Case 1
Purpose
The client, a paper based products manufacturing company wanted to reduce its customer churn in its online business
Approach
Our team aggregated the customer level data into a single source of truth and we then proceed to identify the customer behavior using some simple exploratory data analytics techniques
Due to disparate nature of every customer, to bring some regularization we clustered the customers in different categories based on various behavioral attributes like demographics, average time on site, average number of products they buy etc.
We then calculated customer lifetime journey using proprietary supervised learning techniques like Markov chain models and Bayesian models
The final end to end solution was wrapped around a python web framework with integrated learning algorithms with a thin client facing the customer
Outcome
The client was successfully able to identify more than 90% of its customers that were at the end of its buying life-cycle journey and hence were about to churn
An inclusive strategy was devised to target those customers with lucrative promotional offers and mailers to ensure stickiness to the brand
The customer churn was reduced by more than 30% compared to previous quarter and the overall sales grew by 8-10%
Case 2
Purpose
The client, a leading clothing retail, wanted to understand the best discounts to offer to customers
Approach
The problem statement was taken and broken down into smaller problems:
- Which products would be most responsive to discount promotions
- What are the discounts to be offered on promotionally driven products
We initially collated product data from multiple sources and all the products were clustered using proprietary unsupervised learning techniques
We then built price elasticity models across products to find out where the products stood in terms of promotional responsiveness
The promotionally responsive products were filtered out using SWAT analysis based on price elasticity and customer satisfaction scores.
A differential simulation pricing modelwas used on the filtered products to find the optimized promotional discounts for every product
Outcome
There was a rapid shift from heuristic based pricing to a data driven pricing in the client organization which led to quick and insightful decision making
The client saved $1.5 million in promotional and marketing spend and the overall sales shot up 12% during the promotional period
Case 3
Purpose
A leading baby products manufacturing company wanted to gauge the impact and effectiveness of its promotion strategy
Approach
This was a special engagement as we helped the client in entire experiment design and we were engaged with the client throughout the journey
Initially we collated all the customer related attributes and then classified the customers based on their behavioral pattens and affinity towards the customer's products
We then chose a pool of test customers to roll out the promotion by gauging their behavioral attributes and using sampling across stratas of customer
The behavior of test pool customers was compared to the control pool customers. The customers in control pool were chosen using hypothesis testing and statistical analysis.
The final promotional effectiveness was gauged using a multiple regression model which used customer behavioral factors, market trends and product attributes to create a robust all-inclusive solution
Outcome
Using this experiment our clients were able to identify the product segment and region combination that were most receptive to a particular type of promotion
The final solution helped them to simulate multiple test promotions on product segments and then gauge the promotion effectiveness to select the most optimal promotion for particular product