What's the best way to ruin your business? Ignore your customers, charge high prices, and deliver less value than your competition. Each driver, homeowner, renter, business owner, factory operator, risk manager, agent, broker, or web user will perceive you as inaccessible, expensive, and inept when compared with others in the market.
Whether you're at a mutual company or a stock company, a multiline multivertical multinational or a monoline monostate carrier, you need to be competitive where it matters most to your customers. You should be easy to work with, highly reputable, dependable, and financially stable, while meeting your shareholders' expectations of profitable growth, smart expense management, and agile innovation in the market. These are the things consumers want in an insurance company — along with a great price.
Every executive in the market feels the pressure to compete on analytics, but few have a road map showing their strategy, actions, investments, and governance. They all want — and need — to know how to make decisions on predictive modeling goals and how they compare with their competition.
Where am I now? How did I get here? Where do I go next?
Most insurance carriers have yet to produce highly capable, mature analytics — mostly because of the size of the organization and how their executives feel about predictive modeling. Table 1 indicates that more than 75 percent of companies (255 out of 327) serving the private passenger automobile market have premiums less than $100 million, with an estimated customer base of less than 100,000 customers. For small commercial lines carriers, the relative counts are even smaller.
U.S. PERSONAL AUTO INSURERS AS OF 2010:
MARKET SHARES AND POLICY COUNTS BY SIZE COHORT (USING $1,000 PER POLICY)
|A.M. Best 2010 / U.S. Personal Lines Insurance Groups by Premium and Market Share with Customer Estimates (using $1,000 per policy)|
|Premium segment||$1 billion+||> $500 million||$400M||$300M||$200M||$100M||$50M||< $50M|
|Number of insurers
|> 1M||500,000||400,000||300,000||200,000||100,000||50,000||< 50K|
A wide disparity exists in the size of companies in the marketplace, with most companies having less than 100,000 customers. The systems, processes, and methods that work for a company serving 50,000 customers are not necessarily the same as those that will work for a company with 500,000 customers — or 5 million customers.
Source: ISO analysis of direct written premium data purchased from A.M. Best
Determining "Where am I now?" is linked to your IT architecture, computer platform, consultant budget, expenses for proprietary data and third-party scores, and ability to hire and train experts in both business context and predictive analytic knowledge, skills, and abilities. That said, surprisingly, not all the largest carriers are experts; and more surprising, many smaller carriers are very fast learners.
Understanding customers is what competing on analytics is all about: What do they want? How do they want to purchase products and services? Why did they choose you? What makes them stay or leave? What makes them recommend you to family and friends? Which ones cost the most to serve? And do your promises, prices, and products provide for each market segment?
The questions evolve as you learn more, and you learn more as your capabilities to compete on analytics mature. The more effectively you navigate the predictive analytics capability maturity model (Figure 1), the better you can group customers with similar needs, preferences, and other important attributes such as riskiness, loyalty, ability to add other products, and so forth. Moreover, you can more consistently select risks and accurately forecast their cost of losses, expenses, retention, and lifetime profitability.
WHERE DO YOU FALL IN THE SPECTRUM OF PREDICTIVE ANALYTIC CAPABILITIES?
Companies vary widely in their abilities to create and use predictive information to manage their customer portfolios. There are seven stages of development for predictive analytic capabilities in insurance, and each has a level of investment and an expected return. The companies with the most mature capabilities will have invested in all seven stages shown in the illustration and, depending on individual jurisdictional restrictions, will have deployed analytic models to serve their customers and compete for others.
For your specific portfolio of customers and their various segments, you should ask — and be able to answer — a few general questions: Compared with last year, are your current customers better than before? Did your most desirable segments grow? Who left, why did they go, and where are they now? If you reach your goals, what will next year's strategy look like?
The sharpest part of the cutting edge is the ability to translate consumer insights into faster, easier, cheaper ways to let customers join your brand in ways that appeal to them — through a smartphone, agent, telephone, or the Internet. That's a much more efficient approach than expensive manual underwriting steps. You can take advantage of predictive analytics to introduce a data-driven process flow into your company and simultaneously use analytic capabilities in advertising, marketing, underwriting, risk selection, distribution, channel management, fraud prevention, customer service, and claims integrity. Like most proficiency-based expert capabilities, analytics works better the more you use it. So the challenge is to increase the predictive analytic capability in all aspects of your business.
The capability steps help you determine where you stand currently. But your answer to "How did I get here?" may reveal avenues of varying quality leading from the chief actuary, CFO, CMO, CIO, COO, CTO, smaller groups such as claims and underwriting, and inside operating groups such as personal, commercial, or specialty lines. Generally, personal lines are the universal leaders. But which area specifically championed the first effort and how broadly capabilities have penetrated the organization will vary. Most smaller carriers are relying on consultants to advise them because they haven't built internal teams yet. A few of the largest organizations have now added an analytics executive to their board room team — the emergence of analytics into the culture of the organization has begun.
No matter how you got where you are, "Where do I go next?" is a function of time, expense, and goal priority. But with price and profitability always a concern, analyzing loss costs is imperative since segmenting customers based on risk is the surest way to be an enduring competitor. Actuarial and product management analysis groups favor loss segmentation predictive analytics, most claims organizations focus on fraud analytics, and marketing departments rely on strategic analytics to attract and retain customers.
How various carriers compete on analytics will differ, but there are only so many options. Whether your company is large or small and regardless of how it's organized, each analytic team (consultant or internal) must progress through the stages of capability maturity to achieve the best results. Generally, the larger your team, the more money you'll need to produce a model. If your team isn't very large, you may lack the resources — time, money, people, data — to complete a model. If you engage a consultant, because of time constraints, your internal team may not gain sufficient insight on the context of modeling and learn only the rudiments of operating analytic software. As a result, the model they produce may not be suitable for business use.
Many projects fail completely or fail to surpass an industry solution because of a lack of data or insufficient value. A champion-versus-challenger framework exists to assess both the technical merit of alternative models and the cost footprint of equivalent analytic endpoints. Often junior staffers miss the obvious advantage of incorporating component technology to build customer-facing solutions because of their desire to do everything themselves.
To forgo the use of an industry score is ill-advised when you compare the cost and time spent to arrive at the same or less accurate answer through in-house efforts. Consider that many carriers never developed proprietary credit scores, while many more use the same vendor they chose back in the 1990s. Start out using viable industry scores as a benchmark, and try not just to reaffirm your own beliefs by looking at independent graphs. Rather, explore the inter- and intra-segmentation effects of combining new data with existing information on generating customer insights. Two segmentation structures, each performing well, may achieve better performance when used jointly.
It's often easier, cheaper, and faster to exploit new learning than to spend your monetary and intellectual capital trying to make an expensive copy. The knowledge you gain about "what you didn't know you didn't know" can be used to refine your next investment cycle. If you can use a self-sustaining data-driven approach, you'll be able to eliminate the cycle of reanalysis that permeates many areas of traditional actuarial service providers. Deploy your capital and talent to meet challenges and revisit your solutions any time that better data, better analytics, or better decision-support opportunities arise.
Beware of the hype. Plenty of Big Data start-ups are looking for innovation investment dollars, and many IT departments feel it's their turn to contribute to the trend. But a dozen terabytes is not Big Data, so remember what matters is what you do for your customers with the analytics you already have. The last thing you need is to over-promise or under-deliver on projects. And dollars misspent only increase your expenses and hinder your time to implementation.
Always consider build versus buy
Spending more to get the same result is a waste of resources. A small carrier can get a state-of-the-art loss-cost estimating model for under $100,000 a year, and larger carriers can save many times that amount by assigning their staff or consultants to projects where they can add more value. Accuracy and relevance should be integral factors in your decision-making framework.
Time to implementation is always faster — and the risk of miscalibrated results diminishes — if the solution is already approved under state regulations.
Here's a rule of thumb: Buy a product, and measure what you pay for what you get. Try a service, and measure what you get for what you pay. Build a solution using both to establish quick effectiveness and sustainable capabilities. But be on guard. Your consultant and your internal staff should help you be prudent in evaluating your needs. Otherwise, they'll bill you even if you never get a model that's more effective than an off-the-shelf product. There are plenty of challenges to tackle beyond the ones that have solutions already available.
The elegance of refined analytics is in the transparency linking your shareholders' interests with your customers' needs. Commit to your risk-based pricing strategy, take actions to generate customer insights, make investments in predictive analytics across your enterprise, and provide governance that demands discipline in build-versus-buy capability acquisition. The synergistic effects of strategy meeting action and aligning business goals with predictive analytic investments is what makes competing on analytics an irresistible force in today's marketplace and a cogent approach to customer interaction.
This article was originally published in Verisk Review.
Marty Ellingsworth is president of ISO Innovative Analytics (IIA), a unit of ISO focused on advanced predictive modeling tools for the property/casualty insurance industry. Mr. Ellingsworth joined IIA from Full Capture Solutions, Inc., where he was cofounder and executive vice president. He has more than ten years of experience in the property/casualty insurance industry, with a focus on applied analytics and claims. For five years, he served the Fireman’s Fund Insurance Company. Mr. Ellingsworth’s 24-year career also includes positions at RiskData/HNC Software, Workers Compensation Research Institute, Beech Street Managed Care, and the U.S. Air Force. He received his bachelor of science degree in operations research from the United States Air Force Academy and his master of science degree in operations research from the Air Force Institute of Technology.