Primavera

A basic technique is to find new sites using criteria that match successful existing sites. This technique involves running radii or drivetime reports around existing sites to establish baseline demographic measures. Key variables such as age, income, ethnicity, traffic and daytime population are determined by retailers and commercial real estate analysts. As new sites are considered, reports are created using these key variables as well as other data and compared to those of the existing site. Many commercial real estate companies include these reports as part of their client presentations.

Companies can also use "scoring" models to rate potential sites. These models correlate the relationship between the sales revenues of existing sites and the demographic, business and other data variables in market areas. It is critical that sites in the sample very closely match the existing site. Results of these analyses usually yield between six and 10 report variables with either a highly pos itive or highly negative correlation to the sales revenues of the company's existing locations. These variables and their significant ranges are used as a "scorecard" to rank each potential site. Scorecards can also be used for risk mitigation to help avoid potentially unprofitable sites; however, they do not quantify or identify potential sales by site.

When inserting a new site near existing sites, the selection process becomes more complex. The presence of competitors and the cannibalization of existin g sites must be considered. Gravity models can be used to address these issues. Gravity models operate on the probability of customers visiting and purchasing from a particular site based on distance to the site, attractiveness of the site, and distance and attractiveness of competing sites. Gravity models embedded in a geographic information system (GIS) enable analysts to perform "what if" hypotheses, quantify the market share extent by site and provide a vivid picture by mapping the impact of a new site location vis-à-vis existing sites.

Store sales can be predicted by using multiple regression analyses as part of the site selection process. This technique seeks to identify how a site's variables interact to produce a specific sales outcome. The success of this technique depends upon the user's ability to accurately quantify factors such as operations, sites and locations that might affect sales. The difficulty is quantifying subjective variables such as customer service, merchandise mix, landscaping and signage. However, when correctly applied, this technique can accurately predict — within a certain margin of error — the expected revenue of a potential site. Multiple regression models can be expensive, time-consuming to create and compromised by the introduction of new factors. Time and/or change, plus the difficulty by many in an organization to understand the complexities of multiple regression modeling are negative factors for these types of models.

The techniques described above do not include customer information. The best tool to find successful sites is to profile customers who shop at existing profitable sites. Because of the cash-and-carry nature of retail or the lack of systems to collect customer data, most retailers have little customer information. When customer data is collected — from surveys, delivery records, credit card transactions, and other means — customer profiling can be performed. Segmentation data and geocoding are marketing applications that can accurately profile customers.

Segmentation systems work on the theory that people with similar tastes, lifestyles, and behaviors seek others with the same tastes — "like seeks like." These behaviors can be measured, predicted, and targeted. Geocoding is the use of software to append ge ographic codes such as latitude and longitude and segmentation data codes to customer records. Because segmentation systems can drill down to various levels of geography, such as ZIP Code, census tract, block group and ZIP+4 Code, analysts can easily scan markets and rank prospective sites based on the number of desired segments that match the key segment profiles. Typically, six to 10 segments will contain most of a company's customer types. Companies can use these tools to better understand who their best customers are, along with what they buy, how to reach them and where to find more like them.

Site selection is a trial-and-error process for all-sized businesses; what works well for one company might be less successful for another. Companies have conceptual and cultural differences and varying levels of sophistication. To work well, site selection processes must be thoroughly understood and trusted by an organization's analysts and managers and fit the organization's overall strategy.

­ Dan Primavera is retail industry manager with ESRI's La Jolla, California, office.