The Conexis Blog


As we are approaching mid 2012, the hype surrounding big data is reaching a deafening crescendo. Show me the use cases is what we hear from the uninitiated. Let’s talk about one specific customer use case and then plan on providing frequent updates as we have time to add to this space.

Let’s talk about how one customer, a major international retailer based in the United States utilized the convergence of structured and unstructured data to improve customer acquisition, reduce customer churn and provide offers optimized in to increase retention for customers. This company desired to structure a customer acquisition and retention program with a 360 degree view of their customer, and to develop churn indicators and next best offers so they could better predict which customers were likely to leave in a specific time interval and make them an offer to incent them to stay.

This required processing of massive amount of data coming from different systems, both transactional (ERP & CRM) and interaction data (Social, CDR and Web). Required access from CRM system including customer account information as well as customer order and product configurations from their ERP system, social media data like Facebook, call detail records from support, and clickstreams from web logs. This is a challenge for developers because it requires expert knowledge of the source systems and not only do they have to resort to hand coding to access the data but they have to thing about how well it’s going to perform as the data volumes grow and will it impact the source systems and any applications that use that data?

With casual integration, they can access that data whether it’s real-time or batch and they can load that data to Hadoop HDFS file system because the data is accessed through native API’s from the source systems and they can further increase the performance of that data delivery because they can access it with partitioning that partitions the data coming from the source into HDFS for high performance and scalability. We also populate the Hive data tables so that the data scientists can quickly query the data from Hadoop, understand it and validate assumptions as they work to find the algorithms that are used on Hadoop.

Since this customer was doing Big Data processing on Hadoop, they needed to determine customer sentiment, calculate customer churn indicators, and determine next best offers. Once that was all processed the data was delivered from HDFS to their CRM and Data Warehouse. So now, the client has a 360 degree view of their customer that included churn indicators, and the next best offers. Now the customer could better predict which customers were likely to leave within a defined time interval and quickly and proactively make them a next best offer to incentivize them to stay. This customer chose this solution because they could best manage their current information stores without interfering with existing systems, help increase the productivity of their developers and data scientists, and simplify the maintenance and access of any and all data, either Transaction or Interaction via big data processing.