In this project, Market Segment specific Promotion Campaigns were recommended to the Marketing Team of a traditional granite processor company, MG Impex. As their product i.e. granite price is based on area, Our work is projected to improve their profit margin by INR 270/sq. feet.
We leveraged Predictive Market Basket & Graph Analysis to exploit customer purchasing patterns and validate domain knowledge of Subject Matter Experts.
Our sponsor for this project, MG Impex Pvt. Ltd is a 15 years old leading processor of natural stones in India. They process granite blocks and export slabs all around the world. Currently, the utilization of data is manual and heavily relies on the knowledge of a few key executives.
This dependence is inefficient, as marketing agents must consult higher officers before taking any decision. Another drawback is that marketing decisions are often based on domain knowledge without evidence from data. This approach will not be successful in the long term. To ensure that MG Impex continue to be pioneers in their field, it is important that they use data to tackle these problems.
Tuned Apriori Algorithm mined market-segment specific association rules by intergrating customer information and context to generate interesting rules. To meet the business requirements of the sponsor, sequential recommendations were generated instead of baskets of products using product network graphs.
The Market-segment specific undirected graphs weighted by lift was generated using the mined association rules. These product network graph helped validate & visualize domain knowledge.
We proposed two approaches for graph-based profit calculation & maximization:
- Nearest Neighbours
- Greedy Approach
Due to the usage of sensitive and confidential information in this project, the anonymized data has been made available. The approach and results are discussed in detail in the report. The code however cant be made public. Here the metholody has been summarized for reference. Also, if you are working on something similar and want to discuss approach or bounce off ideas feel free to contact me.
