Grupo Lanit – Customer churn prediction
Grupo Lanit
Sector: Mobility
Business Case
The company needs to anticipate behaviours likely to be abandoned by its customers.
Objectives
Anticipate the risk of customer leakage or loss and take preventive measures to retain customers, such as offering them personalised offerings Optimise the resources and budget allocated to marketing and sales actions by focusing on the most profitable customers with the greatest growth potential Improve customer satisfaction and loyalty by a better understanding of customer needs, preferences, purchasing behaviour and expectations
Use case
Supervised model that, in its automatic monthly execution, identifies the relationship of customer behaviours likely to significantly decrease their visits to stores and regular purchases in the company in the next 3 months, so that early action can be taken at an individual (customer) level, taking the appropriate preventive measures.
Infrastructure
On Premise
Technology
Big Data, Machine Learning
Data
Public and Private Datasets Algorithms XGBoost + Bagging + SVR
Resources
Multidisciplinary team with experts in business, massive data processing architectures, data analysts and data scientists.
Difficulties and learning
The greatest difficulty of this project was the alignment with the business when it came to establishing the criteria for defining the ‘Loyal Customer’ variable. Not every customer who has a loyalty card can be considered a loyal customer so it proved necessary to carry out an extensive study of the behaviour of millions of customers in order to identify the right keys in business terms.
KPIs (business impact and metrics of the model)
(%) Customer recovery rate reported with abandonment behaviour. (value) ROC curve (Receiver Operating Characteristics). AUC > 0.9
Funding
Private
Collaborators, Partners
Own resources
