Has the following situation ever occurred in your company? In reviewing a routine report, a claims manager notices a significant increase in claim severity over the past three months. It is important for her to discover what is driving this trend. Luckily, her company implemented a new data warehouse several years earlier. In the past, she might have had only limited access to reports, enabling her to see only whether the trend existed in other states or in major claim types. Now, her department has access to much more information about claims activity.
She asks her department's reporting analyst if he can use the data warehouse to locate the causes of the trend. He conducts an extensive search of the data and creates more than 100 new reports-some that show the concerning trend and some that do not. He combines all the reports into a PDF file for the manager, confident that he has provided the information needed to identify the source of the problem.
When the manager opens the PDF file, she is shocked by the number of reports. After paging through a few of them, she realizes that each report provides a clue to what the drivers of the trend may be, but can she truly identify the causes? Despite access to an entire database, she still is not sure of the causes.
This scenario is all too common in this age of increasing data abundance. Many companies have developed centralized data warehouses to enable users from across the organization to obtain data efficiently. These are significant undertakings that involve integration from across major operational areas. In many cases, the projects require years of effort to plan and execute. However, if the resources are being used only for report generation, the company has yet to realize the true return on its data warehouse investment.
Just as in the example above, many requests for data are motivated by reaction to emerging business issues. Drill-down capabilities enable analysts to embark on a quest to dig deeper into different dimensions of the data in search of key insights. The ability to perform ad hoc and drill-down queries is critical in using data to understand factors that drive the business. However, if a company does not also build analytical capabilities, it will drown in data without meaningful business insight.
Exploring Data Elements
There can be scores of data elements to explore when looking for the drivers of important business trends. Each data element has the potential to indicate a cause. Drill-down queries enable you to search numerous paths-many of which are dead ends. Even if you find some evidence through one of these paths, you can easily lose the context of the data elements that got you there. As a result, you wind up asking yourself questions such as: Which of all the data elements is the one that really matters? If I had drilled down a different way, would I have seen the same pattern? Have I gone so far that the data is too thin to show a meaningful pattern? How can I be confident that I haven't bypassed what I'm looking for?
The solution is to use analytics to search for insight. Drill-down reporting and predictive analytics are both important tools in an analytically driven organization. However, many companies plateau after drill-down functionality is established. Rather than building on their data platform to develop predictive analytics, they remain in a scavenger-hunt mode, generating many reports but lacking the ability to improve the management of their business proactively.
Statistical and other analytical tools are able to detect patterns across a wide array of data elements and bring those patterns to the attention of the analyst. As a result, such tools enable an analyst to focus on likely causes and bring potential drivers to the forefront. The analytic tools can be fed by the same data as the querying and reporting tools. And, analytical tools enable the key data to come to the surface.
Once these insights are identified, management can address the problems brought to the forefront by analytical tools.
David Cummings is VP and chief actuary at ISO Innovative Analytics, a unit of Jersey City, N.J.-based ISO.
