A common problem many large retailers face is that special offers on generic items don’t reach their loyal customers. Instead, unethical grocery stores and one-time customers clear out the stock first thing in the morning. Loyal customers are left with disappointment. 

A quick solution would be to make special offers exclusive to loyal customers. However, “loyalty” is not simply defined by holding a membership card. It is defined by the value each loyal customer brings to the brand. Down the line, this is a combination of the frequency and the amount spent over a predefined period. Loyal customers can be “promoted” or “demoted” dynamically based on their purchasing patterns and entitled to different special offers. 

This process is neither simple nor well-defined, making it a good candidate for Machine Learning and AI. We asked our client to label as many loyal customers as they could as Gold, Silver and Bronze customers. The labelled data was split into training, validation and testing sets to train ML models. Right now we are comparing Deep Learning and GBM (spoiler: GBM seems to be doing better so far). We will share more as we progress further.