Technology forecasting has been developed for several decades, and some of methods were derived from other field such as demography. For example, the trend of the adoption of a product will grow slowly at first, and then it will have rapidly growth. Finally, the growth of the adoptions will look like a sigmoid curve. Some quantitative technology forecasting model, such as Fisher-Pry and Gompertz, can offer a precursor when a new product replaces a mature product, and they can offer the probable trend of the market share of a new product. This chapter first presents a brief background to the technology forecasting, and then the problem discussion will be introduced. The research purpose and motivation are also presented with the overall structure of this research in this chapter.
1.1 Background
Technology forecasting has been developing during several decades and is still in the process of improving since many new technological inventions increases the demands of forecasting tools. Many enterprises need these methods in their focus field in order to improve their projection ability and to know the trend as many as possible. Some consultative companies also use technology forecasting method to offer the projection about some products or technologies. Therefore, technology forecasting is comprehensively used now. The technology forecasting can simply divided into two fields according to quantitative or qualitative way. However, these methods truly offer an auxiliary role when managers need to make a decision whether they use quantitative or qualitative method.
Generally speaking, the life cycle of a product or a service will display a
bell-shaped curve, and this curve can be divided into five parts- Innovators, Tornado, Main Street, Decline, and Obsolescence. As same as the life cycle, technology adoption life cycle can also fall into five parts- innovators, early majority, late majority, and laggards (Meade and Rabelo, 2004). Therefore, the growth of adopters in new products or services will look like a sigmoid curve, and they will also decline as sigmoid way. Some technology forecasting methods can be a tool to fit and forecast this trend, such as Fisher-Pry and Gompertz. However, there are many adapted or new models proposed in this field (Carrillo and Gonzalez 2002, Bhargava 1995, Rai and Kumar 2003). These methods can fit data sets well in some specific products or services, such as the rate of adoption of mobile phone, and it portrayed the trend of a new product or a service.
1.2 Problem Discussion
Technological forecasting models give the projection and fit in the form of graph according to estimation of points during a period. The forecasters first may observe the market share or sales volumes data, and the time measures can be annual or quarterly. Although many adapted or new technological growth curve models which used to fit the data were proposed in recent years, not every model can fit or forecast well in different kinds of data sets. Technological forecasting models can be classified according their characteristics (e.g. symmetric, asymmetric, and flexible), the method of estimation, and so on. These classification may offer a simply and clearly way to identify which model belongs to which categories. Consequently, some researchers try to find some criteria to identify which model will have better fit or forecast performance in some particular data sets (Young 1993, Meade and Islam 1997). These classifications can reduce the biases which technological growth models made. For example, the symmetric sigmoid curve can be fitted well by symmetrical
technological growth model (e.g. Fisher-Pry). However, there are more implications for model selection. Choosing the right model is an important work before researchers prepare to project the time series data points. Some models perform a better forecast but fail to forecast the short term time-series data. Another way to reduce the bias is to use the combined models, and this method will also help the forecasters reducing the error when they did not use an appropriate model to fit the data (Meade and Islam, 1997).
Some growth curve may not reach the saturation level at 100%, because the new technologies may enter the market and adopt the share of the mature one before the mature product reach the saturation at 100%. This phenomenon may make some technological growth models (e.g. Fisher-Pry) can not fit these data well. Therefore, when the limit of capacity is unknown, some models will not fit or forecast well in this situation. Experts’ opinion seems a good method to solve this problem if experts suggest a proper limit of capacity. However, if a model can be adjusted its capacity with time-varying, the prediction of limit of capacity can be ignored. Moreover, the number of observations will affect the forecasting accuracy. The more observation points can be obtained, the more accuracy of the curve can be estimated. The stable and robust estimation can be obtained if the data includes the peak of noncumulative adoption curve (Mahajan, et al. 1998).
There are some generalized, adapted, or new technological forecasting models proposed in recent years. The generalized and adapted models are based on the prior common models, such as Fisher-Pry and Gompertz, and adapted them through estimative methods, number of parameters, and so on. Moreover, the saturation level (or called capacity) is limited or not also affect models drawing the fit or predicted performance. There is an interesting idea about the capacity of logistic model, which proposed by Meyer and Ausubel (1999). They argued that the capacity of logistic
curve will also vary during time. This research will try to find whether a dynamic capacity will help the model have better performance.
According to the discussion about the technological forecasting model in this section before, some research directions can be found. First, this model will also have a flexible inflection point in order to fit different kinds of data sets. That can improves that one model is suitable for some data sets whether they belong to symmetric or asymmetric form. Second, dynamic saturation level will help researcher save the process to find the possible saturation level. Finally, the number of data will affect the outcome of forecasts.
1.3 Research Motivation
The growth of a product or a service is interesting process, because its shape looks like a sigmoid curve as similar as the growth of population. Therefore, some methods which used in demographic field were comprehensively used in fitting or projecting the technological growth. These methods are generally called the technological forecasting models or technological growth models. However, not all of data will follow the sigmoid curve. They may follow linear-like or other forms.
Although there are so many technological growth models can be used in fitting different kind of data sets, some improvements can be found in these models and try to find a better method to fit the data. That is why so many adapted or new models were proposed in decades.
The growth of a product or a new service may not easily be observed when only few data can be gathered, but the similar growth trend may also be happened in mature products. For example, people used canals, railways, road, and airways to be the transportation way, and then the growth of maglev may be seen as similar as these mature technologies (Meyer et al., 1999). Therefore, the fit of a growth model not
only can find the trend of data, but also offer a possible way which a new product or service is followed. Moreover, the growth curve can offer some useful information in managerial information. For example, the point of inflection can provide when the products will reach the maximum rate of penetration, and when the curve will reach the saturation level roughly.
1.4 Research Purpose
The growth curve can offer some useful information in managerial information.
For example, the point of inflection can provide when the products will reach the maximum rate of penetration, and when the curve will reach the saturation level roughly. There are many quantitative technological forecasting models in this field, and many generalized form were also proposed (Bhargava1995; Rai and Kumar 2003) in order to improve the forecast. Therefore, the way to utilize these models in the right case is more important nowadays in order to get the better fit and forecasting performance, and some scholars have proposed some related paper about the model comparison and selecting criterion (Young 1993, Meade and Islam 1997). Although so many technological forecasting models were proposed, less of them can be used to forecast extensively. The articles proposed about new or modified models explain for a particular time-series data (Young, 1993).
Although there are so many new or adapted models have been proposed in recent decades, some improvements are still can be found in these models. The main purpose of this research will try to find a new technological forecasting model according to the time-varying capacity called the generalized logistic model, and make several comparisons between the extended logistic model and Fisher-Pry model, which is most commonly used in quantitative technological forecasting. Moreover, the comparison will be based on the length of data, and the shape of data.
This thesis will try to find a proper model which capacity growth with time-varying and make a comparison between a new extended logistic model, Fisher-Pry model, and Gompertz model. The goal of this research is to identify the extended logistical model will have a better fit and forecast than Fisher-Pry model and Gompertz model, which are comprehensively used in related area. This research will use the different kind of data sets to test these three models, and data sets will be simply classified in order to identify the performance of these two models in different classification.
1.5 Research Content
The scope of thesis can be divided into several parts, and the basic outline of this thesis was shown (Figure 1). The introduction offers the overview of thesis and the outcome roughly, and talks about the purpose of this research. The literature review gives some definitions such as technology forecasting, technology substitution, the life cycle, and so on. The data collecting will describe why this research collects these data sets and then analyze the outcome of the test in these two models. After analyzing, comparing between these two models will be implemented and set the conclusion.
Figure 1. 1 The structure of thesis.
Introduction, Research purpose and
motivation
Literature review
The definitions and methods of technology forecasting
Comparison between models and Result
Data collection and Analyses
The trend of durable goods in Japan Conclusion and
Recommendation