AutoNation: Building a Behavioral Persona Framework from Zero

AutoNation is the largest automotive retailer in the United States, and like much of the auto industry, it has been navigating an accelerating shift from a traditionally in-person, relationship-driven business toward a digital-first customer experience. That transformation requires knowing who your buyers actually are. When I joined the team, that foundational knowledge didn't exist in any structured form.

There were no personas. No prior behavioral research. No segmentation models grounded in how customers actually make decisions. What existed was a general sense of AutoNation's customer demographics — but demographics don't tell you why someone walks out of a dealership after three hours without signing, or what would have made the online research phase feel less like a game they were losing.

This project was my onboarding project. Rather than waiting to be assigned work, I identified the gap and proposed building the research program from the ground up. The goal was to give AutoNation something it had never had: a behaviorally grounded understanding of the U.S. car-buying consumer, developed using rigorous qualitative methods and designed from day one to feed into quantitative validation.

The central design decision for this project was a deliberate shift in how we thought about personas. Prior work in the industry, and in most UX practice generally, tends to anchor personas in demographics: age, income, geography, job title. Those variables are easy to collect and easy to present, but they don't explain behavior. A 35-year-old with a $60K salary can behave like any of a dozen different buyer types depending on their relationship to information, trust, risk, and the dealership process itself.

I wanted to build personas around behavior and emotion — how people actually move through the car-buying journey, what triggers anxiety or confidence at different stages, and where the experience breaks down. That framing shaped every methodological decision that followed.

This was a sequenced mixed-methods program. The qualitative phase came first and was explicitly designed to inform the quantitative instrument and the structural logic of the whole program.

Analysis: Braun & Clarke Reflexive Thematic Analysis

I analyzed the interview transcripts using Braun & Clarke's reflexive thematic analysis (RTA) — a framework that treats the researcher's interpretive engagement as an asset rather than a source of bias to be controlled away. Unlike codebook-driven approaches or grounded theory, RTA is designed for the kind of meaning-making work required here: developing a rich, theoretically coherent account of patterns across a heterogeneous participant set, without forcing data into a predetermined structure.

The analysis produced a hierarchical codebook with 8 themes, 26 sub-themes, and 60+ codes. Each code was anchored to specific transcript evidence. The full audit trail, code evidence matrices, analytical decision logs, and participant-to-archetype assignments was maintained as a first-class deliverable alongside the persona outputs.

The most significant finding from this research is structural: the qualitative data did not support a demographic account of car-buying behavior. Participants who looked identical on paper, same age, same income band, same vehicle category, behaved in fundamentally different ways. The meaningful variation was behavioral and emotional, and it was consistent enough across the dataset to warrant distinct archetype designations.

The behavioral archetype framework generates concrete product and design implications that demographic personas cannot. Because each archetype is defined by its characteristic information needs, trust thresholds, and decision-making patterns, it points directly to where the digital experience is and isn't working and for whom.

A 15-item quantitative survey was deployed via UserTesting.com following the qualitative phase, with a stratified quota enforcing proportional representation of new and used vehicle buyers. The survey was designed to test whether the behavioral patterns identified in the qualitative phase held at scale and to estimate the population prevalence of each archetype.

The research program produced outcomes at multiple levels of the organization.

The six behavioral archetypes were presented to executive leadership and adopted by product and design teams as an active reference framework, the first time AutoNation had a shared, research-backed vocabulary for talking about who its buyers are and how they behave. The archetypes are currently in use as testing lenses in ongoing product work.

The qualitative phase directly informed the design of a 15-item quantitative survey instrument, which is now in deployment. The survey was built to validate archetype prevalence at scale and to enable cluster analysis, turning qualitative patterns into quantitatively defensible segments. This is the infrastructure that makes the archetypes strategically durable rather than a one-time deliverable.

Perhaps most significantly, the research surfaced the new/used buyer distinction as an organizational blind spot, a gap in how AutoNation understood its own customer base. That finding has opened a dedicated research thread.

Building a research program from nothing is a different challenge than running a study within an established infrastructure. There was no prior work to triangulate against, no existing segmentation to validate or challenge, no institutional knowledge about what had been tried before. Every methodological decision was mine to make and defend.