AI-Driven Campaign Targeting
SUMMARY
- Client
- Raiffeisenbank
- Tech stack
- Profinit AcceptAI (P2L Model), Apache Spark, Hadoop, Python
As part of its strategy, Raiffeisenbank CZ wanted to use customer data to increase offer relevance and conversion rates for consumer loan campaigns. The unique AcceptAI solution models customer behaviour from transactional data, using machine learning to calculate propensity scores. High computational complexity was handled through Apache Spark and Hadoop, processing millions of transactions in minutes while integrating with existing campaign tools.
Results
Project Background
As part of its strategy, Raiffeisenbank CZ wants to use data on customers and their transactions in an advanced way to increase the relevance of offers from the customer’s point of view, target offers more effectively, and increase the conversion rate of a large-scale campaign aimed at selling consumer loans.
Full utilization of customer data requires a sophisticated solution that performs highly complex calculations on hundreds of millions of transactions and an even larger number of possible links in the client network and diverse events. Standard relational databases and business intelligence solutions cannot be used to process these networks.
Challenge
The chosen solution needed to be deployed into the bank’s existing big data platform on Hadoop technology. The solution and its campaign outputs needed to be integrated technically with the bank’s campaign management tools and procedurally with the client advocacy and contact policy mechanism.
Business needs
The solution needed to meet the following specifications:
- Increase the conversion rate of large-scale consumer credit campaigns
- Leverage other channels and increase success rates
- Integrate the solution into CRM and campaign management system processes
- Adhere to strict outreach policy requirements
Solution & Results
The unique solution AcceptAI was chosen, which models customer behaviour from transactional data. Based on the relationships and similarities found, AcceptAI uses machine learning to calculate the propensity scores of individual customers for different financial products.
The high computational complexity was ensured technically by parallelization using Apache Spark and the Hadoop platform. This makes it possible to process all transactions occurring over several years in tens of minutes.
The customer scores are then available in the DWH for campaign management tools. The ability to clearly rank and compare individual customers makes it easy to select different channels for them while maintaining the bank’s outreach policy rules.
HEAR FROM THE CLIENT
Thanks to Profinit’s AcceptAI, we achieved a sixfold improvement in call centre conversion rates in a credit campaign targeted based on customer behaviour. The advanced propensity model running on the big data platform achieved a 50% better overall result than the existing model and improved success rates across all channels.
