Data Science in Insurance
Data science in insurance has come a long way from the days of punch cards and floppy disks to the modern data collection capabilities of analyzing thousands of gigabytes per second. Paired with the right software, insurers can use advanced data science to score risk in real-time, increase profit, and improve pricing models.
How is Data Science Impacting Insurance?
Before we had access to data science powered by AI, providers relied on documents stored in dusty filing cabinets (remember those?) and manual risk assessment. While those tactics have improved over time, they’re still ridden with errors from manual labor and unseen bias. And even with tremendous advancements in data science and underwriting capabilities, we still have multi-billion dollar challenges, such as fraud, in the industry.
In today’s article, we’re diving into how data science impacts the insurance industry. Here’s what you’ll find:
- Enables accurate risk scoring
- Combining clickstream data with data science to understand and predict the customer’s behavior
- Improves pricing models and profit ratios
- Allows for dynamic friction during the application and claims process
But it doesn’t stop there. Data science can cross-pollinate and optimize all departments in the insurance industry. For example, intuitive digital experiences can dramatically improve the customer experience. When customers are happier, there are fewer drop-offs, more satisfied customers, and more profit.
Risk Scoring with Data Analytics
It’s hard to imagine how underwriters assessed risk without the tools available to us today. What’s even more impressive is how profitable these carriers have been for tens, if not hundreds of years. The landscape has begun to change as extremely well-funded InsurTech startups sacrifice profits for growth. We’ve discussed the tradeoff between user experience and risk measures before – as one goes up, the other goes down. This inverse correlation creates friction between the Marketing and Risk teams. To try to ease the tension, carriers have adopted many different tools and datasets that fall in two categories – “pre-submit” solutions and “post-submit” solutions.
Pre-submit solutions are marketing and customer experience-focused to help you determine how to optimize campaigns. They analyze user behavior and user flows before a user actually submits the quote/application.
Post-submit solutions are third-party datasets like health records, DMV checks, and credit reports to help you validate customer-submitted answers and determine the risk profile of an applicant after they have submitted their quote/application.
But this leaves a critical gap during the application flow — what happens during the application? The solution lies with real-time data.
Real-time data allows you to make real-time decisions. This is crucial for the digital experiences of tomorrow, so gaining access to real-time data is a must. The challenge is getting real-time data. And if you can find it, what do you do with it? Whether carriers like it or not, and whether their business models will prove to be sustainable in the long term, InsurTech companies have succeeded in “disrupting” digital user experiences. They’ve taken full advantage of the on-demand culture we live in and made instant underwriting the new standard.
So what do you do now that you don’t have days or weeks to assess applications and, instead, need to make underwriting decisions on the spot? The answer lies in understanding your user’s intent, in real-time.
Clickstream Data Collection – where is it happening?
Data scientists today have access to very basic clickstream data but it is almost always data based on website behavior and very basic event data. Things like session replay, heat maps, and basic funnel metrics are the norm and none of them are built for the real-time nature of future digital experiences.
The two main areas of opportunity we believe for additional clickstream and <digital body language> data are on applications and claims forms.
For instance, with companies like ForMotiv, you can use incredibly advanced, real-time clickstream data that analyzes how a user fills out a digital application. And the use cases for this go way beyond just marketing and funnel optimization. With real-time behavioral analysis and intent scoring now possible it means assessing risk in real-time and actually being able to do something about it at the moment.
The same thing can be said for the claims process. Having real-time clickstream data and behavioral analysis while an online claim is being filled out gives much deeper insight into the validity of a user-submitted claim. Users have a tendency to misrepresent or outright lie on their answers when it impacts the money they will receive as a rate or a claims payout. Real-time behavioral analysis helps solve this.
Improved Profit Ratio and Pricing Models
The traditional business model in insurance is built on the ability to invest the “float” when premiums are received from clients but not paid out. The very nature of this business model is where the conflict of interest lies: every dollar an insurance carrier does NOT payout, they get to keep as profit.
In other words, the more claims they reject, the more money they make, often resulting in lots of paperwork, scrutiny, policy manipulation, and delayed payouts for claims.
Remember hearing about Lemonade’s cutting-edge business model? It was big news in the industry when they ditched the traditional business model to instead build a business designed to appeal to millennials. So that means no paper, no phone calls, and no salespeople. Without traditional risk predictions based on face-to-face interactions, Lemonade turned to the data science created by AI and machine learning.
As a result, they could access 100X the amount of digital user data points to create predictive risk models with over 92% accuracy. Now with 400,000+ customers, about $840 million in funding, and over four billion public valuations, their digital-first strategy based on digital sciences is paying off (even if their stock price isn’t…)
Now whether Lemonade will survive long term is up for debate, but they certainly succeeded in changing customer preferences and bringing the on-demand experience into the insurance market. There is no going back now and legacy carriers will need to invest heavily in improving their digital user experiences. The challenge is that today’s datasets are not built for the real-time experiences of tomorrow, but they need to be to keep up.
ForMotiv provides data scientists with an industry-leading behavioral dataset built for real-time predictive models. Our glass-box behavioral science platform allows data scientists to quickly integrate deterministic signals or intent scores into their internal predictive models.
Application Friction – To increase or not to increase, that is the question?
A lot of digital marketers want to use more advanced clickstream data like behavioral intelligence to reduce friction during the customer experience because fewer obstacles lead to faster race times, right?
Until recently, however, this was not possible. With recent advances in behavioral analytics, companies like ForMotiv are enabling carriers to dynamically increase or remove friction. But why increase friction? Simple, because using positive friction helps prevent fraud. When they recognize that a user’s behavior is indicating high-risk or fraudulent activity, carriers can triage those applicants and further qualify them before approving a policy.
There needs to be a balance between a seamless user experience and a well-protected application, and ForMotiv is helping carriers thread that needle.
Dynamically Add Application Friction for Fraud Prevention
Basic application friction is — well, pretty basic. This could look like 2-step authentication factors, checking a box that a bot can’t, or proving that you’re human by identifying all the boxes with traffic lights. If you’ve spent any time on the internet, chances are you’ve encountered these methods.
Using behavioral data science to predict fraud is a bit more involved. It requires advanced machine learning and predictive behavioral analytics which ultimately produces dynamic application friction.
ForMotiv, for example, analyzes a user’s behavior and predicts whether it’s risky or fraudulent while they’re filling out the application —- not just in hindsight. Did the user manipulate their tobacco or medical questions? Did they get a quote, go back and edit important information, and check the quote again? Did they copy/paste or correct sensitive questions too many times? Once they spot the high-risk behavior, insurers can dynamically add or remove friction tailored to the individual user.
Another example is when an agent is acting in their own best interest instead of that of the company. We see countless examples of agents changing answers into already submitted customer applications, ghost broking, and pushing through bad policies that will quickly churn to hit their sales quotas. In these cases, real-time behavior analytics can flag the user as high risk and notify the risk department to intervene.
As carriers move towards consumer complete and accelerated underwriting, having the ability to intuitively triage a risky application can make the difference between a profitable quarter or one in the red.
The Perfect Time for Reducing Application Friction
The golden time for reducing application friction is when a user needs help. Think about it – if a customer was sitting in front of you and you recognized that they were confused, you wouldn’t just sit there and watch them struggle, get frustrated, and walk out the door only to go back and watch the game tape and figure out an A/B test to try tomorrow, would you? No, because that would be preposterous. Yet that’s what carriers do millions of times a day with their digital applicants and they don’t even know it. Online experiences are the new norm the same way hailing a Taxi on a crowded NYC street was the norm, until it wasn’t.
ForMotiv’s mission is to fix this.
If a user is stumbling through an application, their online behavior will likely show it. This is the time to dynamically reduce friction, like a chatbot popping up asking how they can assist, or driving them into a call center.
Looking for data science in insurance? Let us focus on the behavioral datasets, so you don’t have to. See ForMotiv in action.