Assortment planning for retail
based on AI
The solution for the formation of the ideal assortment in the network of electronics stores
Customers don't just buy products, they fill certain needs with purchases. This can be considered to improve efficiency.
Space on a shelf, page or warehouse is a valuable resource and must be used optimally. On the other hand, it may be that there are several similar products on the shelf and although they cover part of the needs of customers, the needs of some customers remain uncovered and this is a lost profit.
The main tool for identifying customer needs is building a tree of customer preferences.
Our solution analyzes customer needs based on website sessions and/or their historical purchases and aggregates all products by need. A customer decision tree (CDT) is gradually built, which is then passed to the category manager for correction, who can add his vision to the markup.
Further, within the framework of each need, taking into account the size of the shelf, premium goods and stores, the product that is most optimal for the company in terms of a combination of revenue, margin and checks is recommended on the shelf. In the process, cross-cannibalization of the goods is also taken into account. A tool has been developed for the category manager that allows him to correct the recommendations of the machine.

Customer Deсision Tree (CDT) shows which product attributes are important to the customer and in what order they should be placed.
In most cases, the CDT is based on product attributes. The lower the level of CDT, the stronger the goods are substitutes for each other. Category managers can be of great help in building a CDT as they have a good understanding of their categories. There are analytical ways to build a CDT, such as transaction graph analysis. To identify the needs, an expert interpretation of the constructed tree is carried out.

On the example of smartphones, new features of goods that are not directly used when filling the assortment matrix have been identified.
The matrix optimizer will be able to recommend the best set of SKUs (stock items) to represent a category in a store.
Matrix optimizer options for selecting the best set of SKUs
for category presentation in stores
  • Input data in stores
    • Store clustering by geography
    • Store clustering by premium segment
    • Number of available seating spaces per category
  • Input data by SKU
    • Similarities by SKU from CDT
    • Receipt data by SKU
  • Manual input
    • Choice of geography-permiality cluster
    • Filters for the period of sales, promo sales
Economical effect:
  • 1
    Building a customer decision tree (CDT) greatly simplifies the work of a category manager, because on the basis of the already created markup and clustering, he can select individual groups of goods and form an assortment based on the actual needs of customers.
  • 2
    Real-data-based analytics helps you make decisions faster, as well as avoiding errors in the interpretation of indirect customer data and getting rid of intermittent and distorted metrics obtained in surveys.
  • 3
    For large retailers, the economic effect is measured in millions of dollars per year.
We already have a solution
The time for adapting the solution to business is two to four months
Integrating with your
IT systems
Building customer
decision tree (CDT)
Identifying Needs
Formation of a product matrix
We have been developing advanced solutions based on neural networks since 2016
Economic effect of cooperation with us -
tens of thousands of dollars a day

Contact us
+7 (967) 215-75-05