1. Improve Data Collection
   In order to truly benefi t from robust analytics, insurance companies have to enhance their collection of cross-organizational data on customers. Data needs to be collected from all business units, e.g., fi nance, marketing, sales, actuarial, etc. In addition to internal data, external data-which has been largely overlooked to date-needs to be integrated as well. Examples of external data include commercial credit data from D&B, consumer credit data from Experian, demographic information from Claritas, or sector-specifi c data from specialized data aggregators. For example, an auto insurer may collect a customer's credit score, motor vehicle violation data, and claims history to improve its models. Data un-availability can pose signifi cant problems in conducting appropriate analytics. For some products, there may be so few customers that analytics cannot be applied. This problem is not unique to the insurance industry-even in the Financial Services Industry analytics cannot be leveraged as extensively in the corporate loan market as in the consumer credit business. Sometimes, drawing parallels between other products within the company leads to better underwriting. For example, while underwriting for large companies, parallels from middle market segments may provide effective guidance.
   
   
2. Prioritize Analytics across the Customer Life-Cycle
   Once the relevant internal and external data has been collected, companies need to identify specifi c areas for conducting analysis.
   
1. Customer Acquisition
   Traditionally, insurers use multiple channels to acquire customers-the health insurance sector relies mostly on brokers; personal lines rely on direct and broker channels. Companies can harness the power of analytics by using it for the following:
• Assessing Channel Economics: Understanding drivers of channel costs, allocating costs across channels, and prioritizing investments across channels
• Improving Channel Effectiveness:
  
  • Direct Channels: Improving effectiveness of marketing techniques, optimizing spend through segmentation, and better understanding of current and prospective customers
  • Indirect Channels: Improving decisions around sales and agent deployment, and analyzing broker incentives and commission schedules
    
• Identifying geographies to improve acquisition coverage
• Assessing characteristics of high-value customers

The end result of these analyses should be to generate marketing lists, such as customer and broker prospect lists, cross-sell prospect lists and broker coverage lists.

b. Underwriting
Experienced underwriters have traditionally used analytics and statistical modeling extensively (e.g., casualty modeling, life expectancy models and claim occurrence models). We believe that there some areas where we can benefi t from improved analytics

• Risk Assessment: Building predictive models to manage the overall risk for the entire portfolio, or an individual product across an industry, product type, or region.
• Product Pricing: Once such risk models have been created, it becomes simpler to create pricing tables that are determined by segmenting customers based on their risk and projected cash fl ows. Some states limit the price that may be charged for a particular type of insurance which can lead to losses on unprofi table customers for the insurer. Analytics can help identify customers that are profi table, and evaluate the risk of insuring the unprofi table customers. Insurers can approach regulators with facts to support a decision to change the price limits. This type of analysis also helps pricing in a hard market, and adjusting the terms and conditions to ensure that price setting is within acceptable bands.
• Loss Reserves: Claims prediction models can help develop appropriate loss reserves which are better aligned to expected needs. This analysis can also encourage an insurance company to negotiate better deals with the credit rating agencies, freeing up capital for deployment elsewhere.
• Product Design: The relationship between customer profi tability and product features is crucial; models can help determine the best product for a particular customer to maximize profi tability, given certain constraints.

c. Claims Servicing
Analytics can be used to improve claims processing in the areas of:

• Identifying and fl agging potentially fraudulent claims
• Identifying claims with potential for subrogation
• Identifying customers who are more likely to switch to a less expensive channel (manual processing costs 50 times more than online processing)

d. Customer Management
Segmentation analysis, logistic, and linear regression models can be used to better manage existing customers:

• Identifying characteristic of customers that are profi table and those that are not
• Identifying customers that are likely to attrite and devising strategies to retain them
• Identifying cross-sell or up-sell opportunities
  
  

• Creating a Strategic Imperative Once the end-to-end customer lifecycle has been analyzed and the key focus area identifi ed, a clear point of arrival vision needs to be developed and communicated widely across the organization to better manage the change
  
  
• Building Data Gathering and Storing Capabilities
  In order to capture internal and external data, an insurer will need to link multiple databases (e.g., marketing, broker information, claims history, product details), which is a signifi cant task and appropriate resources need to be deployed. A key lesson to remember is that getting the right data is important to minimize investments in data mining. Many companies have lost millions of dollars in trying to get "all" data, most of which is irrelevant and low value add for analytics required (Pareto's 80-20 principle applies here as well). For example, when customer level aggregated monthly information will suffi ce, investing millions of dollars in creating a data asset that stores daily information is not required. In addition to the data asset, a bestin- class IT infrastructure is required to support the analytics. Components of this infrastructure include ultra-high performance servers, ample data storage for growth in data collection, analytical tools (e.g., SAS, SPSS, CART and MARS), availability of high-speed connectivity to all users and Business Intelligence tools to support ongoing reporting.
  
  
• Outlining Effi cient Process for Analytics
  For the analytics to be effective and effi cient, processes must be defi ned for the inner workings of the system. Examples of such process include building generic and custom models, and sharing learning and best-in-class practices across the company.
  
  
• Implementing an Analytical Engine
  After data has been properly gathered, cleaned and stored, and an analytics process created, it is necessary to implement the analytical engine. The output of this engine should be simple lists that the front-line staff can act upon. Models need to be also updated to refl ect new policies and new data. Guidelines for data/model access and use should be clearly laid out.
  
  
• Monitoring of Results
  The fifth foundation of this capability is to implement a process to monitor progress and make appropriate refi nements to the engine. To do this, it is vital to choose the appropriate metrics across each element of the customer life cycle, measure performance at regular intervals and compare it to a predefi ned baseline. An appropriate combination of both quantitative (e.g., loss ratio, amount crosssold, customer retention/acquisition) and qualitative metrics (e.g., customer satisfaction, ease of use/ implementation of models) should be used. Clear management reporting needs to be developed which tracks progress against baselines and measures shareholder value creation.
  
  
• Enabling Change: Funding, Governance, and Resources
  It is essential to have senior level support, sponsorship, and clear ownership of analytics functions in order to gain organizational traction. The whole process faces signifi cant risk of failure if it is regarded as another "nice-to-have-idea" and treated as a "utility/support" function, and not as a mainstream line function. Funding, setting goals, and employee incentive structures are important. It is imperative that appropriate resources be available to execute the numerous steps in the analytics process. Employees supporting the analytics functions fi ll the following roles:
  • Business Analysts to understand overall objective and strategic imperative, structure/ prioritize analytic initiatives and identify new opportunities for the application of analytics
  • Modelers to execute analytics and create data asset
  • Technology experts to assess and implement infrastructure needs
  • Analysts to implement and maintain models
    

Summary of Learning

SOPHISTICATED ANALYTICS HAS BEEN APPLIED IN MANY industries. The most successful companies have been able to track customer behavior across the customer life cycle and translate that into informed actions. Insurers have an abundance of data across their organizations, but most have not leveraged the full potential of this data in customer acquisition, underwriting, claims servicing and customer management. Insurers need to improve data collection, prioritize the application of analytics across the customer life-cycle, and build an analytics capability to create a sustained culture of data driven decision making and drive shareholder value.