By Nancy Mann Jackson
As consumers generate vast volumes of behavioral data—from wearable devices to online interactions—insurance companies are reimagining risk assessment. Actuaries face both opportunities and challenges in leveraging this information to create fair, accurate, and ethically sound models.
Lliving in a connected world means consumers generate data constantly—when they make purchases, browse websites, use social media platforms, fill out online forms and surveys, and interact with services through smart devices and apps. These activities produce a vast amount of data, including purchase history, online behaviors, demographic information, and personal preferences.
Companies collect and use that consumer data for a range of purposes, from personalizing experiences to delivering targeted advertising. In the insurance industry, this behavioral data can play an even greater role. Some insurers are using consumer behavior data in areas such as risk assessment, pricing, fraud detection, and customer engagement.
“Actuarial practice has always depended on understanding the relationship between expected and observed outcomes,” says Doron Samuell, an Australia-based behavioral economist and recipient of the Academy’s 2024 Award for Research. “Insurers have historically relied on operational data gathered at underwriting and at the time of claim to make these comparisons, supported by product-specific sources of information. Increasingly, however, consumer behavior data is changing the boundaries of what can be measured and predicted.”
As consumer behavior data proliferates and becomes increasingly available, it presents both an opportunity and a challenge for actuaries, Samuell says. Actuaries have the opportunity to better align expected and actual outcomes, while also facing the challenge of accounting for new sources of variability that traditional data does not capture.
Risk Modeling
Many insurance companies have embraced consumer behavior data for use in risk assessments. Access to this data allows risk models to incorporate more individualized and dynamic indicators. For example, health and life insurers can analyze fitness app usage or wearable data to gain insights into lifestyle and health management. In auto insurance, telematics can provide a detailed view of driving behaviors, including speed, braking patterns, and time spent behind the wheel.
While these types of data can be informative, they can also introduce bias. For example, wearable technology, such as smartwatches used to track heart rate and rhythm, may provide less reliable data when worn by people with darker skin tones, according to the 2022 study, The Effect of Skin Tone on Accuracy of Heart Rate Measurement in Wearable Devices: A Systematic Review.
Applications for Consumer Behavior Data
Risk assessment isn’t the only area of insurance where consumer behavior data is having—or could have—an impact. Here’s a look at other areas that may be affected by increasingly available data.
Pricing. Insurance companies can use customer data to provide personalized pricing. For example, auto insurers may offer premiums based on how a person drives and how much they drive. Health insurers can offer discounts to policyholders who exhibit healthy lifestyle behaviors, such as walking a certain number of steps per day, completing health screenings, or visiting the gym regularly.
Fraud detection. Just as policyholders can be classified based on specific traits for risk modeling, insurers can use alternative data sources to detect patterns that help identify fraudulent claims faster, says Andrew Larocque, health actuary at Risk & Regulatory Consulting , a 2025 Rising Actuary Award recipient, and vice chairperson of the Academy’s Behavioral Economics Work Group.
“Combining behavioral indicators with plausibility estimation systems can allow insurers to stratify customers into different risk categories,” says Doron Samuell, Australia-based behavioral economist and recipient of the Academy’s 2024 Award for Research. “This [approach] is not about deterministic judgments, but about probabilistic resource allocation. Valid claims could be processed more quickly and at lower cost, while those with higher-risk profiles receive additional scrutiny. Without such approaches, honest customers end up subsidizing dishonest ones, which raises both ethical and financial concerns.”
Keep in mind that incorporating behavioral factors into risk models does not guarantee certainty. A customer whom the model predicts to be at high risk of dishonesty may actually act honestly. “Ethical insurers must avoid treating predictive categories as definitive labels,” Samuell says. “Operational guidelines should ensure that all customers are treated with fairness and dignity, even as insurers make better use of available data.”
For example, a group of actuaries and data scientists from Travelers developed a graphical database known as the organized fraud detector to speed up the process of identifying potential insurance fraud enabled by complex social networks. They use the graphical database to identify connections between claims, medical providers, lawyers, and insurers involved in a given fraud case, doing so up to 1,000 times faster than with a relational database. The team layers these complicated networks with a neural network model that uses deep learning algorithms to sift through the data and uncover any connections or participants that appear suspicious.
Insurers can also use telematics and connected devices to collect real-time data on driving behaviors, property conditions, and environmental factors to help them better detect fraudulent claims related to staged accidents, property damage, or theft.
Customer engagement. Behavioral analytics is also changing how insurers engage with their customers. For example, by understanding how policyholders interact with digital platforms—including how often they log in, which features they use, and where they disengage—insurers can identify friction points and design more intuitive, personalized services.
To effectively target people who may have a greater need for services, Wildflower Health uses the national Area Deprivation Index, a scientifically validated measure of neighborhood disadvantage that can be used to evaluate and improve factors affecting health across populations. “The index tells us if someone lives in a particular ZIP code where they may be more susceptible to poverty or other social determinants of health,” says Sara Teppema, chief actuary at Wildflower Health and vice chairperson of the Academy’s Health Equity Committee. “If so, we conduct more aggressive outreach.”
Similar consumer data can be helpful for Medicaid and Medicare populations, and insurers that offer Medicare Advantage plans, Teppema says. Other types of consumer data, such as that generated by home sensors or wearables, help detect changes in the habits or movements of older adults, alerting insurers when to engage with policyholders, she says.
Product design. Tracking consumer behavior can also play a role in insurance product design. “If we really think about how consumers will behave down the line, we might design policies differently,” says Randall Stevenson, president and consulting actuary, Hause Actuarial Solutions, Inc. “For example, combining long-term care with a life insurance policy could allow the policyholder to get a monthly advance on their death benefit if they enter a nursing home. They may be less likely to cancel the policy later in life because they may need the long-term care benefits.” Stevenson is chairperson of the Behavioral Economics Work Group and a past chairperson of the PBR Review Procedures Work Group.
Communication. Consumer behavior data can also inform communication strategies for insurers. For example, “actuaries often want to be very precise, but data shows that can diminish comprehension of communications,” Stevenson says. “Instead of saying ‘23.578%,’ consumers are more likely to remember ‘about a quarter.’ Also, if you give people three options, they will tend to choose the middle option. Penalties are more effective than rewards—people react more strongly to something being taken away than to something being added.”
In addition to choosing the right words and phrasing, selecting the appropriate communication channel will make a difference. According to LIMRA’s 2024 Life Insurance Fact Sheet, 59% of U.S. adults and as many as 84% of Gen Z adults use social media for financial guidance—so insurance companies should consider communicating through social media to achieve results.
Andrew Larocque, health actuary at Risk and Regulatory Consulting, a 2025 Rising Actuary Award recipient and vice chairperson of the Academy’s Behavioral Economics Work Group, says he opts in to allow his car insurer to track his driving behavior through an app in exchange for a discount. “They get real-time data from many drivers, which they use for modeling, predicting, and pricing,” Larocque says. “But it’s important to realize that the data they’re collecting is from a specific population that knows about the program, and that knowledge may cause us to drive differently.”
Some behaviors could signal risk, such as driving during overnight hours. “But if it’s a person who may not have a choice and works the graveyard shift, that assumption of risk may be misplaced,” says Dorothy Andrews, senior behavioral data scientist and actuary at the National Association of Insurance Commissioners. “The problem with some of this data is that it has positive and negative implications for insurance pricing and consumers, and insurance companies and regulators have to balance both sides to determine the appropriateness of including the variable in a model.” Andrews served as a member-selected director on the Academy Board of Directors from 2021 to 2024. She is chairperson of the AI Subcommittee and a past chairperson of the Data Science and Analytics Committee.
In addition to data generated by insured individuals through technology, some insurance companies also use data from government and other sources in risk modeling. For example, one health insurance company purchased the mailing list of Tennis Monthly magazine, says Randall Stevenson, president and consulting actuary at Hause Actuarial Solutions, Inc. “They wanted to market to that audience because they assumed a lower number of health claims,” says Stevenson, who is chairperson of the Behavioral Economics Work Group and a past chairperson of the PBR Review Procedures Work Group.
And some health plans use data related to social determinants of health, “such as noting that people in a certain ZIP code may be more likely to experience food insecurity,” says Sara Teppema, chief actuary at Wildflower Health and vice chairperson of the Academy’s Health Equity Committee. “Social data can be very predictive of health care costs.”
Behavior vs. Behavioral Economics
When using consumer behavior data to assign risk, it’s important for insurance leaders to distinguish between behavioral economics and simple human behavior. “Most insurers currently limit their operational risk modeling to observable characteristics such as age, gender, location, occupation, and, in some cases, lifestyle factors like smoking status,” Samuell says. “These variables provide only a partial view of customer behavior. They do not capture values, preferences, and enduring psychological traits that strongly influence outcomes. Personality, for example, is one of the most consistent predictors of behavior, yet it is rarely incorporated into risk models.”
That’s why customers who appear similar on paper can have vastly different outcomes. For example, two people may experience similar car accidents, yet one may return to work quickly while the other develops a long-term disability. With current datasets, such differences are invisible at the underwriting stage, Samuell explains.
Ethical and Regulatory Considerations
As the use of consumer behavior data expands, so does the need for well-defined ethical and regulatory frameworks. Actuaries and data scientists must be especially mindful of how behavioral inputs might inadvertently introduce bias.
For example, Dorothy Andrews, senior behavioral data scientist and actuary at the National Association of Insurance Commissioners (NAIC), attended a health conference where “a data scientist said confidently, and without presenting any evidence, that if a person owns a dog, they are more likely to be a smoker.” The data scientist explained that a person who smokes will take their dogs for a walk. Many felt that the rationale did not provide an intuitive and credible relationship between dog ownership and cigarette smoking.
Such broad assumptions that are not based in fact are not uncommon and must be avoided. “It’s dangerous to draw conclusions about someone’s personality when you aren’t trained in psychology,” Andrews says. “Not all consumers respond to changes in financial matters in the same way. You can’t use behavioral economics to assess every aspect of a person’s personality or behavior. Not all human behavior can be interpreted through the lens of behavioral economics. You have to understand the policies and frameworks that have influenced the data in order to begin identifying any embedded biases.” Andrews served as a member-selected director on the Academy’s Board of Directors from 2021 to 2024. She is chairperson of the AI Subcommittee and a past chairperson of the Data Science and Analytics Committee.
To combat the prospect of bias and inaccurate assumptions when using data, some lawmakers are pushing to regulate how predictive models are used in insurance, says Andrew Larocque, a health actuary at Risk & Regulatory Consulting, a 2025 Rising Actuary Award recipient, and vice chairperson of the Academy’s Behavioral Economics Work Group. The state of Colorado, a leader in the movement, passed a law in 2021 protecting consumers from unfair insurance practices.
Colorado’s SB21-169, Protecting Consumers from Unfair Discrimination in Insurance Practices, holds insurers accountable for testing their big data systems—including external consumer data and information sources, algorithms, and predictive models—to ensure they do not discriminate unfairly against consumers on the basis of a protected class.
In addition to legal measures, the Actuarial Standards Board has promulgated standards of practice that address actuaries’ responsibilities when risk modeling. “Actuaries must follow both the regulations and actuarial standards of practice,” Larocque says. “And as technology changes, so do the regulations.”
Every person has hidden characteristics, such as honesty and responsibility, that play a big role in their behavior. “Insurance companies are analyzing a lot of data to determine how responsible someone is or how honest someone is,” Andrews adds. “They might use credit scores to determine how responsible someone is, but some people in communities of color have no access to credit, so that’s unfair for an insurance company to automatically assume you’re high risk just because you don’t have credit history. There are so many third-party variables that are behavioral in nature that are not dictated purely by economics.”
How Actuaries Can Balance the Data Equation
An actuary’s role is to establish the criteria for how insurers approach risk management, and new sources of consumer behavior data create the possibility of “richer, more nuanced models,” Samuell says. “Purchasing patterns, digital traces, and social media interactions can reveal traits such as impulsivity, a risk factor relevant across many product lines.”
Insurance can learn from other industries about appropriate segmentation. For example, online retailers and other consumer goods companies group their customers into clusters based on behavioral and demographic data, and then tailor communication and pricing accordingly. “Insurers could benefit from a similar approach,” Samuell says. “Underwriting and communication strategies that consider latent traits would likely produce more accurate risk ratings, higher retention, and faster claims resolution.”
For example, recent research shows that even slight adjustments to the enrollment process can lead to significant improvements. Lower disclosures from customers screened by financial advisors, a study that Samuell co-authored, found that customers reported fewer risk factors when screened by financial advisors for life insurance, resulting in unfairly cheaper and more favorable policies. By eliminating financial advisors from the process, the participating insurer in the study found it could reduce life insurance premiums by 20% without affecting profitability.
Results like these “illustrate the transformative potential of broadening risk models,” Samuell says. “If insurers fail to adapt, they risk being overtaken by non-financial firms that already leverage consumer behavior data at scale.” It also underscores the crucial role of actuaries in designing enrollment and risk assessment processes that ensure accurate data, fair pricing, and reliable risk models.
For insurance companies interested in innovating their risk assessment models, a wise first step is “to establish clusters of customers with common features that correlate with risk,” Samuell says. “Once identified, these clusters could be enriched with additional sources of behavioral data, including wearables, social media, and financial records. Actuaries are well positioned to guide this process, ensuring that new models remain analytically rigorous, ethically defensible, and commercially valuable.”
When deciding which variables should be included, there are many factors to consider. Andrews recommends ensuring that the variables have a true relationship to risk. She recalled one insurance company that wanted to include data in their risk assessments about how much mining occurs in the state of the insured. “The relationship to risk was not intuitive, and no correlation was demonstrated,” she says. “You don’t have to demonstrate causation, but you should show predictor model variables have credible correlations to risk.”
To effectively make those decisions and lead their organizations in using data appropriately, actuaries need to improve their technical expertise as well as their understanding of the data, Larocque says. He recommends achieving this by collaborating closely with data scientists.
“There are so many more data points available now, and actuaries need technical skills to know how to use the data and incorporate it into models without introducing bias or creating a discriminatory model,” Larocque says. “It’s very common for actuaries and data scientists to work together on these issues.”
Nancy Mann Jackson is a freelance writer for Contingencies.