پیش بینی فروش دسته بندی و سهم بازار برای مشترکان تلفن های بی سیم : یک روش ترکیبی
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|13955||2002||21 صفحه PDF||سفارش دهید||10523 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : International Journal of Forecasting, Volume 18, Issue 4, October–December 2002, Pages 583–603
The ability to forecast market share remains a challenge for many managers especially in dynamic markets, such as the telecommunications sector. In order to accommodate the unique dynamic characteristics of the telecommunications market, we use a multi-component model, called MSHARE. Our method involves a two-phase process. The first phase consists of three components: a projection method, a ring down survey methodology and a purchase intentions survey. The predictions from these components are combined to forecast category sales for the wireless subscribers market. In the second phase, market shares for the various brands are generated using the forecast of the number of subscribers that are obtained in Phase 1 and the share predictions from the ring down methodology. The proposed methodology produces the minimum Relative Absolute Error for each market as compared to the forecasts from each individual component in the first phase. The value of the proposed model is illustrated by its application to a real world scenario. The managerial implications of the proposed model are also discussed.
Analysts at AT&T forecasted a total of one million cellular subscribers in service by the year 2000. By 1993, the end of the first decade of availability of cellular telephones, there were 16 million cellular telephones in use, with an additional 14,000 new users coming on line each day (Edwards & Dye, 1996). Table 1 compares the number of subscribers as forecasted by different agencies in 1995 to the actual number of subscribers by the end of 2000. As it can be seen in Table 1, different agencies underpredicted the number of cellular subscribers in Europe by the year 2000 to be between 11 million and 20 million. These validate long-held practitioner beliefs as evident in the above quote.The telecommunications industry is facing continuous technological and regulatory changes. Jurisdictional differences are being removed and leading US carriers are forming joint ventures with foreign companies to enter new markets. As the industry becomes more competitive, consumers have benefited through lower prices, which have stimulated telecommunications demand to unprecedented levels. Techniques for forecasting the demands for product offerings over a planning horizon of several years are highly complex. Quantitative forecasting methods such as time series and econometric modeling have become less accurate because the industry no longer has the stable historical relationships that these models rely upon. The forecaster therefore needs to incorporate perceived future industry dynamics into the model (Ozturkmen, 2000). There are several reasons why prediction in the telecommunications industry is a very difficult task. A representative issue is that the boundaries between television, computers and telecommunication products are being progressively eroded through the growth of the Internet and its service providers. Although data and message traffic is growing to enormous levels, this demand is being counterbalanced by substitution of technology (for example, email substitutes telephonic conversation) (Fildes, 2002). The introduction of new information technology significantly affects the demand for telecommunication services (Cristiano, 1993). A person with forecasting responsibilities has three options. As the first option, he or she can use an intuitive approach—preparing forecasts based on his or her best judgment. The second option for the analyst would be to use a quantitative approach—preparing forecasts using statistical techniques such as regression and time series analysis. Finally, a combination of quantitative and intuitive approaches can be used. Specifically, a quantitative approach is used to arrive at a baseline forecast, and then the baseline prediction is adjusted by overlaying judgment to arrive at a prediction interval (Armstrong & Collopy, 1998). Instead of relying on intuition or using judgment, one could take the average of the independent forecasts to arrive at a more reliable forecast. This is the essence of the combination technique. The data can be from different sources or the methods used to forecast can be different, or both (Armstrong, 2001). The literature suggests that including forecasts from different statistical methods generally improves accuracy when significant trends are involved (Armstrong & Collopy, 1998). Using several sources of forecasts can add useful information and may adjust for biases. Instead of trying to choose the single best method, one should combine the results from different methods, which would help in reducing errors arising from faulty assumptions, biases, or mistakes in the data (Armstrong, 2001). For example, Fildes (1991) used three sources: a panel of experts, a naı̈ve extrapolation and an econometric model for forecasting construction output. An equally weighted combined forecast from the three methods reduced the Mean Absolute Error (between predicted construction output and observed output) by 8%. It is difficult to classify the wireless telephone industry as product-based or service-based. Wireless telephones can be categorized as a service with relatively slow replacement rates since the primary utility is enhanced communication facilities. However, they can also be categorized as a product, since the younger generation, in addition to the service features, derives utility from wireless telephones as a fashion accessory. The latter case leads to progressively accelerating replacement rates and low loyalty. The mobile telecommunication churn rate hovers around 30% (the reported values are based on a press release from Telebright Corporation and can be obtained from http://www.telebright.com/PressRelease0530.asp). This churn rate has a direct effect on the market share of each firm. The above mentioned issues make it necessary not only to capture the growth in the product category, but also to capture the shift in sales across firms and technologies (in other words, brand switching). The brand switching behavior can be captured by implementing a brand level diffusion model. In order to implement such a model, one needs brand level data from the introduction of the product in the market. However, consider the following situation: category level sales data is available from introduction (1983/84) until the current time period (1992). Managers need category level subscriber forecasts and brand level market share estimates for the next two years. Given that data on the number of subscribers at the brand level is not available for the past, it is necessary to improvise to be able to address the needs of the manager in generating forecasts. In this study, we propose the use of multiple forecasting techniques to make accurate forecasts of future category sales of a new product and market shares of brands in that product category. In line with the genuine concerns mentioned above, we use a multi-component model, called MSHARE, which involves a two-phase process. In the first phase, a ring down survey methodology, a purchase intentions survey, and a projection method are used to forecast category sales for the wireless subscribers market. In the second phase, estimates of market shares for various brands in the market are generated. The ring down method involves calling up a sample of telephone numbers for each firm in each market to estimate the number of wireless subscribers by brand in each market. In the projection method the total number of subscribers at the national level is estimated from historical secondary data on category sales, and these estimates are then scaled to the market level. The purchase intentions method involves conducting a survey and asking respondents about their likelihood of subscribing to a wireless service in the near future. We propose a methodology to combine the forecasts from these three different techniques that updates weights used in the combination of forecasts in a dynamic fashion. The next section describes the literature on various market share forecasting models. Section 3 discusses our proposed framework, MSHARE. In Section 4 we provide an empirical illustration using data on subscribers for wireless service in three markets in the United States and discuss the results from our framework. We conclude the paper with discussions and implications of our proposed framework.