How Insurers’ Big Data “Moneyball” is a Game-Changer for Agents

Agents often say they want a way to differentiate in a crowded and noisy marketplace.  This underwriting revolution presents a sustainable competitive advantage to those willing to invest in gaining knowledge and expertise.

The National Council of Compensation (NCCI) Experience Rating plan was created to adjust premium costs to reflect "the unique claims experience of each eligible individual employer relative to other employers within the same industry group, according to NCCI. Using broad job classifications and aggregating data from across a state, the experience rating plan helps insurers charge the appropriate premium for an individual employer's work comp policy. Or, as one actuary stated, "the experience mod is a predictive indicator of future losses." Traditionally, a higher experience mod predicts that the employer will have greater than expected losses in the coming policy period, so the insurer needs additional premium for the risk.

Most experts would agree that the experience rating plan has historically served the insurance industry well. But we are entering a new era where individual insurers are building their own predictive analytics models because of:

• Recent and swift explosion of huge databases
• Inexpensive computing power and storage
• Advances in data acquisition and aggregation from multiple sources
• Applied statistics
• Machine learning techniques.

Computer hardware and software advancements now allow insurers to quickly process millions of calculations, analyze the data they produce and promptly validate their emerging predictive models.  

Before these technological advances made it easy and cheap for insurers to crunch their own data and create their own predictive models, they relied on the rating bureaus to collect the data and apply their models to leverage economies of scale. 

Big data has made it possible for insurers to play their own version of "Moneyball," Michael Lewis's 2003 nonfiction book about the business of baseball. General Manager Billy Beane, faced with the franchise's unfavorable financial situation, takes a sophisticated analytical approach toward scouting and analyzing players. He leverages the inefficiencies in the system to compete with a smaller payroll than his competitors.

There are significant inefficiencies in the rating system that data-savvy insurers can leverage to gain a competitive advantage. They can analyze their own data instead of relying on the rating bureau's broader, aggregate view to create a competitive advantage.

Let's assume the rating bureau's data indicates that claim costs are rising for plumbers in a given state. The rating bureau will likely increase advisory rates and expected loss rates for plumbers in the entire state. However, an individual insurer analyzes its own book of business and sees a decrease in claims costs for that state's plumbers. The carrier could set a lower premium for plumbers and capture greater market share from competitors that only use aggregated rating bureau data.

Premium growth in the insurance industry has been less than 5 percent over the past 10 years, so taking profitable business form competitors with a better data mousetrap is a good way to grow. Data is a vast new natural resource, according to IBM CEO Ginni Rometty: "Organizations that do a better job of finding, extracting, and refining data resources to produce insight will gain competitive advantage."

It's no surprise that large global actuarial and consulting firms are working with insurers to develop and enhance predictive models. Insurers already possess a treasure trove of data just waiting for the "data nerds" to spin it into gold. 

As one actuary from a well-known consulting firm said at a recent industry conference, "Underwriters have been using about six data points to determine acceptability and pricing of a risk. We can build them a model with 400 data points." 

While consulting firms and insurers are not likely to share their specific list of data points, you can assume that it will include both changeable and unchangeable data. For example, business location may be a heavily weighted item, but it is not likely the business will move to get better insurance rates.  However, many data trends can be influenced by effective business practices. For example:

• Is frequency increasing or decreasing?
• Are claims developing at a higher or lower cost than expected based on diagnosis?
• Are lost work days consistent with the expected disability duration?

Big data brings big opportunities to carriers and agents, but with any collision of old-world and new-world methodologies, there will be some challenges and casualties. An underwriter receives an application for a workers' compensation renewal. The applicant's experience modification factor is renewing lower than the prior year, as is the governing class code advisory rate. 

Read related: "Analytics Investments Increase, but Use of Big Data Lagging: Novarica."

However, the insurer's predictive model indicates that this class of business shows eroding performance that is not yet reflected in the rating bureau's data due to the time lag in its data collection. As a result, the underwriter removes the scheduled credit and adds a scheduled debit to the pricing. Now the agent has to explain an unexpected higher premium to the client.

Or worse, the underwriter cannot even make an offer because the maximum allowed scheduled debit will not provide the pricing needed according to the predictive model. In this case, the applicant's reduced experience modification factor actually prevented the employer from getting a renewal offer from its current or preferred insurer.

This may seem crazy, but when you add more and new data to a pricing model, you often get a different indicator. 

Enhanced data analytics can turn conventional rating and pricing upside down. The purpose of the rating bureau's experience rating plans is to assist the insurers appropriately set a price for the risk. However, with advanced analytics and regulations mandating the use of the experience mod, employers may find themselves in the residual market because the insurer was unable to make an offer at their price.

Assume there is a Jan. 1, 2014 renewal. The payroll and claims data typically used for the 2014 experience mod is from policy periods 2012, 2011 and 2010. The most recent policy period, 2013, is not used for experience rating.

However, 2013 is heavily weighted in the insurer's predictive model, which does not give as much weight to the 2010 policy period. In this scenario, you are likely to get very different indicators of future losses. 

Experience mod and advisory rates go down and the pricing goes up. Experience mods and advisory rates can go up and the pricing goes down, the opposite of what you'd expect.

Agents will also see pricing variances from accounts within the same governing class code. Consider the wide range of risks that are eligible for the "auto service and repair" class code, including franchised auto dealers, tire stores, oil change operations and independent repair shops. What happens when an insurer discovers through data analytics that franchised auto dealers perform better than tire stores? The insurance company can leverage the inefficiencies in the advisory loss cost system and adjust its pricing accordingly. 

Workers' compensation experience rating and experience modification factors are not going away any time soon; they are enmeshed into each state's regulatory and statutory framework. And not all insurers will create and utilize their own predictive models, so they will continue to rely on the rating bureaus. However, you're probably beginning to see anomalies between the old world of "predictive indicators of future losses" and the new world of insurance-specific predictive analytics.

Agents must not only be aware of these underwriting changes, but must educate their clients and prospects. The brightest future belongs to employers that can move the loss data in the right direction over the long term. The agent's role is to help them establish processes to make that happen.

Read related: "4 Views on Improving Data Quality."

Having this conversation with clients will also help prepared agents take on the most dreaded response from prospective clients: "We are happy with our agent." It's an opportunity for agents to engage clients in a new conversation and shift the dialog. Buyer surveys have affirmed that they want to learn about any emerging changes that are currently off their radar screens. Agents can ask prospects and clients some of the following questions:

• Are you aware of how the "big data" revolution is impacting your insurance program and pricing?
• Has anyone shared with you how the insurance company's underwriting process is going through its most dramatic change in more than 100 years?
• Have you taken steps to adapt and align your business objectives and risk management practices to leverage this new approach?