Developing a store attributes predictive model is one of the most important initiatives a growing chain can implement.  With a finite number of resources and investment capital, companies need to ensure that their efforts provide the greatest return and without a predictive model, how else would you determine the success or failure of future capital investments?

A store attributes predictive model gathers all the aspects of a particular store – i.e., size, demographics, product mix, location, etc. – and attempts develop patterns to their overall financial performance.  By identifying correlations that can be made based on these attributes contribute, the organization can better predict future results of new locations.

So, for multi-unit operators, slicing and dicing their store portfolios into store demographic geo-types is the first step.  This way they can bucket like-stores together with similar attributes to gauge whether the financial performance of those locations mimic one another.  If they do, those store attributes may provide a predictive model for future locations.

Here is an approach to developing a predictive model:

Analyze Existing Locations:  Group your stores together based on similar attributes.  For instance, you may consider Store Demographic Geo-Types:

• Suburban
• Rural
• Urban Neighborhood
• Urban Business District
• Commuter and Turnpikes
• Truckstop

Next, you may want to group those stores into various Operational Store Types (for example):

• Basic (smaller, older locations)
• Classic (midsize with typical c-store product assortments)
• Premier (higher-end products, larger locations)
• Premier Plus (higher-end products, large locations with a strong foodservice component)

Competitive and Demographic Analysis:  While store types and operational configurations can give you one slice of the equation, overlaying competitive and demographic information provides a better representation of store performance.  For instance, here are five (5) potential examples on how a stores performance can be impacted by outside forces:

• Store scenario 1: Store surrounded by 25 competitors vs. 5 competitors?
• Store scenario 2: Store that is a 65% Hispanic demographic, how do we alter merchandising mix?
• Store scenario 3: Store with six (6) QSR’s nearby?
• Store scenario 4: Store has site access issues
• Store Scenario 5: Suburb location where people leave to go to work as opposed to suburb site where people come to work

Each of these scenarios offers an influential weighting factor to the overall predictive model of store performance.

Develop a "Predictive" Model:  This model should be applied to existing (for past performance validation and vetting) and future stores and sites, which will help determine the following:

• Financial performance metrics
• Merchandising mix
• Store size and offering
• Dispenser allotment
• Store budget (new builds and remodels)
• Decision point on whether to remodel or rebrand, to purchase or lease a site
• Decision point on whether to do an acquisition
• Align potential new sites with “like” sites already performing at the company with similar attributes

The more attributes your team can measure, the tighter your predictability for future store performance will be.

Put the Predictive Process in Motion:  Once you have identified key metrics for evaluation, consider the following:

• Traffic generators (i.e. schools, factories, etc.)
• Functional constraints (i.e. natural barriers such as rivers, etc.)
• Competitive retail (both c-store and foodservice)
• Geography (i.e., urban residential, suburban, etc.)
• User type (blue collar worker, professional worker, etc.)

Develop a Scoring System:  Next, begin to assess your locations with a scoring system that includes a summary page and background data.  The scoring system should be dovetailed with existing financials – is there a correlation with scoring?  In addition, third-party company data points can be layered over your data to help validate your decision matrix.  A good testing ground is to apply your decision matrix scoring system against existing sites to see if your matrix hypothesis holds up across your network. 

Next steps and Recommendations:  Based on the results, tweak your model to become a better, more accurate predictor in the future. Obtain input on adjustments and additions to scorecard from key resources within the organization to not only create a more accurate predictor model but also to engage the organization in a deeper approach to site selection.  Key resources to tap into are as follows:

• Operations
• Merchandising
• Facilities
• Real Estate
• Finance
• Cross-reference third-party data

In summary, using a store attributes predictive model can validate data from existing stores to help the company become a better projector of results for future locations.  It is through this understanding of what makes one store perform better than others to validate a predictor scorecard on existing sites and applying to future sites.   With capital spending at such a premium, it’s best to give yourself a fighting chance for success utilizing the rigors of predictive modeling.

Want more ideas?  For more information on Gray Cat Learning Series, visit: https://www.graycatenterprises.com/gray-cat-learning-series