news details type

Getting modeling to click in automotive CRM

Daniel Vai

September 21, 2023

Steve Jobs and Henry Ford have great quotes about not just giving customers what they wanted, but instead giving them something beyond what they could imagine.

The Model T and iPhone were revolutionary products that forever changed their respective industries. Both products were launched without any serious competition. You may not have bought the perfect car or phone for you, but at least the decision was easy.

Fast-forward to the well-established automotive industry of 2023. There are no world-transforming modes of transportation currently on the horizon (though I’m no Henry Ford). The problem now is that there are over 400 different models, each with multiple grades, trims, packages, and colors, totaling tens of thousands of possible builds that cover almost every possible combination of customer needs. This makes finding the exact right vehicle challenging for both the buyer and seller. At RAPP, we recognize the auto options exist because no two consumers needs are exactly alike. Thankfully, we now have more data than ever to help consumers navigate to that perfect match… at least at an individual manufacturer level.

As a CRM Experience Strategist for Toyota and Lexus at RAPP, one of my core challenges is determining which vehicles to market to which customers, within what time frame, and how to highlight each vehicle’s niche.

By the end of this year, there will be over 30 Toyota vehicles to choose from (including HV and PHEV variants)! Most consumers probably aren’t aware of more than 4 or 5 Toyota vehicles. Deciding on which combination of powertrain and features best supports their family, driving habits, and lifestyle, can be overwhelming. 

RAPP has created some very sophisticated statistical modeling suites for both Toyota and Lexus. Modeling suites use hundreds of customer data points and behaviors to predict future behaviors. Our predictive models not only predict what vehicle a might fit a given consumer’s needs, but also contain hundreds of data points that help us predict when they might be in-market, their likelihood to engage with a given communication, and much more.

These statistical models are then used as input in our decisioning engines and dynamic communication templates to determine who is in-market for a vehicle and which combination of Toyota’s 30 vehicles they should be sent. These inputs are a core pillar of a larger performance marketing engine that RAPP has setup for TMNA.

Who we send to

The core model that determines who gets triggered into a vehicle sales email, predicts when someone will purchase their next vehicle. This model scores every consumer in the database for their likelihood to purchase. Profiles that hit a certain score threshold are then pulled into these campaigns. This model is constantly tested and refreshed for accuracy and to find new consumers.

What we send them

Once we figure out who to send to, we use another statistical model to determine what vehicles to send each recipient. In a more complex campaign that utilizes both modeling and behavioral capabilities, the first touch features multiple vehicles to choose from, then later touches drill down to a single vehicle. This model scores every consumer for each vehicle. This means theoretically, every profile in the database has an ordered list of vehicles they are most likely to purchase from 1 to 30. We typically feature the top 3 to 4 of these in an initial email.

Strategy’s challenge is conveying the lifestyle, features, and benefits of each of these 30 vehicles within a few lines of copy and an image or two. Each vehicle needs to feel unique, and its niche must be clear enough that someone scanning over 3-4 of them in an email (in under 30 seconds) can identify these differences and which vehicle might be right for them. From there, we can track exploration and behaviors on .com, narrowing in on what vehicles are garnering the most attention. These behaviors also give direction on where an individual may be in their shopping journey based on lower funnel actions taken (e.g., vehicles built, quotes requested, offers viewed, etc.).

Bringing it all together

Once the audience for a given program is determined, the decisioning engine and predictive models work with dynamic email templates to personalize and populate each email according to the strategy and resulting business rules. Personalized support content such as offers, and regional events can also be populated based on an individual’s profile.

Hoping it all ‘clicks’

Once this ultra-sophisticated lure is cast, we wait for a bite…

As soon as someone clicks on a vehicle (or vehicles) the priority of content dynamically switches from the modeling suites to customer engagement. Models are amazing tools to show shoppers the vehicles that might be right for them, but only they know which one is right for them. By continually tracking what a customer is interested in and progressively profiling them, we can serve up the exact right communication to fit their needs. When we account for all the possible combinations of vehicles, features, copy and content an individual can receive, there are tens of thousands of email permutations that can be created to fit (almost) any set of needs.

Using this optimal combination of modeling and behavioral indicators we aim to fast-track vehicle buyers to their perfect ride…. At least until the next Model T comes out.

news details separator

Similar stories