Drive through most major towns and you are likely to see a phenomenon known as “agglomeration,” where retailers congregate in the same general trade areas. Lowes and Home Depot, Walgreens and CVS, and of course, McDonald’s and Burger King all seem to have their brick-and-mortar establishments in proximity to one another. The rationale is that each competitor doesn’t want to let the other have a meaningful location advantage. However, the days of using the “golden arches” site selection strategy of finding solid, profitable locations in sight of Micky D’s for your new hamburger joint, or anything else, is obsolete. Almost.
The best brick-and-mortar retail site locations are a finite and scarce resource. Today, selecting “where” is the domain of the geospatial data scientists that have an abundance of information available to them. However, because the access to digital site selection data is rich and vast, the competition is brutal.
The definition of “perfect competition” considers that each competitor has access to exactly the same data, that is “perfect information.” As such, it is likely that the same property becomes the optimal choice in the model scenario of every likely buyer. In short, the best sites may already be taken and the hard work of selecting a less ideal location begins.
Retail Site Selection Criteria
Competitors, even within the same industry, such as quick service restaurants (QSR), have different site selection criteria when choosing a retail store location, whether to lease or buy. Even the variables that are required to decide whether to extend an existing lease are different. The question becomes, which variables, how current, and to what geographic extents provide the most exacting information about the revenue potential and profitability of any given site.
Whether you are using five or fifty variables to evaluate a given retail space, benchmarking the set of variables and using spatial statistics, such as spatial regression analysis, will help determine a correlation of significant variables. This is a necessary step as there are literally thousands of location-based data products that could support site selection analysis. Likewise, there are a number of location-based data suppliers that must be considered, including those offering demographics, mobility, traffic flow, weather, property, and many other data types.
Pandemic Impacts on Site Selection
Even before the pandemic, the challenge of selecting profitable sites became complicated by the balance retailers needed to strike between brick-and-mortar and e-commerce, and the option of buying online and pick up in store (BOPIS). Today, even good sites must be re-evaluated, and the criteria challenged based on whether the site fits into a different model.
For example, regional malls that were already changing to a mixed-use of office, residential, and retail, might also now include warehousing and fulfillment center operations due to an increase in e-commerce activity that was 21% of total retail sales in 2020, according to Digital Commerce 360.
Likewise, some retailers are opting for a change to their store formats, such as Starbucks that is experimenting with both curbside pickup and drive-through-only locations. Chipotle Mexican Grill was already undergoing a transformation to add drive-through lanes and now counts over 100 “Chipotlanes” that reach a clientele still hesitant to rejoin the inside dining experience.
The Retail Site Selection Process
Let’s look at how to prepare the retail site selection process.
Consumer Data – “Retail Follows Rooftops”
One of the standard baseline criteria for retail site selection by real estate brokers is to look for new housing developments in what is referred to as “retail follows rooftops.” But where do you get “rooftop” data? New housing data is the domain of local tax assessors, but these data are being incorporated on a regular basis with address validation and geocoding software like those available from Precisely, Loqate and others.
These data can be geoenriched with property information such as assessed value, number of rooms, roof composition, etc. This can provide a picture of neighborhood economic conditions and whether the area’s dwelling type is composed of single-family, multi-family, condos, gated communities, or other types of dwellings. Core consumer data will augment home location data with standard demographics such as age, income, education, and ethnicity but also retail expenditures by category, credit card usage, disposable income, consumer vitality, credit worthiness, and social affinity.
Geodemographic data infers the predominant lifestyle characteristics of neighborhoods. Precisely’s CAMEO for the United States, and Environics Analytics’ PRIZM for Canada are examples of segmentation solutions that identify the socio-economic profiles. Geodemographic segmentation data describes the lifestyle characteristics of households by identifying family social status, characteristics of children living at home, discretionary income spending behavior, financial status, hobbies, and interests. It helps the commercial real estate site selection team, as well as the marketing manager, to better understand the consumer’s lifestyle choices and the geographic similarities of individual cohorts.
And, if “retail follows rooftops” then there will be competitors lurking close by. Therefore, a robust data set of business locations, business classification, and zoning data will be required. This layer provides an added view of business activity. The retail site analysis should include annual average daily traffic (AADT) to reveal a picture of business health, but more importantly, detailed historical traffic (DHT). Even including weather data is a predictor of seasonality for stocking certain products and predicting the length of the buying season for items from snow cones to snow shovels.
Lastly, footfall traffic delivers a hyperlocal view of the propensity and frequency to visit local store brands. Foursquare, for example, delivers a set of insights that help retailers on both selecting sites as well as evaluating existing stores for continued operation. Foursquare provides:
- Venue-level granularity to analyze total visits to specific stores, brands, and other points of interest.
- Foot traffic data to specific chains, categories, and markets.
- Normalized data to mitigate demographic bias.
- Weekly and monthly indices that are normalized for visit fluctuations.
These data provide a more real-time demographic profile of consumer behavior since they are derived from first-party sources, and would thus conform to regulations such as GDPR and CCPA privacy guidelines.
Using Retail Site Selection Models for Location Analytics
In 1931, William J. Reilly proposed his which assumes that buyers will travel longer distances to larger shopping centers. Conversely, the further the travel distance, the less likely a buyer is to travel, and hence convenience becomes a key factor. John Quincy Stewart, in 1947, proposed “demographic gravitation”, suggesting that large numbers of people can be an attraction, a model relevant to both cities and regional malls. Dr. David Huff proposed in 1961 that there are other “attractiveness” factors influencing retail sites such as the number of stock-keeping units (SKUs) available for purchase, the number of parking spaces, and even the amount of lighting in parking lots.
Together, these data are model inputs to spatial regression analysis that would identify in better statistical relevance whether geography, demographics, or another data element is more significant in understanding site viability. Most geographic information system software today can execute distance decay, spatial regression as well as the Huff model.
The Next Best Location – Store Analogs and Cannibalization
Franchise development dictates that retail site selection is not just about a single location but the ability to optimize an entire network of highly performant stores. Model development, as suggested above, supports the identification of specific criteria that can predict sales revenue. However, the use of a store analog also provides validation that the model works at an existing location.
As Joe Rando points out in an article titled Why I like Analog Models, “Analog models use the trade area and market characteristics of a given site to compare against your existing stores and find the stores that most closely match the site in question. The output of the analog model is a list of your existing stores that most closely resemble the site in question along with a score of how closely they match.” While Rando suggests using the model for an existing store, the same methodology and data inputs could be applied to a potential new site without existing sales data.
Taken together, both a predictive model and a store analog provide the confidence that the network can be optimized for maximum revenue performance and to justify to franchisees that adding a new location will not cannibalize sales at existing stores.
Combining Software Tools and Data for a Competitive Advantage
While perfect markets and, consequently, perfect information, do not exist, data is more available than a decade ago from a variety of geospatial data marketplaces. In addition, software solutions that perform location analytics can integrate more disparate data types and implement a variety of retail models. So, both software tools and data have become a competitive advantage to those that invest the time and resources to develop location intelligence. Those without the existing internal resources can contact Korem today!