Yesterday, while watching our kids play, my neighbor asked me about my specialty at work. He knows I’m actively looking for job opportunities. I took a moment to think and responded, “My specialty is modeling human behavior using data science.”
This common thread connects all my previous roles. As it turns out, modeling human behavior is both complex and incredibly valuable in various industries. If I were to create another course as an adjunct professor (I previously taught Storytelling with Data at BYU), it would be titled Modeling Human Behavior Using Data Science.
The Challenge of Modeling Human Behavior
What makes modeling human behavior different from other data science disciplines? The answer lies in human unpredictability. Machines, like those I worked with at General Electric, can generate erratic sensor data, but they never behave irrationally—there’s no human brain involved. Humans, on the other hand, are highly diverse, and our decisions are influenced by emotions and circumstances.
While predicting any one person’s behavior is difficult, modeling behavior across large groups is more feasible. We may be irrational, but we are predictably irrational—a concept explored in one of my favorite books, Predictably Irrational by Dan Ariely.
Why Businesses Care About Behavior Modeling
Modeling human behavior is particularly useful for businesses developing products. Key questions companies want answered include:
- Will people buy this product? If so, at what price?
- Which features should be prioritized?
- What messages will encourage purchases?
- What should the packaging and design look like?
- Where should paywalls be placed in software?
- How will customers react to market, economic, or political changes?
- Which employees might be bad actors?
Answering these questions requires a mix of survey and behavioral data. Predictive models—like SARIMA, Prophet, and Exponential Smoothing—help forecast sales. Feature prioritization and pricing can be tackled using survey methods such as Conjoint and MaxDiff, powered by Hierarchical Bayes and Monte Carlo Markov Chains. Identifying bad actors relies on entity resolution and graph databases. Customer segmentation, enhanced through KMeans and KMedoids clustering, improves message testing.
Of course, my neighbor wasn’t interested in the technical jargon. What he really wanted to know was how these methods are applied in surprising and impactful ways. Here are three real-world applications I’ve worked on.
Cast Studies in Behavioral Modeling
1. Analyzing Voice Audio for Emotion-Based Price Quotes
While working for a national shipping broker, I encountered an interesting pricing strategy. Companies would call in to request a semi-truck trailer shipment. The brokerage had its own drivers but also relied on a network of independent truckers.
The company developed software to analyze emotion in callers’ voices and factor that into price quotes. The result? Angry customers often received better rates. This underscores how behavioral modeling can be leveraged in unexpected ways.
2. Optimizing the Right Offer at the Right Time
I’m currently consulting for a startup optimizing when Shopify users see promotional pop-ups. The challenge is finding the right timing: presenting a discount to a customer who was going to buy anyway reduces revenue, but waiting too long could mean losing the sale entirely.
I’ve worked with this company to align the models with business outcomes. This is the art of behavior modeling—balancing unknowns and assumptions to make informed decisions.
3. Using Email Domains for Loan Default Scoring
I once consulted for a personal loan lender. Certain demographic factors (such as race, ethnicity, gender, and age) cannot legally be used in loan decisions. However, I discovered an interesting predictive signal—email domains.
Applicants with Gmail addresses were the least likely to default, while those with Hotmail accounts were the most likely. My neighbor, who uses Hotmail, found this amusing. While surprising, insights like these help businesses make better decisions while staying compliant with regulations.
These three examples highlight how human behavior modeling plays a crucial role in various industries. Do you have any surprising applications of behavior modeling? Share them in the comments!