Case One
Medihelp: Transforming Traditional Databases into Business Intelligence

Medihelp is South Africa's third largest health insurance company. It covers about 220,000 people with plans ranging from R828 to R2,700 (about U.S. $80 to $260) per month per person. Medihelp needed a better way to access and analyze data on customers, claims, and third-party providers in order to monitor the effectiveness of its insurance products and to fine-tune and create new products as needed.
Medihelp's problem was that its data was stored in a traditional database. As discussed in this chapter, traditional databases are not designed to support decision making. They're not efficient at the types of information retrieval that decision making uses. With Medihelp's existing database, reports took unacceptably long times to run. Reports based on the content of the full database couldn't be run at all.

This inefficiency detracted from Medihelp's ability to make informed business decisions.

For example, Medihelp's claim file had about 55 million rows. Each row contained about 35 data values describing medical conditions, treatments, and payments. Each row was associated with one of 15 million rows of historical member data, which held another 15 or so data values describing the member and his or her coverage. For claim processing, accessing the data for a single claim was fast and efficient. Combining data from thousands of rows, as decision support calls for, was not.

"Logging into our [traditional] database ... was not providing us the information we needed to make the best business decisions," explains Jan Steyl, senior manager of business intelligence at Medihelp. "We needed a dedicated, high-performance data warehouse."

After looking at several options, Medihelp made a preliminary selection of Sybase IQ as the basis for its data warehouse. Working with B.I. Practice, a Sybase subsidiary in South Africa, Medihelp carried out a proof of concept to confirm that Sybase IQ could deliver the needed performance at an acceptable cost. This procedure involved loading a subset of the tables from the operational database into Sybase IQ, executing queries that used only that subset of the data, and evaluating the results

Because these queries now used a database designed for queries, performance improved dramatically. Response time was reduced by an average of 71:5 percent. Response time for ad hoc queries, those which were not programmed into the database system ahead of time, was reduced by an average of 74:1 percent. One query's response time dropped by 92:8 percent.

Theo Els, Medihelp's senior manager of client relations, likes the new system. "Health insurers supply data to mployer groups. These demographic and claims profiles are essential for employer groups seeking to understand their employees' health risks ... and the consequent impact that risk can have on business productivity. Brokers and healthcare consultants also use this information in their annual client reviews to ensure that employees receive the most suitable coverage. It is imperative that the ... data warehouse provide accurate information in a format that is easily understood."

The biggest beneficiary of Medihelp's data warehouse is its product development team. The team uses data from the data warehouse to understand trends in claims by benefit code, condition, area, age group, and other factors. Medihelp also uses its data warehouse to determine what financial effects changes to a benefit in a specific product will have. This provides the sales force with the right offering at the right price for specific target markets in South Africa.

Discussion Questions
1. How did Medihelp's use of a traditional database limit what it could do with its data?

2. How does Medihelp's Salesforce now use the information in the company's data warehouse?

Critical Thinking Questions
1. How did B.I. Practice improve the efficiency of queries run by Medihelp employees?

2. How does the new query system help Medihelp make better decisions for the company? Does Medihelp also use the system to help their clients make better decisions? If so, how?