Retail analytics and site selection methods have come a long way since I started consulting almost 25 years ago.
           Back then, car wash developers and investors relied on equipment dealers to help define trade areas and make decisions about where to locate a new store. Site selection procedures included a combination of experience, intuition, and predictive models, including checklists and analog-based models.
           Such models are static. In other words, they can’t adjust for changes in business performance or account for things like inflation. Consequently, site selection was mainly a subjective process and relied on the experience and skill of the local equipment dealer.
           Today, AI- and machine-learning-powered software provides a more comprehensive approach to retail analytics and site selection.
           The cause for this shift is the simple fact that consumers are now easily tracked. Digital tracking records consumer behavior through permission-based tools such as cookies, device IDs, IP mapping, and geolocation.
           A cookie is a piece of data from a website stored within a web browser that the website can retrieve later. Cookies are used to inform the server that users have returned to a particular website.
Device ID identifies a user device like a smartphone or tablet based on an identifier, which is a unique string of letters and numbers.
           IP mapping is the process of identifying the IP address of computers based on their geographical position to define the exact location of the device connected to the Internet.
Geo-location tracks a mobile device’s whereabouts using GPS, cell phone towers, WiFi access points, or a combination of these tactics.
           Consequently, the data landscape has expanded from census data (demographics) that explain who people are to psychographics that explain family lifestyle to geo-social that explain what people do.
To help define trade areas and develop forecasts, analysts have taken this data and divided it into two camps: proximity and personal data. Personal data tells who lives in an area, whereas proximity data tells what happens there.
           For example, proximity data tells the percentage of the population that drinks coffee, whereas personal data enumerates how many households in the area have the lifestyle characteristics of coffee drinkers. Proximity data comes from geo-tagged social media posts whereas personal data comes from mobile visitation. Proximity data is most useful for capturing changes over time and activity volume whereas personal is most useful for predicting sales.
           Segmentation methods include leveraging psychographics to target family lifestyles, internal customer records, or sales correlation (statistical approach).
           A sales forecast — such as the number of households with an affinity for a brand, product, or service — is produced by combining proximity and personal data.
           Segmentation and sales forecasting are the black boxes of methodology. A black box refers to a system where the inputs and outputs can be seen, but its internal workings remain hidden.
Despite this lack of transparency, experts say AI and machine learning can improve forecast accuracy and site selection decisions since the technology can handle mountains of data and complicated calculations at lightning speed.
           Arguably, this is the case if someone plans to build six or seven new stores in an area. After all, this magnitude of stores requires evaluating a much greater number of candidate properties. This isn’t easy unless you have the right tools and know how to use them.
           However, to say AI will necessarily lead to better forecasts and site selection decisions seems to fly in the face of experience. For example, the car wash industry is well known for having an economic failure rate far below the national average.
           Arguably, the principal reason for this success is the industry has experienced professionals such as equipment dealers and consultants who excel at evaluating markets and sites and determining if they are suitable for a car wash.

Bob Roman is a car wash consultant and can be reached at bobr427@protonmail.com.