this will be the results and just sum it all so this is the result so x and y again you just need to multiply these two and now we will be substituting it again for um from the equations of b and a so this is these are the results so b is equals to 0.25 and a is 1.75 so let's see the results
• 66:30 - 67:00 so y hat is equals to 1.75 plus point 25 x so sales is equals to 1.75 plus point 25 of payroll so based on this this will be the equations that can predict the sales based on the its relationship to the payroll and here is the comparison of the data points versus the regression line
• 67:00 - 67:30 if payroll next year is estimated to be 600 million then sales of 1.75 plus 0.25 multiply by 6 sales will be approximately 325 thousand dollars so here the sales is 600 million this will be the approximate sales if the payroll i mean is 600 million dollars
• 67:30 - 68:00 standard error of estimate the forecast of 325 thousand dollars of nodal sales in example is called point of estimate.y the point estimate is really the mean or expected value of a distribution of possible values of sales this figure shows this concept to measure the accuracy of the regression estimate we must compute the standard error of the estimate which is the s y x
• 68:00 - 68:30 here s y x this computation is called the standard deviation of the regression it measures the error from the the dependent variable y to the regression line rather than to the mean so as you can see in this equation it is similar expression to that two that found in most statistical books for computing the standard deviation of arithmetic mean so
• 68:30 - 69:00 s y x is equal to square root of summation of y minus y c squared over n minus 2. so y is equals to y value of each data point y is the computed value of the dependent variable from the regression equation and n is the number of data points confused computationally this equation is considerably easier to use so here we use the standard error to set up the
• 69:00 - 69:30 predictions intervals around the point estimate so it is based on the result of the regression line so y and this my summation of y squared minus a summation of y minus b summation of x y over n minus 2. so here is the results so the result of the standard error estimate is 0.306 the standard error of the estimate is
• 69:30 - 70:00 around 30 600 in sales so there will be a variation on that about the predicted and the possible results correlation of coefficients of regression line the regression equation is one of the one way of expressing the nature of relationship between two variables regression lines are not cause and effect relationships they are merely describe the relationship among variables the regression equation shows
• 70:00 - 70:30 how one variables relates to another value and changes in another variable another way to calculate the relationship between two variables is to compute the coefficient of coefficient of correlation this measure expresses the degree of strength of linear relationship usually identified as r the coefficient correlation can be any number between positive one and one
• 70:30 - 71:00 to compute r we use much of the same data needed earlier to calculate a and b for the regression line the rather than the equation for r is r is equals to n summation of x y minus summation of s x and summation of y over the square root of n and the summation of x squared minus summation of x squared this is the s will be like also
• 71:00 - 71:30 the summation for this one and multiply by n summation of y squared by the summation of y squared y squared so correlation coefficient still this one will be the formula so perfect positive is will be like a straight line going up upward the correlation will be positive one
• 71:30 - 72:00 and positive correlation will be like zero between zero the r will be great between zero and one and then no correlation at all will be straight line and the data points are scattered here some of the data points are scattered and not within the line next it is perfect negative correlation r will be minus one so perfectly
• 72:00 - 72:30 downwards and the point fitted to the line although the coefficient of correlation is the measure of most commonly used to describe the relationship between two variables another measure does exist it it is called the coefficient of the termination and is simply the square of the coefficient correlation namely r squared
• 72:30 - 73:00 the value of r squared will always be a positive number that range from zero and the r squared from zero to one or equal to one the coefficient of the termination is the percent of variation of the dependent variable y that is explained by regression equation in adults case the value of r squared is 0.81 indicating that 81 of the total variation is explained by the regression equation
• 73:00 - 73:30 here multiple regression analysis multiple regression is a practical extension of simple regression model we just explored it allows us to build a model with several independent variables instead of just one variable for example if no constructor construction wanted to include average annual interest rate in its model for forecasting renovation sales the proper equation would be y hat
• 73:30 - 74:00 is equals to a plus b one multiplied by x one plus b two x two and so on so y hat still the dependent variables such as the sales a will be the constant x one and x two values of the two independent variables area payroll and interest rate respectively b1 and b2 is the coefficient for the two independent variables
• 74:00 - 74:30 so this is the new equation in an example so y hat is equals to 1.80 plus point 30 x1 minus 5.0 x2 an improved correlation coefficient of r is equals to 0.96 means this model does a better job of predicting the change in construction sales so here sales is equals to 1.8 plus 130 multiplied by 6 minus 5.0 multiplied by 0.12 is equals to 3 or around 300 000
• 74:30 - 75:00 dollars so basically we will not compute it manually we can just use our our statistical tools in doing this one since it will be it will take long for us to do this one monitoring controlling forecast so tracking signal one way to monitor forecasts is to ensure that they are performing well is to use a tracking signal a tracking
• 75:00 - 75:30 signal is a measurement of how well the forecast is predicting actual values as forecasts are updated every week month or quarter the newly available demand data are compared to the forecast values the ratio of running sum of forecast error r s f e to mean absolute deviation m a d indicates that good traveling signals
• 75:30 - 76:00 has low values if forecasts are continually high or low the forecast has a bias error next monitoring and controlling forecasts still traveling signals is equals to rsfe over m80 so traveling signals is equals to summation of actual demand in period i minus forecast demand in period i divided by the summation of absolute value of the actual minus forecast over n
• 76:00 - 76:30 so here positive trapping signals indicate that demand is greater than forecast negative signals mean that demand is less than a forecast a good tracking signal that is one with low rsfe has about much positive error at his has negative error in other words small deviations are okay but positive and negative error should be balance one another so that tracking signal centers
• 76:30 - 77:00 closely around zero a consistent tendency for forecast to be greater or less than the actual values that is for a high rsfe is called bias error here so here one example so first we need to get uh to [Music] to get the [Music]
• 77:00 - 77:30 the minus 100 so the error will be negative 10 then 95 minus 100 it will need negative five 115 minus 100 will be approximately 15 100 minus 110 will be negative 10 125 minus 110 will be 15 and 140 minus 110 will be positive 30. so basically here will be the error so rf s rsfe will be the cumulative so negative 10 will be
• 77:30 - 78:00 here plus negative 10 um plus negative 15 i'm a negative 5 i mean will be negative 15 plus 15 will be zero so zero plus negative ten will be ten negative ten plus fifteen will be positive by five plus thirty will be thirty five so this will be the absolute forecast error here so basically we'll just copy this one but it should be in the in the absolute value and the cumulative absolute forecast error will be just the
• 78:00 - 78:30 sum of this 1 10 10 plus 5 will be 15 15 plus 15 will be 30 30 plus 10 will be 40 40 plus 15 will be 55 5 plus 30 will be 85 and then you will get the m80 and then get the m80 by like one like 10 divided by 1 15 divided by 2 30 divided by 3 equals to 10 40 divided
• 78:30 - 79:00 by 4 is equal to 10 55 divided by 5 is equal to 11. 85 divided by 6 will be 14.2 so that's how you compute it so here so the variation of the traveling signal between negative 2 and positive 2.5 respectively since you just need to [Music] here the rfs rsfe divided by the m80
• 79:00 - 79:30 which is equal to negative 1. negative 15 divided by 7.5 is equal to negative 2 0 divided by 10 fifth negative 10 divided by 10 positive pi divided by 11 35 divided by 14.2 so here adaptive forecasting refers to computer monitoring of tracking signals and cell adjustment if a signal
• 79:30 - 80:00 passes a preset limit for example when applied to exponential smoothing the alpa and beta coefficients are first selected on the basis of values that minimize error forecasts and then adjust it accordingly whenever the computer notes an errant trapping signal this process is called adaptive smoothing focus forecasting rather than adapt by choosing a smooth constant computers
• 80:00 - 80:30 allow us to try a variety of forecasting model such as an approach is called focus forecasting focus forecasting is based on two principles sophisticated forecasting models are not always better the symbol one second there is no single technique that should be used for all products and services so this approach uses historical data to test multiple forecasting models for individual items the forecasting model with lowest error is then used to forecast the next demand
• 80:30 - 81:00 forecasting and service sector forecasting in the service sector presents some unusual challenges a major technique in the retail sector is tracking demand by maintaining good short-term records for instance a barber shop catering to men expect big lows on fridays and saturdays indeed most barbershops are closed on sunday and monday and many call it extra help on
• 81:00 - 81:30 friday and saturday a downtown restaurant on the other hand may need the track conventions and holiday for effective short term forecasting fast food restaurant forecast so basically this is one of the uh graphs that indicates the percentage of sales based on time which is the lunchtime so as you can see there's a peak here during the lunch time from 12 to 1 and
• 81:30 - 82:00 continuously decreasing again and then increasing again from five to six and seven to seven six to seven i mean so this is the forecast for a fast food restaurant so i think you already know and it's it is already self-explanatory summary a critical part of operations manager function demand forecast drive affairs production capacity and scheduling system and affect the financial marketing and personal planning function there are a variety of qualitative and
• 82:00 - 82:30 quantitative forecasting techniques qualitative approaches employ judgment experience intuition and a host of other factors that are difficult to quantify quantitative forecasting uses historical data and causal or associative relations to project future demands so basically forecasting calculations are seldom performed by hand most operation managers turns to software packages such as forecast pro
• 82:30 - 83:00 the sap the afs the sas spss or excel so basically that ends the lesson too and thank you for listening