Tag Archives: data source

Eating the Big Data Elephant

How do you eat an elephant? One bite at a time.

What an old joke with a great premise. No matter how big the task, taking things one bite at a time makes any daunting task seem easier to swallow.

Take the big data challenge. By and large, insurance companies and traditional businesses are used to relying on paper files, mailrooms, fax machines and call centers as incoming data streams. Designed to handle internal data collected from limited sources, the systems showed their first hint of trouble with an inability to incorporate emails and SMS text messages into policyholder and claim files. Inefficiently integrated best-of-breed IT environments further complicated the issue by putting data in silos and restricting access to users.

Today, integration of systems has improved, and the move toward suites has enabled additional collaboration and data sharing benefits. However, big data, marked by its volume, velocity and variety, still has insurers stymied. And the move toward omni-channel distribution, the Internet of Things (IoT) and the connected world has amplified the need for insurers to incorporate even more data streams (both internal and external) into the risk assessment process. Cue the analytics software and reporting solutions, neither of which alone will make a legacy system more able to digest information from new data sources for rating and underwriting purposes.

Meanwhile, the big data behemoth is growing into the proverbial elephant in the room. The problem is no longer just Incorporating this data; analyzing it and acting on it are equally incomprehensible.

Buying data from traditional data sources –including motor vehicle reports (MVRs), historical flood data and credit reports on the property and casualty (P&C) side or health and medical records or test results on the life and health side is expensive. Furthermore, traditional data sources don’t allow insurers to pick and choose what may be most useful based on line of business, let alone product or policy type, geographic area or purchasing preferences.

Alternative data sources such as social data exist, but the unstructured nature of the information makes it especially difficult for insurers to internalize. Consider that today’s consumers, who are both existing and potential new policyholders, are creating mountains of data that could contribute to better risk decision making, but right now that data doesn’t make it to the underwriter’s desk. Social data is a silver bullet that can provide a predictive enhancement layer for traditional data sources, leading to more accurate underwriting and making insurers better able to select the best risks.

By breaking the traditional data collection and utilization mold as it relates to risk assessment, insurers can integrate social data with core administration systems, making unstructured social data both accessible and actionable across all industry segments and lines of business. By capitalizing on the explosion of social data as a resource for better insurance risk assessment, insurers can improve underwriting, streamline the claims investigation process, decrease loss costs and potentially make insurance relevant to a whole new generation of insurance consumer.

The scope of the big data problem is just dawning on insurers. In an effort to not bite off more than can be chewed at one time, insurers can start to consume and absorb big data by incorporating social data into rating and underwriting. But keep in mind that social data is just the first bite of a very important meal.

3 Game Changers — and How to Survive

The follow-the-leader principle works on a trail that has proven to be relatively safe from perils and predators. However, when new frontiers are breached, a new kind of leadership is required for survival.

Insurers have generally been able to just follow the leader for ages, but now a new frontier has been breached. The insurance industry is vulnerable to three game changers that consumers are eager to embrace.

Drawing on remarks I made recently at a keynote for the National Association of Mutual Insurance Companies Annual Conference, here are the game changers:

The first big disrupter is data collection. Insurance is built on the principle of using accurate data and statistics to build underwriting financial models that serve to predict behavior and events from an actuarial or probability standpoint. London’s Edward Lloyd figured this out when he opened his coffee shop in 1688, and people started selling insurance to merchants and ship owners. His motto was fidentia, Latin for confidence. We now refer to “confidence factors” when estimating future losses.

Insurers have been notorious for using forms to collect data. But, today, a person is subjected to more new information in one day than a person in the Middle Ages saw in his entire life. If modern competitors to the insurance industry can obtain more accurate data in a faster and more in-depth manner, they may beat insurers at their own game.

With cloud computing and its infinite data storage/retrieval capability, trillions of bits of information relating to insureds are available. Data sources track things like profile patterns, such as personal Internet searches or satellite surveillance data. Relevant data can be mined and analyzed to build a risk model for every insurable consumer or business peril from property and vehicle insurance to earthquake and weather insurance.

The five biggest data collectors on the planet are Google, Apple, Facebook, Yahoo and Amazon. These high-tech companies have the ability, financial resources and potential desire to foray into the insurance industry. Keep in mind that in 2014 the world’s top 10 insurers received $1.2 trillion in revenue, yet surveys have shown that people around the world have grown to use and trust the products and services provided by the five biggest data collectors.

Accessibility and familiarity are allowing profitable new brands to replace old brands. Consumers also prefer and use third-party validation and independent comparisons found on websites.

What does this spell for the insurance industry? Sadly, consumers have grown more uncomfortable with reliance on and interaction with agent relationships. John Maynard Keynes once said: “The difficulty lies not so much in developing new ideas, as in escaping from old ones.”

The second emerging threat to insurance is botsourcing — the replacement of human jobs by robotics. The robots haven’t just hatched in agriculture or auto assembly plants — they’re expanding in a variety of skills, moving up the corporate ladder, showing awesome productivity and retention rates and increasingly shoving aside their human counterparts.

Google won a patent recently to start building worker robots with personalities. Move over, Siri.

Author and entrepreneur Martin Ford, in his book Rise of the Robots, argues that artificial intelligence (AI) and robotics will soon overhaul our economy. Increasingly, machines will be able to take care of themselves, and fewer jobs will be necessary.

Reassessment of the way we employ our workforce is essential to cope with this new industrial revolution. The lucrative insurance realm of personal and product liability insurance lines and workers’ comp is being tempered as human risk factors — especially in high-risk areas — give way to robotics. The saying goes: “Management is doing things right, but leadership is doing the right things.”

How will the insurance industry react to the accelerating technology of bot-sourcing?

The third emerging threat to the insurance industry that has received enormous attention this past year autonomous vehicles. More than a half-dozen carmakers, as well as Google and Uber, predict that self-driving vehicles will be commonplace on our roads between 2017 and 2020. Tesla Motors CEO and general future-tech proponent Elon Musk has predicted that human drivers could someday be outlawed. Humans cannot outperform an autonomous vehicle, which can assess and react to more than 7,000 driving threats per second. There are no incidents of driver impairment, reckless driving, DUIs, road rage, driver texting, speeding or inattention.

With a plethora of electronic distractions, increased safety can only be achieved when human drivers are removed from the equation. Automakers have employed an incremental approach to safety in their current models. These new technologies are clever and helpful but do not remove the risks. There’s a phenomenon called the Peltzman Effect, based on research from an economist at the University of Chicago who studied auto accidents. He found that, when you introduce more safety features like seatbelts into cars, the number of fatalities and injuries doesn’t drop. The reason is that people compensate for it. When you have a safety net in place, people will naturally take more risks. Today, 35,000 vehicle occupants die in the U.S. because of auto accidents. Autonomous vehicles are expected to cut auto-related deaths and injuries by 80% or more.

One of the biggest revenue sources to insurers is vehicle insurance. As autonomous vehicles take over our roads and highways, you need to address all the numerous unanswered questions relating to the risk playing field. Who will own the vehicles? How can you assess the potential liability of software failure or cyberattacks? Will insurers still have a role? Where will legal liabilities fall? Who will lead the call to sort these issues out?

Clearly, the lucrative auto insurance market will change drastically. Insurance and reinsurance company leadership will be an essential ingredient to address this disruptive technology.

As I told the conference: Count on Insurance Thought Leadership to play a significant role in addressing these and other disruptive technologies facing the insurance industry. A Chinese proverb says: “Not the cry, but the flight of a wild duck, leads the flock to fly and follow.”

A Secret for Comparing Workers’ Comp Costs

Workers’ compensation claims and medical managers are continually challenged by upper management to analyze their drivers of workers’ comp costs. Moreover, upper management wants comparisons of the organization’s results to that of peers.

The request is appropriate. Costs of doing business directly affect the competitive performance of the organization. Understanding drivers of workers’ comp costs is key to making adjustments to improve performance. Still, it’s not that simple.

Executing the analysis is the lesser of the two demands. More challenging is finding industry or peer data that is similar enough to create an apples-to-apples study. In a recent article, Nick Parillo states, “Regardless of the data source, whether it be peer-related or insurance industry-related, risk managers must be focused on aligning the data to their respective company and its operations.” Parillo emphasizes that the data should be meaningful and relevant to the organization.

Aligning the data to the situation can be challenging. Industry or peer data may not be situation-specific enough or granular enough to elicit accurate and illuminating information. State regulations vary, as do business products and practices, along with a multitude of other conditions that make truly accurate comparisons difficult.

Variability in the data available for benchmarking can be especially disconcerting when considering medical cost drivers, which now account for the majority of claim costs. Differences in state fee schedules and legislation such as required utilization review (UR) and the use of evidence-based guidelines can produce questionable comparative results. Additionally, whether the contributed data is from self-insured or self-administrated entities can skew the results.

Other variables that make comparing industry or peer data less valid are unionization, physical distribution of employees, employee age and gender, as well as industry type and local resources available. Potential differences are unlimited.

External sources such as local cultural and professional mores, particularly among treating medical providers, can play a significant role in disqualifying data for comparison. For instance, my company’s analysis of client data has uncovered consistent differences in medical practice patterns in one large state. In one geographic sector, referrals to orthopedists with subsequent surgery and higher costs are far more frequent than in another sector of the state for the same type of injury.

Parillo continues, “Given the uncertainty and limitations on the kinds of peer group data a risk manager would need to perform a truly “apples to apples” comparison, the most “relevant and meaningful” data may be that which a risk manager already possesses: His own.”

Analyzing internal data can be highly productive. First, the conditions of meaningful and relevant are guaranteed, for obvious reasons. The geographical differential across one state was found in one organization’s internal data, which ensures that data variability is not a factor.

Analyses can be designed that dissect the data at hand. Follow up to the above example might include looking for other geographic variables in costs, in injury types and in medical practice patterns. Compare physician performance for specific injury types in the same jurisdiction and then look for differences within. To gain this kind of specificity and relevance, drill down for other indicators.

Evaluate how costs move. Look at costs at intervals along the course of claims for specific injury types. In this case, utilizing ICD-9s is more informative than the National Council on Compensation Insurance (NCCI) injury descriptors. One client found that injury claims that contained a mental health ICD-9 showed a surge in costs beginning the second year. Now, further analysis can begin to discern earlier indicators of this outcome. In other words, dive further into the data to find leading indicators.

Industry data is not likely to contain the detail necessary to evoke subtle mental health information during the course of the claim. Most analysis ignores the subtlety and sequence of diagnoses assigned. Few would uncover the mental health ICD-9 because few bother with ICD-9s at all.

Drilling down, analyze claims that fall into this category for prescriptions, legal involvement and other factors that might divulge prophetic signs. It is an investigative trail that relies on finite internal data analysis.

Too often people disrespect their own data, thinking it is too poor in quality, therefore of little value. It’s true, much of the data collected over the years is of poorer quality, but it still has value. Begin by cleaning or enhancing the data and removing duplicates. Going forward, management emphasis should be on collecting accurate data.

Benchmarking data sourced from the industry may be useful but should not necessarily be considered the most accurate or productive approach. Internal data analysis may be the best opportunity for discovering cost drivers.