Behavioral Customer Segmentation

DILEMMA

Genworth already segmented its customers — on risk stratification, propensity to claim, and program appropriateness. But every one of those facets was inside-out: they described what the business needed to know about a customer in order to manage its own operations, not who that customer actually was. The consequences showed up in performance. Campaigns and outreach produced mediocre results, and the organization had no clear understanding of the people it was trying to reach. At the same time, there was a genuine push to build relationships of trust with policyholders — something the business had not previously prioritized, having been oriented toward running its operations as efficiently as possible rather than knowing its customers. The philosophical shift this work sits inside is the move from an inside-out to an outside-in view of the business.

OBJECTIVE

Understand, identify, and measure the behavioral tendencies of customers within specific decisioning contexts — building a foundation for engagement grounded in how customers actually behave rather than in what the business needs to know about them, and doing it in a way the organization could act on at scale.

MY ROLE Behaviorist, Research Program Manager

CAPABILITIES LEVERAGED Behavioral Economics, Qualitative Research (IDIs, SME Focus Groups, Call Listening), Quantitative Research & Model Validation, Experimental Design, Behavioral Customer Segmentation, Voice of Customer, Cross-Functional Data Partnership

PROCESS

Qualitative foundation. The work opened with a qualitative research sprint spanning literature review, 100 recorded customer calls reviewed, 200 outreach calls, and three SME focus groups of two to four participants each. Cross-synthesis across those sources produced a first set of directional customer segments and a working customer journey. An intercept survey on the customer portal added further VOC signal.

Grounding in behavioral science. A second literature review focused specifically on behavioral economics within the insurance and financial services context, followed by experiments designed to surface testable hypotheses about how customers approach these decisions.

Modeling the decision. In-depth interviews with 16 long-tenured policyholders established the decisioning model customers actually follow when facing this context — not a theoretical decision path, but the reasoning real policyholders use.

Scaling and validating. Findings were scaled through a quantitative study with lookalike audiences (500 responses) to validate the model at volume, followed by a second quantitative study with 300 actual Genworth customers to confirm the model held for the real policyholder population rather than a proxy for it.

The dimensions. The research produced three primary behavioral dimensions, each grounded in the Voice of Customer of people who had actually navigated these decisions:

• Financial Capability — financial guidance, financial literacy, economic means

• Willingness to Engage — social support, institutional trust, tolerance for uncertainty

• Vulnerability — protection breadth, coverage adequacy, propensity to claim

A secondary set captured technology tolerance, social identity, and attitudes toward insurers.

Making segments observable. A segmentation model is operationally useless if the organization can't tell which segment a given customer belongs to. Two workstreams ran in parallel toward that problem and converged on congruence. The data team identified operational data sources and facets they hypothesized would indicate the biases and constructs the research had surfaced; those hypotheses were verified and mapped to specific survey questions. The survey produced clusters of customers based on their responses, and the data team then took those groupings back into the operational data to observe whether the same patterns appeared there independently. Where the two converged, the segmentation became something the organization could apply to its actual customer population rather than only to survey respondents.

TENSIONS NAVIGATED The central constraint was the operational data itself. Some dimensions of customer reality the research had identified simply had no observable proxy in the existing datasets, and others proved difficult to trace reliably. The segmentation could only become operational to the extent the organization could actually see the behavior it described — which meant accepting that parts of a well-evidenced model would remain, for the moment, unmeasurable at scale.

RESULTS

(Work is ongoing; findings below are directional.)

• Five behavioral segments identified to date, built on three primary dimensions and validated across both lookalike and actual customer populations

• Segments mapped to observable operational data, making the model applicable to the full policyholder population rather than survey respondents alone

• Replaced a purely business-led view of customers — risk, claim propensity, program fit — with one grounded in who customers actually are

• Changed how the organization interacts with its customers, introducing deliberate strategic complexity, with content, outreach, and engagement differentiated by behavioral pattern rather than applied uniformly

• Activated the organization's product strategy for long-tenured policyholders on a behaviorally grounded foundation