Tag Archives: statistics

Is the Data Talking, or Your Biases?

In April, a large life insurer announced plans to use Fitbit data and other health data to award points to insureds, providing impressive life insurance discounts for those who participated in “wellness-like” behaviors. The assumption is that people who own a Fitbit and who walk should have lower mortality. That sounds logical. But we’re in insurance. In insurance, logic is less valuable than facts proven with data.

Biases can creep into the models we use to launch new products. Everyone comes to modeling with her own set of biases. In some conference room, there is probably something like this on a whiteboard: “If we can attract people who are 10% more active, in general, we will drive down our costs by 30%, allowing us to discount our product by 15%.”

That is a product model. But that model was not likely based on tested data. It was likely a biased supposition pretending to be a model. Someone thought he used data, when all he did was to build a model to validate his assumptions.

Whoa.

That statement should make us all pause, because it is a common occurrence – not everything that appears to be valid data is necessarily portraying reality. Any data can be contorted to fit someone’s storyline and produce an impostor. The key is to know the difference between data cleansing/preparation and excessive manipulation. We continually have to ask if we are building models to fit a preconceived notion or if we are letting the data drive the business to where it leads us.

Biases hurt results. When I was a kid, my Superman costume didn’t make me Superman. It just let me collect some candy from the neighbors. Likewise, if insurers wish to enter into an alternate reality by using biased data, they shouldn’t expect results that match their expectations. Rose-colored glasses tend to make the world look rosy.

Here’s the exciting part, however. If we are careful with our assumptions, if we wisely use the new tools of predictive analytics and if we can restrain ourselves from jumping through our hypotheses and into the water too soon, objective data and analytics will transport us to new levels of reality! We will become hyper-knowledgeable instead of pseudo-hyper-knowledgeable.

Data, when it is used properly, is the key to new realms, the passport to new markets and to a secure source of future predictive understanding. First, however, we have to make it trustworthy.

Advocating good data stewardship and use.

In general, it should be easy to see when we’re placing new products ahead of market testing and analysis. When it comes to insurance, real math knows best. We’ve spent many decades perfecting actuarial science. We don’t want to toss out fact-based decisions now that we have even more complete, accurate data and better tools to analyze the data.

When we don’t use or properly understand data, weak assumptions begin to form. As more accurate data accumulates and we are forced to compare that data with our pre-conceived notions, we may be faced with the reality that our assumptions took us down the wrong path. A great example of this was long-term care insurance. Many companies rushed products to market, only later realizing that their pricing assumptions were flawed because of larger-than-expected claims. Some had to exit the business. The companies remaining in LTC made major price increases.

Auto insurers run into the same dangers (and more) with untested assumptions. For example, who receives discounts, and who should receive discounts? Recently, a popular auto insurer that was giving discounts to drivers with installed telematics, announced that it would begin increasing premiums on drivers who seemed to have risky driving habits. The company had assumed that those who chose to use telematics would be good drivers and that just having the telematics would cause them to drive more safely. The resulting data, however, proved that some discounts were unwarranted; just because someone was willing to be monitored didn’t mean she was a safe driver.

Now the company is basing pricing on actual data. It has also implemented a new pricing model by testing it in one state before rolling it out broadly – another step in the right direction.

When we either predict outcomes before analyzing the data or we use data improperly, we taint the model we’re trying to build. It’s easy to do. Biases and assumptions can be subtle, creeping silently into otherwise viable formulas.

Let’s say that I’m an auto insurer. Based on an analysis of the universe of auto claims, I decide to give 20% of my U.S. drivers (the ones with the lowest claims) a discount. I’m assuming that our mix of drivers is the same as the mix throughout the universe of drivers. After a year of experience, I find that I am having higher claims than I anticipated. When I apply my claims experience to my portfolio, I find that, actually, only the top 5% are a safe bet for a discount, based on a number of factors. Now I’ve given a discount to 15% more people than ought to have had it. Had I tested the product, I might have found that my top 20% of U.S. drivers were safe drivers but were also driving higher-priced vehicles – those with a generally higher cost per claim. The global experience didn’t match my regional reality.

Predictions based on actual historical experience, such as claims, will always give us a better picture than our “logical” forays into pricing and product development. In some ways, letting data drive your organizations decisions is much like the coming surge of autonomous vehicles. There will be a lot of testing, a little letting go (of the driver’s wheel) and then a wave of creativity surrounding how the vehicle can be used effectively. The result of letting the real data talk will be the profitability and longevity of superior models and a tidal wave of new uses. Decisions based on reality will be worth the wait.

The Science (and Art) of Data, Part 1

Most insurers are inundated with data and have difficulty figuring out what to do with all of it. The key is not just having more data, more number-crunching analysts and more theoretical models, but instead identifying the right data. The best way to do this is via business-savvy analysts who can ask the right strategic questions and develop smart models that combine insights from raw data, behavioral science and unstructured data (from the web, emails, call center recordings, video footage, social media sites, economic reports and so on). In essence, business intelligence needs to transcend data, structure and process and be not just a precise science but also a well-integrated art.

The practitioners of this art are an emerging (and rare) breed: data scientists. A data scientist has extensive and well-integrated insights into human behavior, finance, economics, technology and, of course, sophisticated analytics. As if finding this combination of skills wasn’t difficult enough, a data scientist also needs to have strong communication skills. First and foremost, he must ask the right questions of people and about things to extract the insights that provide leads for where to dig, and then present the resulting insights in a manner that makes sense to a variety of key business audiences. Accordingly, if an organization can find a good data scientist, then it can gain insights that positively shape its strategy and tactics – and gain them more quickly than less-well-prepared competitors.

What it takes to be an effective data scientist

The following table highlights the five key competencies and related skills of a qualified data scientist.

Competencies

Key Skills

Business Impact

1. Business or Domain Expertise

   Deep understanding of:

  • Industry domain, including macro-economic effects and cycles, and key drivers;
  • All aspects of the business (marketing, sales, distribution, operations, pricing, products, finance, risk, etc.).
  • Help determine which questions need answering to make the most appropriate decisions;
  • Effectively articulate insights to help business leadership answer relevant questions in a timely manner.

2. Statistics

  • Expertise in statistical techniques (e.g., regression analysis, cluster analysis and optimization) and the tools and languages used to run the analysis (e.g., SAS or R);
  • Identification and application of relevant statistical techniques for addressing different problems;
  • Mathematical and strategic interpretation of results.
  • Generate insights in such a way that the businesses can clearly understand the quantifiable value;
  • Enable the business to make clear trade-offs between and among choices, with a reasonable view into the most likely outcomes of each.

3. Programming

  • Background in computer science and comfortable in programming in a variety of languages, including Java, Python, C++ or C#;
  • Ability to determine the appropriate software packages or modules to run, and how easily they can be modified.
  • Build a forward-looking perspective on trends, using constantly evolving new computational techniques to solve increasingly complex business problems (e.g., machine learning, natural language processing, graph/social network analysis, neural nets, and simulation modelling);
  • Ability to discern what can be built, bought or obtained free from open source and determine business implications of each.

4. Database Technology Expertise

  Thorough understanding of:

  • External and internal data sources;
  • Data gathering, storing and retrieval methods (Extract-Transform-Load);
  • Accessing data from external sources (through screen scraping and data transfer protocols);
  • Manipulating large big data stores (like Hadoop, Hive, Mahoot and a wide range of emerging big data technologies).
  • Combine the disparate data sources to generate very unique market, industry and customer insights;
  • Understand emerging latent customer needs and provide inputs for high-impact offerings and services;
  • Develop insightful, meaningful connections with customers based on a deep understanding of their needs and wants.

5. Visualization and Communications Expertise

Comfort with visual art and design to:

  • Turn statistical and computational analysis into user-friendly graphs, charts and animation;
  • Create insightful data visualizations (e.g., motion charts, word maps) that highlight trends that may otherwise go unnoticed;
  • Use visual media to deliver key message (e.g., reports, screens – from mobile screens to laptop/desktop screens to HD large visualization walls, interactive programs and, perhaps soon, augmented reality glasses).
  • Enable those who aren’t professional data analysts to effectively interpret data;
  • Engage with senior management by speaking their language and translating data-driven insights into decisions and actions;
  • Develop powerful, convincing messages for key stakeholders that positively influence their course of action.

While it may seem unrealistic to find a single individual with all the skills we've listed, there are some data scientists who do, in fact, fit the profile. They may not be equally skilled in all areas but often have the ability to round out their skills over time. They typically tend to be in high-tech sectors, where they have had the opportunities to develop these abilities as a matter of necessity.

However, because of the increasing demand for data scientists and their scarcity, insurers (and companies in other industries) should consider if they want to build, rent or buy them. Although buying or renting capabilities can be viable options – and do offer the promise of immediate benefits – we believe that building a data science function is the best long-term approach. Moreover, and as we will address in our next post, in light of the shortage of data scientists, a viable approach is creating a data science office of individuals who collectively possess the core competencies of the ideal data scientist.

Tackling Underwriting Profitability Head On

For many years, insurance companies built their reserves by focusing on investment strategies. The recent financial crisis changed that: insurers became incentivized to shift their focus as yields became more unpredictable than ever. As insurance carriers looked to the future, they know that running a profitable underwriting operation is critical to their long term stability.

Profitable underwriting is easier said than done. Insurers already have highly competent teams of underwriters, so the big question becomes, “How do I make my underwriting operation as efficient and profitable as possible without creating massive disruptions with my current processes?”

There are three core challenges that are standing in the way:

  • Lack of Visibility: First, the approach most companies take to data makes it hard to see what's really going on in the market and within your own portfolio. Although you may be familiar with a specific segment of the market, do you really know how well your portfolio is performing against the industry, or how volume and profit tradeoffs are impacting your overall performance? Without the combination of the right data, risk models, and tools, you can’t monitor your portfolio or the market at large, and can't see pockets of pricing inadequacy and redundancy.
  • Current Pricing Approach: You know the agents that underwriters engage with every day want you to give them the right price for the right risk, and it's not easy. In fact, it's nearly impossible. Underwriters are often asked to make decisions based on limited industry data and a limited set of risk characteristics that may or may not be properly weighted. As an underwriter reviews submission after submission, you need to make decisions such as, “How much weight do I assign to each of these risk characteristics (severity, frequency, historical loss ratio, governing class, premium size, etc.)?” Imagine how hard it is to do the mental math on each policy and fully understand how the importance of the class code relates to the importance of the historical loss ratio or any other of the most important variables.
  • Inertia: When executives talk about how to solve these challenges around visibility and pricing, most admit they're concerned about how to overcome corporate inertia and institutional bias. The last thing you want to do is lead a large change initiative and end up alienating your agents, your analysts, and your underwriters. What if you could discover pockets of pricing inadequacy and redundancy currently unknown to you? What if you could free your underwriters to do what they do best? And what if you could start in the way that makes the most sense for your organization?

There's a strong and growing desire to take advantage of new sources of information and modern tools to help underwriters make risk selection and pricing decisions. The implementation of predictive analytics, in particular, is becoming a necessity for carriers to succeed in today's marketplace. According to a recent study by analyst firm Strategy Meets Action, over one-third of insurers are currently investing in predictive analytics and models to mitigate against the problems in the market and equip their underwriters with the necessary predictive tools to ensure accuracy and consistency in pricing and risk selection. Dowling & Partners recently published an in-depth study on predictive analytics and said, “Use of predictive modeling is still in many cases a competitive advantage for insurers that use it, but it is beginning to be a disadvantage for those that don't.” Predictive analytics uses statistical and analytical techniques to develop models that enable accurate predictions about future event outcomes. With the use of predictive analytics, underwriters gain visibility into their portfolio and a deeper understanding of their portfolio's risk quality. Plus, underwriters will get valuable context so they understand what is driving an individual predictive score.

Another crucial capability of predictive modeling is the mining of an abundance of data to identify trends, patterns and relationships. By allowing this technology to synthesize massive amounts of data into actionable information, underwriters can focus on what they do best: they can look at the management or safety program of an insured, anything they think is valuable. This is the artisan piece of underwriting. This is that critical human element that computers will never replace. As soon as executives see how seamless it can be for predictive analytics to be integrated into the underwriting process, the issue of overcoming corporate inertia is oftentimes solved.

Just as insurance leaders are exploring new methods to ensure profitability, underwriters are eager to adopt the analytical advancements that will solve the tough problems carriers are facing today. Expecting underwriters to take on today's challenges using yesterday's tools and yesterday's approach to pricing is no longer sustainable. Predictive analytics offers a better and faster method for underwriters to control their portfolio's performance, effectively managing risk and producing better results for an entire organization.

Predictive Analytics And Underwriting In Workers' Compensation

Insurance executives are grappling with increasing competition, declining return on equity, average combined ratios sitting at 115 percent and rising claims costs. According to a recent report from Moody’s, achieving profitability in workers’ compensation insurance will continue to be a challenge due to low interest rates and the decline in manufacturing and construction employment, which makes up 40% of workers’ comp premium.

Insurers are also facing significant changes to how they run underwriting. The industry is affected more than most by the aging baby boomer population. In the last 10 years, the number of insurance workers 55 or older has increased by 74 percent, compared to the 45 percent increase for the overall workforce. With 20 percent of the underwriter workforce nearing retirement, McKinsey noted in a May 2010 Report that we will need 25,000 new underwriters by 2014. Where will the new underwriters come from? And more importantly, what will be the impact on underwriting accuracy?

Furthermore, there’s no question that technology has fundamentally changed the pace of business. Consider the example of FirstComp reported by The Motley Fool in May 2011. FirstComp created an online interface for agents to request workers’ compensation quotes. What they found was remarkable. When they provided a quote within one minute of the agent’s request, they booked that policy 52% of the time. However, their success percentage declined with each passing hour that they waited. In fact, if FirstComp waited a full 24 hours to respond, their close rate plummeted to 30 percent. In October 2012, Zurich North America was nominated for the Novarica Research Council Impact Award for reducing the time it takes to quote policies. In one example, Zurich cut the time it took to quote a 110-vehicle fleet from 8 hours to 15 minutes.

In order to improve their companies’ performance and meet response time expectations from agents, underwriters need advanced tools and methodologies that provide access to information in real-time. More data is available to underwriters, but they need a way to synthesize “big data” to make accurate decisions more quickly. When you combine the impending workforce turnover with the need to produce quotes within minutes, workers’ comp carriers are increasingly turning toward the use of advanced data and predictive analytics.

Added to these new industry dynamics is the reality that both workers’ compensation and homeowners are highly unprofitable for carriers. According to Insurance Information Institute’s 2012 Workers’ Compensation Critical Issues and Outlook Report, profitable underwriting was the norm prior to the 1980s. Workers’ comp has not consistently made an underwriting profit for the last few decades for several reasons including increasing medical costs, high unemployment and soft market pressures.

What Is Predictive Analytics?
Predictive analytics uses statistical and analytical techniques to develop predictive models that enable accurate predictions about future outcomes. Predictive models can take various forms, with most models generating a score that indicates the likelihood a given future scenario will occur. For instance, a predictive model can identify the probability that a policy will have a claim. Predictive analytics enables powerful, and sometimes counterintuitive, relationships among data variables to emerge that otherwise may not be readily apparent, thus improving a carrier’s ability to predict the future outcome of a policy.

Predictive modeling has also led to the advent of robust workers’ compensation “industry risk models” — models built on contributory databases of carrier data that perform very well across multiple carrier book profiles.

There are several best practices that enable carriers to benefit from predictive analytics. Large datasets are required to build accurate predictive models and to avoid selection bias, and most carriers need to leverage third party data and analytical resources. Predictive models allow carriers to make data-driven decisions consistently across their underwriting staff, and use evidenced-based decision making rather than relying solely on heuristics or human judgment to assess risk.

Finally, incorporating predictive analytics requires an evolution in terms of people, process, and technology, and thus executive level support is important to facilitate adoption internally. Carriers who fully adopt predictive analytics are more competitive in gaining profitable market share and avoiding adverse selection.

Is Your Organization Ready For Predictive Analytics?
As with any new initiative, how predictive analytics is implemented will determine its success. Evidence-based decision-making provides consistency and improved accuracy in selecting and pricing risk in workers’ compensation. Recently, Dowling & Partners Securities, LLC, released a special report on predictive analytics and said that the “use of predictive modeling is still in many cases a competitive advantage for insurers that use it, but it is beginning to be a disadvantage for those that don’t.” The question for many insurance executives remains: Is this right for my organization and what do we need to do use analytics successfully?

There are a few important criteria and best practices to consider when implementing predictive analytics to help drive underwriting profitability.

  • Define your organization’s distinct capability as it relates to implementing predictive analytics within underwriting.
  • Secure senior management commitment and passion for becoming an analytic competitor, and keep that level of commitment for the long term. It will be a trial and error process, especially in the beginning.
  • Dream big. Organizations that find the greatest success with analytics have big, important goals tied to core metrics for the performance of their business.