VISUALIZING RETAIL HOT SPOTS IN CANADA

By Anita Zelic and Paul Dunphy, CSCA

Competition plays a major role in shaping the contemporary retail environment. The competitive landscape reflects both the location and distribution of stores and the resulting concentrations of retail sales. In order to (i) observe the dynamics of the competitive environment and (ii) identify the development of retail hot spots, researchers can now examine changes in annual retail sales at the postal code level. Numerous factors can be used to interpret temporal variations in sales. These can include shifts in local demography, change in personal income levels, or new commercial developments. We have observed that the greatest increases in retail sales in Canada over the 1989-1998 period are located within selected 'hot spots'. This research letter, using data visualization and associated data mining techniques, links the growth in retail sales to the locational tendencies of the major retail formats in Canada and provides insights into the spatial characteristics of the new retail economy.

The main objective of this research letter is to illustrate, using data visualiztion techniques, the spatial distribution of the retail 'hot spots' and to examine the relationship among the main retail formats with the highest retail sales across Canada. The basic spatial unit used in this analysis is the Forward Sortation Area (i.e. the FSA, or 3-digit postal code area). The time frame covers the period between 1989 and 1998. This work was undertaken using the data mining facilities that are available at the Centre for the Study of Commercial Activity (CSCA). The analysis was performed using Mineset and MapInfo software packages and Silicon Graphics hardware that was acquired by the CSCA through the support of the Canada Foundation for Innovation and the Ontario Innovation Trust.

MAJOR RETAIL FORMATS

Retail chains, shopping malls and power centres collectively comprise the major components of retail activity. For this analysis, a retail chain is defined as a commercial organization that operates with four or more stores. These chains adopt a variety of spatial strategies to deal with the competition. These can include: avoidance, direct competition, competitive clustering, or some combination of these. More than one hundred of the most prominent retail chains with the highest number of stores were chosen for this analysis. Priority has been given to chains that are widespread across Canada and those that cover different commercial sectors such as food, clothing, hardware, jewellry and pharmaceuticals.

Shopping malls, which vary in size and mix of retail tenants, are the dominant nodes for retail distribution in the country. Their size ranges between 50,000 and 3.8 million square feet and is closely related to the type of business existing in a particular mall. The larger regional shopping malls (>700 000 sq feet) and power centres were selected for this analysis.

Power centres are unenclosed commercial formats that typically consist of two or more free standing big box stores. Previous studies have highlighted the rapid growth of this format over the last decade, and the emergence of "power nodes", that is the clustering of power centres.

The rest of this research note provides selected illustrations of how data visualization and associated data mining techniques can be used to examine recent changes in the Canadian retail economy. By fusing various data sets, it is believed that more informed decisions could be made. Mineset, a software package that allows animated 3D mapping of temporal data, was used to facilitate the analysis. The software was used to examine the changes in annual retail sales at the FSA level on an annual basis between 1989 and 1998.

Case 1: Temporal Changes in the 'hot spots'

Figure 1 illustrates temporal changes in retail sales between 1992 and 1998 for the Montreal CMA. Two snapshots of the resulting animation are shown for the same geographical unit, and differences in the level of retail activity in the FSAs can be clearly identified over time. Clearly, these patterns raise many questions from both a retail and property development perspective: What factors are associated with the growth in sales? How effectively are our properties or stores positioned? Where is our major competition? Are our stores located in high-performing FSAs?

Figure 1. An illustration of temporal changes in total retail sales in the Montreal area; left, retail sales in Montreal in 1992; right, retail sales for the same area in 1998. Random colours are used to differentiate FSAs, and their height represents retail sales in the FSA.

Case 2: Data Fusion

In addition to visualizing numerical data such as retail sales, other databases can be incorporated into Mineset and accessed simultaneously. Figure 2 provides an example of such "data fusion", showing the additional information on power centres, shopping centres and major chains that can be retrieved during the animation process. This allows analysts to see which, or how many, retail formats are present in an FSA, while temporal changes are being observed dynamically. This analysis could also be performed on different levels of geography, such as enumeration areas or census tracts, depending on the interests of the researcher and the availability of data.

Figure 2. An example of fusion of the total retail sales data with CSCA's power centre, shopping centre and retail chains databases.

Case 3: Spatial Distribution of the 'hot spots' and location of the new retail
formats

In addition to the 3D maps produced by Mineset, this study has also employed the 2D (i.e. conventional GIS) environment of MapInfo. The statistical approach of "natural breaks" was applied to the retail sales data - FSAs within 33% of the highest retail sales in 1998 were selected. Two areas of interest in the Toronto and Vancouver regions are shown and compared.

Although there are indications that the role of shopping malls as the dominant retail destination is under threat, it is evident from Figures 3 and 4 that they continue to play a major role in generating retail sales in the Toronto and Vancouver regions, respectively. Both major shopping malls and retail chains are positively related to retail sales within both regions. In addition, Figure 5 suggests that there is a stronger positive correlation between the number of chains and retail sales than between the number of tenants in power centres and retail sales. The Toronto region has more power centres than the Vancouver region, and many of them are located within FSAs that have more than a few chains and shopping malls. Compared with the shopping malls, the power centres are expanding into less profitable FSAs, which are mostly new residential or grayfield areas. It is apparent that power centres have become most rapidly engaged in retail activity during the past decade (Figure 6); this trend will be of interest in future studies, since it may shift retail sales into a different spatial distribution.

Figure 3. Relationship between retail sales and different retail facilities in Toronto

Figure 4. Relationship between retail sales and different retail facilities in Vancouver

Figure 5.Relationship between retail sales and a) number of chains b) number of power centres

Figure 6. Growth of Big-Boxes in the Greater Toronto Area during the past decade

Case 4: The Relationship between Wal-Mart and Retail Sales Growth

In Figures 7 and 8, the location of Wal-Mart is compared to both the overall growth in retail sales and the locational decisions of major retail chains. The maps present these two relationships for the Toronto region. The positive relationship is evident between both the tendency for national chains to seek out locations in close proximity to a Wal-Mart and the trend of Wal-Mart locations to be associated with areas of increased retail sales. The map patterns indicate that the number of stores either increases or remains constant within FSAs that have positive growth and high retail sales. In addition, the location of Wal-Mart outlets typically does not coincide with locations of shopping malls or power centres in the Toronto area.

Figure 7. General trend on major chains and Wal-Mart in Toronto

Figure 8. Relationship between retail sales, Wal-Mart and retail facilities

Case 5: Spatial Distribution of Home Depot and Revy

The final illustration examines the distribution of two major competing home improvement retail chains. Figures 9 and 10 indicate the locations of Home Depot and Revy stores in the Toronto and Vancouver regions. Home Depot dominates the Toronto region, while more Revy stores could be found in the Vancouver region. In most cases, both chains are found in FSAs that have high retail sales - hence dominant retail nodes. Further, the majority of these home improvement big-box centres are not located in FSAs that contain shopping malls.

Figure 9. Relationship of retail sales and spatial distribution of Home Depot and Revy in Toronto

Figure 10. Relationship of retail sales and spatial distribution of Home Depot and Revy in Vancouver

CONCLUSION

This research letter illustrates how data mining and data visualization techniques can be used to examine changes in the retail system. The research could be expanded to include other areas in Canada or the United States, and the techniques used to explore relationships between various retail formats. For instance, one could visualize the relationship between the locations of various commercial structures and retail sales. Similarly, factors that relate to demographics, income and ethnicity can be easily incorporated into the analysis in order to gain a better understanding of the spatial distribution of 'hot spots', and to predict the direction retail may take in the future. In this context, the trend of the power centres may be of particular interest due to its spatial deviation from other retail formats.

Finally, it should be noted that the cases presented here were by necessity static; however, the system that produced these images facilitates temporal animation. With a longitudinal database such as SARTRE or those housed at the CSCA, temporal changes can be presented in map form. We believe that this type of dynamic, spatial modeling is one of the future directions that geomatics research and associated GIS technologies will take. If you would like a prototype presentation of such a system, contact the CSCA for a demonstration or visit the CSCA's visualization web site at www.businessgeomatics.com.