Summary & Goals


  • Issue with high churn (60%)
  • Low level of customer understanding


  • Reduce customer churn rate
  • Build behavioural insights and churn prediction model for new customers
  • Build propensity modelling to estimate adoption rate of couponing

What I did

1. Exploratory Data Analysis of 120k customer base transactional data (data understanding and cleanup, correlation analysis, profiling)

2. Feature engineering to feed more information to the machine learning models

3. 'Manual' Customer segmentation (RFM) and Cohort analysis for LTV

4. Test of existing libraries for calculation of 'remaining' Lifetime value at single customer level

5. Built model to evaluate customer preferences for booking products (dates, specific products) to provide specific offers (i.e customer has a birthday, or specific time of the year preference to book) and understand if this could be helpful in predicting churn

6. Built RNN (recurring neural network) to predict churn and customer clustering to see similarity of customers at each stage of their lifetime

7. Understood couponing activities impact on customers (i.e. re-activation from churn, incrementality) to outline expected value of running couponing and adjust accordingly future campaigns. Predicted propensity to use couponing.

8. Created collaborative recommender system to provide product recommendations on website.

9. Ran marketing activities on new built customer segments and preferences

Customer Behaviour prediction


  • Churn reduction of 10% in 6 months
  • Increase of CLV of 25% and CAC reduction
  • Creation of new user-level dashboards to understand trends and changes in behaviour through time
Scroll to top