Propensity Modeling for Motor Vehicle OEMs: A Plain-Language Guide

Demographic targeting tells you that 45-year-olds in a certain ZIP code tend to buy your product. Propensity modeling tells you which specific households are likely to buy next. That difference is worth real money.

Propensity modeling is a statistical method that estimates the likelihood that an individual or household will take a specific action, such as buying a vehicle, within a defined window. It uses purchase history, geo-demographic data, and behavioral signals to produce a probability score per person. Unlike demographic targeting, which groups by broad traits, propensity modeling ranks actual likelihood to act.

What is propensity modeling in plain language?

Propensity simply means likelihood. A propensity model looks at everything it knows about a person and outputs a score: how likely are they to do the thing you care about, like buy a specific vehicle in the next year.

Instead of treating everyone in a demographic bucket the same, it scores each person individually and ranks them from most to least likely. You then focus your marketing and sales attention on the people most likely to act, which is where this kind of consumer analytics pays off.

What inputs does propensity modeling use?

Three broad categories of input feed a propensity model. Purchase history is the strongest signal, including what a person has owned, when they bought, how often they replace, and where they service. Geo-demographic data describes who and where they are, including location, household characteristics, and life stage. Behavioral signals capture what they are doing now, such as browsing, configuring, requesting brochures, or other indicators of active shopping.

The model learns the patterns that historically preceded a purchase, then looks for those patterns in your current audience. The richer and cleaner the inputs, the sharper the score.

How does propensity modeling differ from demographic targeting?

Demographic targeting segments people by shared attributes and assumes everyone in a segment behaves alike. It is broad, cheap, and blunt. It will happily spend budget on a 45-year-old in the right ZIP code who has no intention of buying.

Propensity modeling scores individuals on actual likelihood, blending demographics with behavior and history. Two people with identical demographics can receive very different propensity scores because one is actively shopping and the other is not. The result is targeting based on probable intent rather than stereotype, which cuts waste and lifts conversion. The payoff is measurable: McKinsey finds 78 percent of consumers are more likely to buy again from brands that personalize, and BCG, in research with Google, found that companies using first-party data for targeting and other key functions achieved up to 2.9 times the revenue uplift.

How do OEMs use propensity modeling differently than dealers?

Dealers use propensity at the level of individual prospects, deciding who to call, who to retarget, and who to prioritize for a test drive. Their horizon is the next sale.

OEMs use it at scale and for strategy. Aggregated propensity reveals where intent is concentrating geographically, which informs network planning, marketing allocation, and inventory decisions. For powersports and other OEMs, propensity scores rolled up to a market show where demand is forming before it shows up in registrations, which is a planning advantage dealers do not need but OEMs do.

Frequently Asked Questions

Is propensity modeling the same as demographic targeting?

No. Demographic targeting groups people by shared traits and treats the group as uniform. Propensity modeling scores each individual’s likelihood to act using behavior and history as well as demographics. It identifies probable intent rather than assuming it from category membership, which makes it far more precise.

What data do you need for propensity modeling?

You need purchase history, geo-demographic data, and behavioral signals. Purchase history is the most predictive, since past buying and ownership patterns strongly indicate future behavior. Behavioral signals add timing by showing who is actively shopping now. Cleaner, richer inputs produce more accurate and more actionable scores.

How accurate is propensity modeling?

It does not predict any single person with certainty; it ranks likelihood across a population. Its value is in prioritization: focusing effort on the highest-scoring prospects reliably outperforms broad demographic targeting. Accuracy improves with better data and with retraining as new purchase and behavioral signals accumulate.

Sources: McKinsey & Company, on personalization and repurchase (mckinsey.com); Boston Consulting Group with Google, “Responsible Marketing with First-Party Data,” 2020 (bcg.com).