Data science can help insurers assess risk more accurately, price policies fairly, and move away from historically inequitable underwriting data.
Insurers have always used data like age, location, and driving record to write insurance policies, but the results haven’t always been precise, accurate, or equitable. Now, insurance industry entrepreneurs are developing new models that are more customized and less dependent on flawed and implicitly biased older data categories.
Root Inc. is a Columbus, Ohio-based auto insurance company that’s using telematics to collect real-world policyholder driving data for more accurate risk assessment and pricing. Root is also a Data Science for All (DS4A) Employer Partner, which means they contribute to DS4A scholarship and mentorship opportunities — and they get to connect with potential candidates while they’re training for their data careers.
Annie Yang, Root’s Director of Data Science, spoke with us recently about her company’s work, the importance of sourcing and developing diverse data talent and her perspective on building a career in a mostly male field. Annie oversees Root’s pricing and underwriting data science teams, who “build statistical and machine learning models that power [their] pricing and underwriting algorithms.”
Because people can easily get multiple quotes online to compare prices and coverage before they choose a policy, “the better an insurer can match price to risk, the more profitable the insurer will be. Root’s algorithms use many indicators instead of the traditional demographic variables like age or credit. Root combines variables like age or credit with more individual data allowing Root to assign risk in a way that companies using traditional data sources cannot.”
Becoming a Data Leader as a Woman in a Male-Dominated Field
Despite her fast rise to the director level after five years in data science, Annie says she got into the insurance industry “kind of serendipitously” after earning her bachelor’s and master’s degrees in mathematics. “I was looking for something that was quantitative but also in industry.”
After roles in claims and pricing analysis for other insurers, Annie joined Root in late 2019. “Initially I led a team that was responsible for developing new features for pricing, and I grew myself over the years to lead our entire pricing data science function.”
Like many women in STEM fields, Annie has often found herself among a small group of female peers or even as the only woman in a group. “Studying math in undergrad and during my masters, the balance of men and women was very skewed toward men. At my first two companies, the balance was about even. Then I was one of about four women on a team of 30 or so. It's hard for me to pinpoint having ever been the subject of discrimination, but at the same time, it's something that I never really forget about myself.”
“I certainly suppressed parts of my personality at work, particularly some of the more ‘girly’ parts of my personality. I don't think that suppression affects my performance, but it might add a bit of cognitive load, always having this filter in the back of my mind.”
Attracting More Diverse Data Talent
Rather than accept the status quo, Annie has “advocated for hiring practices that can mitigate this imbalance—for example, sourcing talent from historically Black colleges, expanding the top of the funnel to a more diverse talent pool, and creating a program that trains people from non-data backgrounds to become data scientists.”
For other data science leaders seeking to diversify their teams and attract more talent, Annie recommends “starting at the top of the talent funnel. Sponsoring programs like DS4A / Women can help employers build relationships with new talent while they train for data-related roles.”
She also suggests “having opportunities in place to hire people who might not meet every single requirement immediately, but who can be trained quickly tostart adding value. A formal mentorship program can be helpful, too.” Informal ongoing learning has been valuable at Root. Monthly lunch-and-learns where someone presents on a recent project or new technique, reading groups, and book clubs can help interested team members pick up new ideas and build their strengths.
More Perspectives, More Ways to Approach Problems
A more diverse team delivers innovation and recruitment benefits, too, Annie said. “Diversity in terms of educational or occupational background can bring new ideas to the table. And a diverse team allows you to attract top talent that wants to work in that kind of diverse creative environment.”
Different perspectives also allow analysts to more easily understand the implicit biases in some datasets and rethink how and when to use certain types of data. “I won't make the argument that, for instance, having more women on the team would allow us to price more fairly based on gender. But there are certainly biases in the way that insurance data is collected and potentially used.”
One example is the use of credit scores to set premiums, despite a long history of credit discrimination against Black, Hispanic / Latinx, and people from lower income levels. “Credit can help segment risk, but ultimately, we'd love to get to where we don't have to use it. We're lobbying very hard to make credit a banned rating element. Relying more on causal factors can go a long way in making our pricing fairer.”
Owning and Solving Problems with Data
Root’s approach to addressing a problem like credit score data aligns with Annie’s advice for women seeking data careers: Adopt an ownership mentality.
“As a data scientist, you're not just there to collect data and build a model. You also have to solve a business problem through that modeling. Engaging with stakeholders, identifying the right problems to solve, deploying your model correctly, monitoring the results, and demonstrating that you improved some metric in the business — all of that is ownership.”
Developing that mindset, she said, makes the difference between doing the job adequately and growing a career by making business impacts through data.
Curious to learn how your company and its data leaders — like Root Inc. and Annie Yang — can help nurture and diversify data talent? Keep reading.