Case Study: Machine Learning based Effective Campaign Management
Company spends lots of money to promote their products. This is a success story about how ThirdEye managed their campaign effectively to optimize ROI (Return on Investment).
The customer is a leading budget telecom provider headquartered in California, US. It has business spread across 2 continents and 4 countries and continues to grow rapidly. The company’s aim is to expand its customer base and retain existing customers by optimizing company’s profit.
The company approached ThirdEye with two business problems:
The company wanted to promote their products based on the customers’ behavior. The behavior includes usage, spending, and is also based on certain customers’ attributes like region, subscription, etc. The company wanted to increase the ROI for each campaign and optimize revenue.
At the same time, the company wanted to optimize its templates that were used to send messages through several channels in order to increase customers’ engagement.
ThirdEye proposed Machine Learning based solutions to accomplish customer needs. The machine learning solutions were supervised so that it can be tweaked as per needs.
The first solution was a Multi ARM Bandit algorithm-based machine learning process to choose the right template for the right customer. This process is called Template Optimization.
The second solution was the Decision Tree machine learning approach to select the right subset of customers to campaign a product to ensure a higher conversion rate.
The process steps were as follows:
- Extract Data from source and load the data into NoSQL (Couchbase) database.
- Setup campaign.
- Create an initial segment of customers and set a number of templates for the campaign.
- Run template optimization.
- Send emails to users.
- Capture the user’s response.
- Once the campaign is completed, run Decision Tree algorithm to find the right set of customers to promote a product.
- Couchbase Server – NoSQL Database
- Elastic Search – To offload indexing from Couchbase Server
- Hadoop Framework – For processing data
- Python – Machine Learning and other data process
- Sring Boot – To build the APIs on top of Elastic Search and Couchbase Server
The customer got better than expected ROI (Return on Investment) with a huge opportunity to increase the customer base with better engagement towards campaign. ML based campaign management with automated message sending option significantly reduces communication gap, spending, manual effort, etc. which are trade off from operation prospective.