Our client, one of the world’s largest food and beverage companies, wanted to build a retailer recommendation engine which looks at the historical order placement data and suggests products for the next month.

The objective is to provide retailer-level recommendations to the company by unlocking key patterns in the historical purchase order at channel x retailer x region/district level.


Data Pre-Processing and Feature Engineering

  • Optimization for unique hierarchy ID assignment in combined databases.
  • Data cleaning and validation.
  • Focusing on the top 90% retailers and products based on revenue.
  • User-item matrixas implicit feedback derived from transactional data.

Features based clustering.

  • Retailer segmentation with comparable purchasing behaviors based on their sales channels.
  • Algorithms used such as RFMand KNN.
  • Segmentation based on revenue such as:
    • Premium
    • Medium
    • Occasiona

Recommendation Engine

  • Implemented product-based recommendations for Retailers through Collaborative Filtering techniques via ML Models like Neural Collaborative Filtering.
  • Applied Association Rules to find popular items and co-purchases via Apriori Algorithm.

User Interface

  • Updated SQL table with tailored recommendations post model training for consumption over web and mobile app for Distributors and Retailers.


  • Projected improvement in revenue by ~3%
  • brand consumption improvement from 4 to 6
  • Strategically upselling new product variations by increasing product awareness through recommendations.
  • Improve Precision: & accuracy by 80%
Learn more about TransOrg’s value proposition, solution methodology and implementation approach