You might say that auto insurance is a crystal-ball industry, a business that relies heavily on the ability to make predictions about risk and accidents. Insurers constantly are trying to improve their predictions.
In that quest, Allstate in 2011 became the first insurer to tap the power of "crowdsourcing" to polish its crystal ball. Allstate's move could shape the future of the industry and could provide consumers with more accurately priced (and, perhaps, cheaper) auto insurance premiums.
What is crowdsourcing?
Crowdsourcing may sound complex, but the idea is rather simple. According to Crowdsourcing.org, the term refers to outsourcing a task that otherwise would be handled in house by a group of people or an online community. In other words, if a company has a problem, it looks to a "crowd" for solutions.
For example, Netflix uses crowdsourcing to provide subscribers with movie recommendations based on what others with similar tastes (those in the “crowd”) have watched.
In Allstate’s case, the problem was coming up with an even better model for determining insurance premiums, which are based on a variety of factors, including a driver’s age, geography and vehicle type. To find a solution, Allstate launched the Claim Prediction Challenge in July 2011.
The contest was run by Kaggle, a San Francisco company that specializes in statistical analysis and predictive modeling competitions. Through Kaggle, Allstate made available information on drivers’ injury and vehicle claims from 2005 through 2007. Then the analysts who were part of Kaggle’s “crowd” were given the challenge of looking for connections between these claims and the types of vehicles driven to accurately predict 2009 claims.
What made this contest unique was that it focused entirely on the link between injuries and vehicle characteristics, such as car length, horsepower and engine cylinders. Essentially, Allstate wanted to know how these factors might determine the likelihood that their policyholders would be prone to an injury in an accident.
When all was said and done, Allstate got a staggering result. The first-place winner, Matthew Carle, an actuarial consultant in Australia, developed a solution that was 340 percent more accurate than Allstate's existing methods. He was awarded $6,000 for his efforts.
“I haven’t heard of any other insurance company doing anything like this, and frankly it strikes me as ingenious,” says Bob Detlefsen, research director at the National Association of Mutual Insurance Companies, a trade group. “Someone basically developed a predictive model for Allstate for a fraction of what it would have cost the company had they done this internally. I think they’ve broken new ground with this one.”
Indeed, no other American insurer has tried this before, and the success of Allstate’s competition has led many in the industry to wonder: Is crowdsourcing the wave of the future for determining rates?
The pros of crowdsourcing
The biggest cost for any insurer is claims -- easily running into the hundreds of millions a year for major companies -- and insurers always are trying to pinpoint better ways of predicting the frequency and severity of claims. John Espenschied, owner of CaliforniaLife.com, a division of Insurance Brokers Group LLC, says the use of crowdsourcing to achieve this is promising.
“Crowdsourcing and statistical analysis of the facts, in my opinion, give a lot of credibility to rates,” Espenschied says. “There’s still no magic pill to make these predictions 100 percent accurate, but I think Allstate’s idea of using crowdsourcing is a great thing. Anytime a company can get a better handle on this, it ultimately helps control costs, which means better rates for consumers.”
According to several experts and analysts, crowdsourcing is only going to become a more popular tool for insurers moving forward. Here are three reasons why:
1. Crowdsourcing taps into a pool of experts who otherwise might not be known to an insurer.
“There are many experts outside of an insurance company’s staff who can help with untapped expertise,” says Dan Weedin, an insurance and risk management consultant in Seattle. “It’s going to be a broader pool than what the company has in house, which means more interesting solutions to a problem.”
2. Crowdsourcing is relatively cheap.
“This contest was incredibly affordable for Allstate, and I imagine their return on investment was very good,” says Joseph Turian, a statistical modeling expert who founded consulting firm Metaoptimize.com. “If you consider the total number of man-hours that were spent on this project, it was much higher than what Allstate eventually paid for.”
3. Crowdsourcing could result in lower premiums.
“Trying to find the bad apples in a pool of insured consumers ultimately helps us all,” Espenschied says. “You spread out the risk, and that makes insurance even cheaper for those who are lower-risk drivers.”
Because it's still in its infancy, the use of crowdsourcing by insurance companies doesn’t come without potential snags. For example, Andrew Schrage, insurance and real estate expert at MoneyCrashers.com, says insurers need to make sure they're offering rewards that are substantial enough to merit a serious effort on the part of the crowd.
“If you’re not willing to offer up a fair reward to the winner, you run the risk of receiving inaccurate data,” Schrage says. “This can have a huge effect on a massive number of consumers, especially when large businesses are using crowdsourcing.”
Also, using the crowd as a source for problem solving means publicizing critical -- and sometimes sensitive -- insurance data. Turian says some insurers may be reluctant to release this data for fear that their rivals would pounce on it.
“The results of this Allstate competition are now public, so competitors can now use the same results without paying anything,” Turian says. “The public nature of crowdsourcing could be its greatest drawback.”
Indeed, Detlefsen says crowdsourcing probably will be used in a limited way by insurers.
“I don’t imagine large established companies will attempt very often to solve problems or develop new technologies based on a public contest that takes place over the Internet,” Detlefsen says. “It was very clever and ingenious of Allstate to use the crowd in this instance, and you’ll see more of it. But I would be surprised to find this as a standard business practice across the board.”