Logo Dataclair

CUST2VEC Deep neural networks for Customer Lifecycle Management

mobile 1

About CLM

A "customer lifecycle" is a journey of various stages a customer goes through, which perhaps begins when a potential customer decides to make a purchase. There are five typical actions (they differ slightly, depending on the source) we can take, to engage the customer life cycle. Specifically: reach, acquisition, conversion, retention, and loyalty (Sterne and Cutle, 2000).

Each company has a chance to control and guide the customer’s journey and thus understanding your customers is one of the most crucial steps when you want to change from product-centric to customer cycle-centric strategies.

Why should we care about CLM

Moreover, there are reasons why to consider CLM as an essential ingredient of your business. At first, with constrained customer acquisition channels, the customer acquisition costs will proceed to rise unabated. Positive information about happy customers is spread by bringing additional customers and thus results in more purchases. Last, but not least, it is easier to convert and retain an existing customer than to acquire a new one.

Our Approach

Improvement in Machine Learning (ML) and Big Data technologies provide us the possibility to shift from conventional linear regression to deep learning (DL) methods. The main reason is that the customer journey is not linear in time and DL techniques have recently shown incredible strength to produce significant, if not state of the art, results on multiple kinds of problems, ranging from image recognition to text translation. By DL approach we build a model, which will identify the customer stage once it is already acquired. Consequently, we represent each customer as a vector of events in time, using history and record actual behavior to improve the next prediction. Before we had a first outcome, we faced up numerous challenges like class imbalance and representation of the ML problem. Finding up the cost function allows us to reduce customer's attrition, increase lifetime value while reflecting maximum benefit from our services.

Credits

Jan Romportl

JAN ROMPORTL Director

show more
Michal Pleva

MICHAL PLEVA Data Science Team Lead

show more
NEXT CASE STUDY CELL2VEC BTS EMBEDDINGS FOR ADVANCED NETWORK DIAGNOSTICS DISCOVER MORE mobile 1