This guest blog is penned by Bhavani Rao, Solutions Marketing Manager at Search Technologies, a Cloudera partner specializing in the design, implementation and management of search and big data applications.
It is estimated that workers’ compensation fraud costs the industry $5 billion per year. Fraudsters have grown increasingly sophisticated making it hard to identify and even harder to substantiate and prosecute. We at Search Technologies successfully launched a solution based on Cloudera Enterprise that addressed this issue.
One of our customers collects nearly a $1 billion in premiums yearly and has 1500 employees, of which 500 are attorneys. A good bulk of these attorneys along with a cadre of business analysts are part of the Special Investigations Unit (SIU), whose job it is to identify and prosecute fraudsters.
Sifting through multiple data sources for individual line item details, compiling the data, and completing the analysis to create a solid legal case was extremely slow and laborious – the process could take an entire month. Most SIU agents were not technical so any new requests needed to be sent to business analysts, who would build queries, execute the request against a data warehouse or eight separate databases, and then compile a report. In addition, there was no way to fetch the original bill scans without writing down record numbers and going to the designated server and retrieving them individually. There were fraudsters that just got away because they could manipulate billing changes faster than the customer could audit.
Search and big data to the rescue
The customer wanted a modern analytics platform based on Cloudera Enterprise and asked Search Technologies to architect an entirely different approach. We built a custom application using Cloudera’s Distribution of Hadoop (CDH), Cloudera Search, Zookeeper, our Query Processing Language and a custom user interface. The database search was replaced with Cloudera Search, which enabled faster and easier access to data because it eliminated the need for SQL. Cloudera Hadoop is used to streamline the data preparation process and indexing pipeline, a massive job because there are nearly 100 indexes in production. Query Processing Language from Search Technologies brings great value to feature rich and scalable Solr – the core of Cloudera Search. Query Processing Language enables sophisticated aggregations and processing of search results, at query time. The results are quickly and easily presented via visual and insight deriving charts, in a user-friendly GUI.
From a business perspective, this application has become a competitive edge in the fight against fraud. Investigators can directly query large datasets, without intervention from or delay by a business analyst – we call this “self-service analytics”. Data is available in near real-time with data sets added dynamically if the user has permissions. Facets or filters allow SIU agents to drill down quickly and interactively to individual claim scans of relevance that can be used to build a legal case.
By flattening their database information, the customer immediately saw results. The application scales easily to analyze tens of millions of claims, related records and hundreds of millions of bill line details. And because it’s built on a search engine rather than a data warehouse, queries are retrieved in seconds not hours. SIU agents use a simple search interface instead of creating SQL and can compare statistical patterns across all the data to find outliers, rather than just sampling.
Most importantly, the application has demonstrated immediate ROI by compiling the evidence to prosecute $150 million+ fraud cases. The customer received kudos from a judge who commented on how evidence can be delivered in a matter of minutes rather than three days – faster than any other organization in the workers’ compensation industry.