Over the past decade, predictive analytics in the insurance industry has evolved from a tentative experiment to a competitive necessity. Any insurer that hopes to see the end of the next decade will need to augment traditional actuarial analytics with strong predictive analytics in marketing, operations, cash flow management, risk assessment, pricing, and claims. Although the Ph.D.s producing sophisticated models may get the headlines, the fact is that effective analytics must start with quality, relevant data. Good data can sometimes compensate for mediocre analysis, but the opposite is never true: Bad data or data implementation will always lead to bad results — no matter how skilled the analyst.

Insurance companies collect vast amounts of data. Policy data, billing information, underwriting data, and claims data are among the most important sources. The analyst's job is to make good use of all the data available within the company. However, this means that the original raw data for analytics is sourced from systems designed for purposes other than analytics. Rarely will a company collect data specifically for use in advanced analytics. In other words, analytics is an opportunistic user of data. The implication is that analytics rarely sets the quality requirements for the data. The concern for analytics often is as much about understanding and evaluating the characteristics and limitations of the data as the ability to manage data quality.