Customer Segmentation

Introduction

Customer segmentation and personalised marketing has become the norm in driving customer attrition and acquisition. High value customers drive over 70% of value for companies; targeted marketing enhances loyalty and customer engagement by tailoring to each contextual user experience. Moreover, BCG (2017) noted that “brands that create personalised experiences…are seeing revenue increase by 6% to 10% — two to three times faster than those that don’t”. This case study outlines the delivery of a customer segmentation model that greatly enhanced our client’s marketing understanding and strategy.

 

The Problem

The client is a fast-growing online retail marketplace that specialises in particular verticals of the Australian liquor industry. Their inventory encapsulated 400+ individual brands and 2000+ products. The client required a more efficient way to target and engage new customers in addition to retaining existing customers. A customer segmentation model was developed in aim of increasing the overall customer engagement and loyalty by personalising the experience to behavioural patterns.

 

Our Approach

Our approach for the customer segmentation model can be outlined in four stages:

– Data preprocessing and feature engineering

– Model development

– Iterative validation and selection

– Deployment

 

Data was collected from trailing three years of customer and purchasing data. Features were created and preprocessed through various interactions with domain and subject matter experts in the business. A number of candidate semi-supervised models were subsequently developed to segment customers into actionable groupings, through the AWS SageMaker platform. The model selection and validation process then ensured that the best candidate models were identified and over-specification minimised. Lastly, the final machine learning solution was deployed to the cloud as an API endpoint.

 

Results

Overall, different customer segments in the business were identified, giving the client actionable insights in terms of which people (age, gender, and location) should receive what products and promotions. The client developed a better understanding on how to target and acquire new high-value customers based on their most responsive products and promotions. Furthermore, this allowed the client to better tailor their re-marketing campaigns to increase engagement with existing high value customers – ensuring that their marketing spend was directed at customer segments with the highest value potential.

 

References

https://www.bcg.com/publications/2017/retail-marketing-sales-profiting-personalization.aspx