Predictive analytics for enhanced customer experience

Turning disparate datasets in a hybrid environment into actionable customer engagement strategies

BUSINESS OBJECTIVE

An online and offline retailer wanted to merge customer data of an acquired company with its existing customer data to enhance its advanced analytics programs. How could it collect, centralize and analyze data held in multiple locations to ensure accurate, reliable results?

SCENARIO

• The retailer wanted to exploit sales transaction data to take advantage of data insights and predictive analytics.
• With multiple subsidiary systems, the company struggled to aggregate hundreds of terabytes of data to ensure it was analyzing the widest possible datasets.
• Synchronizing data from subsidiary Microsoft Azure HDInsight clusters to its on-premises Hadoop clusters was hindered by high WAN latency.
• Requirement to replicate data consistently to the parent company's data lake to enable corporate-level analysis and insight.

CHALLENGES

• Ensure global data consistency and accessibility of hundreds of terabytes of data across a complex multi-location, multi-vendor cloud environment:
  • Hortonworks, on-premises Hadoop and Azure HDInsight.
  • Separate corporate networks and with many firewalls between zones.
  • Different underlying file systems between zones (Azure WASB, Azure HDInsight)

SOLUTION

Use WANdisco Fusion's multi-platform proxy server architecture to provide the LiveData platform despite the highly complex environment.

ENVIRONMENT, TECHNOLOGY, AND TOOLS

• Hive
• Spark
• Hortonworks
• Azure HDInsights
• Azure WASB
• WANdisco Fusion

CAPABILITIES INTRODUCED

WANdisco Fusion powers a LiveData platform that provides realtime data replication for high availability and near-zero RPO and RTO disaster recovery.

BENEFITS & OUTCOMES

• 'Always-on everywhere' data management strategy delivers critical analytics capabilities and provides enterprise disaster recovery.
• Enterprise License Agreement enables the corporation and all its subsidiaries and acquisitions to deploy unlimited petabyte-scale replication.
• Heterogeneous data copying from inbuilt heterogeneous platform support.
• Always-on unidirectional replication set up with optional bi-directional replication of specified paths on demand.