The fact that predictive analytics can generate a scenario based on how a set of variables interact makes them superior to static business rules evaluating single variables serially in certain applications.

While business rules are useful in narrowing down lists of suspected or future fraudulent activity, for example, they are too rigid as they consider the input variables in isolation. What makes predictive analytics a compelling business case is the fact that domain experts can use their own extensive knowledge to consider not so obvious variables and to harness data mining to take into account emerging experience.

While there are many business uses for predictive analytics, the bad news is that developing and validating the predictive model requires a good deal of business domain and analytical expertise: for establishing a goal; for identifying predictors; for developing a model that is accurate, valid and useful; for evaluating predictive results; and for deploying the predictive application within the organization.

These activities represent a major investment in effort as well as expertise; which may not always be available from within your own organization. However, the knowledge of experts is vital for building an effective predictive analytic model. The good news is that vendors are continuing to develop preconfigured predictive applications and models for diverse areas of the business such as credit scoring, marketing and cross-selling and fraud detection.

While most business users cannot participate in the process of data mining, trend analysis or design of prediction models, many business users are already using the products of predictive analytics, for example:

Customer Service Representatives:  Predictive analysis in applications for marketing campaigns, sales and customer services are becoming increasingly common. The insights from predictive analytics are used to build scripts that are used by customer service representatives to deliver campaign messages, proactively identify and mitigate issues that have the potential to cause client dissatisfaction with services, or to improve customers' sentiments about the company.

Claims Adjusters:  Fraud is a big problem for many insurers and can be of various types—primarily the filing of claims for loses that were not incurred, or inflating the value of an incurred loss. Predictive analytics has identified indicators of the possibility of fraud, such as the frequency of claims, the proximity of the claim to the inception of the contract, an unreasonable distance between the service provider and the claimant's residence, etc. The system can score the claim based on such factors and refer potentially problematic claims to specially-trained adjustors—making more efficient use of the normal adjustors' time.

Underwriters:  Underwriters have used predictive analytics since before the term was invented. The estimation of the probability of loss based on the clients' socio-demographics, past history is the definition of underwriting, and an application of predictive analytics. The automated application of those predictive models is called automated underwriting: it has been used to reduce the underwriters' workload by permitting the simple decisions to be made automatically and requiring human intervention only for the more ambiguous cases.  

Back Office Processors:  The widespread adoption of business process management (BPM) systems to handle transaction routing in the back office has opened the possibility of using predictive analytics to assist the back office. With predictive analytics, it has become possible to gauge the propensity of a transaction to require more work based on characteristics of the customer and/or the request. For example, a model that predicts the number of follow-ups needed to complete the underwriting requirements of a life insurance policy based on applicant age, gender, address (rural vs. urban, etc.) can be used to predict how much work will be needed to complete the application. The BPM assignment logic can thus be used to equalize workload taking into account the expected difficulty of the task, not just the number of transactions to be handled.

Distributors:  As more information is obtained, and as more insight is developed on the buying behavior of insurance clients, guidance to distributors on the likelihood of converting leads into applications and applications into policies will harness the powers of predictive analytics to make better use of distributors' time.

Distribution Management:  The average age of an insurance agent is 57. As retirements reduce the number of agents and advisors, the need to recruit replacements becomes acute. The investment, in time and money (in start-up financing), required to ensure a recruit's success is increasing. Predictive analytics offers the possibility of identifying the recruit characteristics that agency builders can use to improve their chances of developing a successful producer.

Actuarial/Marketing:  These disciplines are built around predictive analytics—to design and price insurance products and determine the population to which the products would be attractive. The data explosion simply makes their respective applications of predictive analytics more effective.

By using predictive analytics, insurers can position themselves to take advantage of opportunities in the future. They can gain insight into their information, reduce risks, and improve efficiencies across all areas of their business.