Optimizing Geofencing for Location-Based Services: A New Application of Spatial Marketing
Since the beginning of modern retailing, retailers have used geographic information to determine the trading area around their stores (Christensen and Tedlow 2000). Subsequently elaborated spatial marketing techniques have emerged and sophisticated models
- PDF / 174,315 Bytes
- 4 Pages / 595.276 x 790.866 pts Page_size
- 90 Downloads / 191 Views
o be considered. The traditional methods and models used in spatial marketing to calculate trade areas may be applicable to geofencing. APPLYING SPATIAL MARKETING METHODS AND MODELS To date customizing the shape and size of a virtual fence has not been a mainstream option for retailers. Defining the geofence perimeter is a complex task that offers constancy and a certain similarity to the definition of a traditional trade area around a retail store. The purpose of this study is to review the existing spatial marketing models and to determine their potential application in geofencing. Applebaum (1966) delineated trade areas according to the proportion of clientele through customer spotting and spacedistance approach and proposed to divide the trade area into three sub-areas forming circles around the shopping area: primary, secondary and tertiary (or marginal) according to the proportion of customers attracted by the shopping area. Today GIS (Geographical Information System) software is able to draw trade areas from customer addresses obtained by for instance loyalty cards and show that most of the time these trade areas have nothing to do with circles. Early models (Reilly 1931; Huff 1964) were built for prediction purposes mostly around the notions of distance and store size where newer models (Gautschi 1981) incorporated additional parameters such as store image (Lindquist 1974; Nevin and Houston 1980), transportation mode (Brand 1973; Sherret and Wallace 1973; Domencich and McFadden 1975), or store perception (Cliquet, 1995; Nakanishi and Cooper 1974). It has been used for grocery stores (Popkowski, Sinha and Sahgal 2004), convenience stores (Achabal, Gorr and Mahajan 1982), furniture store (Cliquet 1995; Huff 1964) and so on. These models are difficult to implement according to categories of products and specific features of these models. Attraction models used in retailing can be either strictly gravity models (Reilly 1931; Huff 1964) or both gravity and market-share models like the MCI model (Cooper and Nakanishi 1988; Nakanishi and Cooper 1974). The MCI model took over because it is more flexible according to the category of products: some categories and the specialized stores which sell them imply a continuous attraction whereas others (e.g. furniture stores, Cliquet 1995) needs to consider thresholds (Malhotra 1983). For mathematical reasons (transformation of the MCI model into a regression model), ratio scales should be used for data (Gautshi 1981) and when interval scaled data are only available, a zeta squared transformation has been proposed (Cooper and Nakanishi, 1983) to get ratio scaled data. These models need to divide the market area into geographic cells and the choice of this division can alter the results. To control the possible nonstationarity of the regression parameters, a typology of cells (Cliquet 1995; Ghosh 1984) can be used to overcome this difficulty. Using a regression model, a too limited number of shopping areas considered in the study can disturb the estimation of the regression co
Data Loading...
