Store-level models are focused on specific locations where existing or planned stores are, be them direct (i.e. managed directly by our client) or affiliated (in this case our client can be both the franchisor and / or the franchisee company / candidate).

Depending on the characteristics of the analyzed store and on the size of the network it belongs to, we use two different kinds of models:

Contendible Potential Models.
they are gravitational geostatistical models and are especially suited to estimate the potential of stores who belong to small-sized networks and / or which implement a new retail format.

Building the Model
Contendible Potential Models are developed following a recursive process.

1. Determine the Catchment Areas for the Stores (either existing or planned openings) involved in the analysis. The union of all the Catchment Areas will be the territory which will be analyzed to calculate the Contendible Potential of the Stores.
   
   
2. For each census block lying within the area of analysis we will determine the probability with which the resident population will direct its purchases towards each of the surrounding Stores. As shown in the picture below, such probability ("P") will depend on two type of factors ("Forces"):
   
   
   
   
   
   • Attraction Forces which pull residents towards a particular Store. In the classical formulation of this model the typical attraction force is the Stores's size (the larger the size of the Store, the more it will be able to attract customers). Leveraging our expertise on territorial profiling we have integrated this classical approach with the residents' Affinity Index to the product / service / retail brand as a force of attraction. The higher the the Affinity Index for a Store, the stronger will be its "pull" on the resident population.
     
     
   • Attrition / Repulsion Forces which pull the residents away from a particular Store. In the classical formuation of this model the main attrition force is distance (the farther the Store, the higher the attrition force). Leveraging our expertise on territorial profiling we have integrated this classical approach, taking into account the different importance of Store proximity for different customer profiles.
     
     
   Applying the probabilities to the total category consumption in that census block, we are able to calculate the quota which will be directed towards each Store.
   
   
3. Repeating the calculation for all the census blocks within the analysis area and adding - step by step - the quotas, we will obtain the Contendible Potential for each store.

Expected Business Models.
They are regressive geostatistical models and are especially suited to estimate the potential of stores who belong to medium and large-sized networks which implement an established retail format.

Building the Model
CTB has consolidated over the years a methodology for Expected Business Models development structured in three phases:

1. Stores Clustering: The existing sales points are split into homogeneous groups, based on characteristics of the location (ie. Internal captive locations such as shopping malls, stations, airports or on the street) or macro-physical characteristics or offer ( eg. size);
   
   
2. Analysis of Variables by Cluster: For each cluster, four classes of variables are analyzed, in order to determine those that best explain the variability of turnover:
   • Specs of Store (Type management, Opening, Shop, Windows, Size, Layout, etc.);
     
     
   • Environmental and Socio-Demographic (Population, Work Population, Available Income, level of consumption of the category, etc.) and Psychographic-Behavioral (Lifestyles, affinity to the Target Company, etc.) - Learn more on micro-territorial profiling ;
     
     
   • Environmental or Structural (General Commercial Intensity and of the category, Presence of Traffic Generators and location of the competition, etc.);
     
     
   • “Captive” VAriables (GLA SM/ Shops / Shoppinh Center key traffic generator business, Number of Visitors / Passangers / Number of Vehicles per year);
     
     
3. Model Processing: Based on an analytical work, a forecasting model is developed for each cluster identified in the first phase. The technique used is usually that of multiple regression (linear or not), using a number of variables that generally varies from 8 to 15, selected from among those analyzed in the second stage in order to maximize the model predictive ability.

Model Implementation
Once built, the model can be made available to the Client in many forms:

• Performance Analysis of the Points of Sale (managed or served) vs. potential. The analysis includes the comparison - for each cluster - including POS close to the potential and those significantly far from it;
  
  
• Evaluation Report or prioritization of one or more POS opportunities in terms of potential, both gross and net of the cannibalization of other Stores. In case of network rationalization, the support is very similar: calculate the loss of business, gross and net of impact on others Stores.
  
  
• Evaluation report and prioritization of one or more distribution expansion opportunities in terms of potential, both gross and net of the cannibalization of other Stores served. In the case of distributive rationalization, the intervention is absolutely similar: calculate the loss of business, gross and net of impact on others Stores.
  
  
• Integration on the Target Planner^® Geomarketing platform , via an interface that guides the user in using the instrument and provides a few clicks in a robust and reliable estimate accompanied by customizable reporting.