Consumer preference has been a core concept to explain customer choice behavior. There are two schools of thought regarding the preference (Hoeffler and Ariely, 1999). The Economic school is based on the assumption that consumers’ choices signal their underlying need and wants, since preferences are revealed when consumers choose. Each consumer has stable and coherent preferences and maximizes those preferences via choice behavior (Rabin, 1998). Preference in Economics is considered as utility, and such choice behavior is called rational utility-maximization choice. On the other hand, Psychology believes that consumer choice looks more like irrational and is affected by status quo or reference products in his mind. That means each choice behavior is governed by preference that is constructed on the spot. The constructive processing approach in Psychology assumes preferences are constructed on the basis of the task and context factors during choice or evaluation. Ever since Huber, Payne,and Puto (1982) have shown the attraction effect, the effect has been one of the main topics in consumer decision research. By adding an asymmetrically dominated alternative to a choice set, they have demonstrated that the newly added alternative helps the alternative that was the most similar, thereby violating the similarity hypothesis and regularity as well (Huber and Puto, 1983). No one can say which school is correct or not in simple and definite terms. The reality might be in the middle, where consumers construct their preferences in a new category and stay there in a consistent way with more experience. Marketing managers should be able to understand not only the origin and formation of preference, but also how to utilize it to make successful performances. Key components of successful marketing strategy include prediction or estimation of sales and optimal design of product characteristics. In order to estimate sales of new products, marketing models usually employ models based on either statistical techniques or micro economic theory. Recently, there was a growing interest in structural models which is based on sound theory, not just forecasting formula (Chintagunta et al., 2006). Existing BDT research builds theory with only a small number of alternatives. However,quantitative choice modelers need to test behavioral theories using secondary data which is collected in more realistic choice environments (Kivetz et al., 2008). The proposed model collects customer choice behavior data from three alternative sets and estimates a choice response surface from the choice behavior data, which is a smoothed preference surface function from empirical choice distribution on joint product attributes. The purpose of the paper is to investigate how preference structure changes due to increment of number of competing products. It can be done by comparing each choice probabilities of space where different number of product exists. Next, the paper tries to find out any context effect by changing relative position of products nearby. Finally, there are two cases of introducing additional product to the existing choice space. Introducing the second product where only one product exists, and introducing the third product where there are already two existing products. By comparing these two cases, we can find out whether the effect of introducing additional product is different depending on number of existing products, and whether it comes from local configuration or from global configuration. The proposed model consists of three parts: 1) collecting data for preference surface with only one alternative, and with two alternatives, and three alternatives, 2) obtaining the preference surface using smoothing technique, multiple local linear regression and 3) applying fisher exact test to find out whether choice space is different from each other. We collect data from quasi experiments. Hypothetical product, antibacterial towel for display monitor of computer or cell phone is selected. Each product has two attributes, duration time as antibacterial status and number of effective wiping, and the values of two attributes represent one of 78 product locations on the region. So subject chooses his best product or evaluates purchase probabilities with one or two or three alternative products. With the data on 78 locations within region, choice probabilities on entire region are estimated by smoothing method, multiple local liner regression. And comparison between choice probabilities of only-one-product situation, two-product situation, and three-product situation, uses fisher exact test between odd ratios (Agresti 2002). The smoothed choice space and test results are shown in two dimensional graphs. The results show that choice probabilities are significantly changed as either number of competing products or location of products varies. When there are two products, it will get better if two products are quite different. And the context effect near preferred product is stronger than near non-preferred product. Compromise effect exists but it is shown in a little bit different way as suggested in the literature. Marketing managers can develop a new product and positioning strategy based on the results. Marketing modelers also use the curvature of choice probability on the region to develop a proper function or apply raw data directly in their model construction.