Research
Automotive CDPs and the Consideration Layer
CDPs have become very good at reconstructing the customer journey. The next opportunity is understanding what the customer is evaluating while that journey is happening.
Dealerships can see more of the customer journey than they could even a few years ago.
A customer data platform can connect a website visitor to a known CRM record, combine activity across devices and channels, show which vehicles that person viewed, incorporate service and ownership history, and place the activity into a chronological customer timeline.
Fullpath's Shopper Timeline, for example, can include dealership website activity, VDP views, forms, calls, CRM activity, appointments, marketing interactions, service repair orders, sales events, and matched VDP activity from Autotrader and Kelley Blue Book.1 Orbee similarly combines web sessions, CRM interactions, service visits, and offline transactions into an identity-resolved customer timeline.2
Enterprise CDPs go further as general-purpose infrastructure. Tealium can accept events from websites, mobile applications, servers, APIs, and data warehouses, then use those events to build visitor attributes and audiences.3 Adobe Real-Time CDP is designed to combine known and anonymous data from multiple sources into profiles that can be segmented and activated across channels.4
The industry has become very good at collecting customer activity.
There may be another useful layer underneath it.
We previously wrote about that layer from the shopper's side of the decision: mapping the consideration phase itself. This essay asks a different question: where that record sits in the customer data stack, and why a CDP cannot produce it on its own.
A simple way to think about the customer data stack
Most of the customer intelligence being assembled today can be organized into three broad layers.
Identity asks: Who is this shopper?
This includes CRM records, contact information, device identifiers, household data, ownership history, and service records. Identity resolution is one of the fundamental functions of a CDP. Adobe describes its Real-Time Customer Profile as combining information from multiple online, offline, CRM, and third-party sources into a unified customer view.4 Segment similarly uses identifiers such as user IDs, email addresses, phone numbers, and device IDs to connect activity to a profile.5
Activity asks: What did the shopper do?
The customer visited the website. They viewed an F-150. They came back two days later. They filtered the inventory. They opened an email. They submitted a form. They called the store. They scheduled service.
This is where automotive CDPs have become particularly sophisticated. Fullpath can even distinguish between dealership inventory and market inventory viewed on participating third-party shopping sites.1 Orbee describes its customer-level exploration product around a similar question: who is the customer, and what did that customer do?2
Prediction asks: What may happen next?
Once identity and activity have been assembled, systems can model them. A dealer can prioritize customers who appear to be in market, estimate which vehicle they are likely to purchase, predict propensity, or determine a next action.
automotiveMastermind is a good example. Its Behavior Prediction Score separates In-Market, Vehicle, and Deal scores, designed to help answer when a customer may be entering the market, which vehicle they may purchase, and how they may transact.6 Segment also offers predictions and recommendations on top of unified customer profiles.7
Identity, activity, and prediction form an increasingly powerful stack.
But consider what happens when a shopper actually starts deciding whether a particular vehicle is right for them.
A VDP view can contain several different decisions
Imagine two shoppers who each visit the same used Toyota RAV4 VDP five times.
At the activity layer, they can look almost identical.
Both generated five VDP views. Both returned several times. Both spent meaningful time on the page. Neither has submitted a lead yet.
Their actual research can be completely different.
One shopper may be trying to understand whether the asking price is reasonable. Another may already accept the price but is concerned about long-term reliability. Someone else may be comparing the XLE and Limited trims. Another shopper may have noticed something in the vehicle history and returned specifically to investigate it.
A conventional event stream could represent all four shoppers as repeated engagement with the same VIN.
That is accurate. It is also incomplete.
The missing information is the subject of the research.
We think of this as the consideration layer:
What is this person actually evaluating about this vehicle?
Price and valuation are consideration signals. So are ownership cost, history, reliability, trim and equipment, specifications, buyer fit, and EV-specific questions.
The behavior around those subjects creates another set of signals. Which subject did the shopper investigate first? Which one did they return to? How deeply did they explore it? Did their research shift to something else? Did they come back to the same subject the following day?
These signals describe the structure of the research itself.
Why a CDP does not automatically have this information
The important distinction is architectural.
A CDP can store remarkably rich data. Tealium supports custom event sources and visitor attributes, including funnels and timelines.3 Adobe supports custom ExperienceEvent schemas and can incorporate partner-provided fields into a customer profile.8
So there is nothing preventing a CDP from storing a field such as:
research_subject = ownership_costThe challenge is producing that field in the first place.
A CDP normally receives information from other systems. A website sends pageviews. The CRM sends lead activity. The DMS sends ownership or transaction data. An ad platform sends campaign engagement. A call provider sends call activity.
The meaning available to the CDP depends on the meaning created by the source.
If the source sends:
page_view
VIN: ABC123the CDP knows that a vehicle was viewed.
If the source sends:
accordion_click
VIN: ABC123the CDP knows that an interface element was used.
But if an experience understands that the accordion contained a five-year ownership-cost analysis, it can generate a more useful event:
VIN: ABC123
research_subject: ownership_cost
interaction: expandedThat difference becomes more important across several interactions.
A CDP could receive:
7:41 PM VDP viewed
7:42 PM Valuation researched
7:44 PM Ownership cost researched
7:47 PM Ownership cost revisited
9:16 PM Shared Page returned
9:17 PM Vehicle history researched
8:05 AM Trim information researchedNow the customer record contains something different from a sequence of pageviews.
It contains a record of the customer's vehicle research.
The deeper signal comes from the experience that generates it
This is why the type of product a shopper uses affects the type of customer data that can exist.
A marketplace can understand search criteria because it owns the search experience.
A conversational platform can understand the subjects a customer asks about because it owns the conversation.
A structured vehicle research product can understand which areas of a specific vehicle the shopper chooses to investigate because it owns that research experience.
Carvia's current experience already divides vehicle research into defined subjects such as valuation, verified records and history, specifications, ownership cost, buyer-fit information, and EV data.9 The product also records interactions it directly operates, including report navigation, carousel interactions, accordions, cost controls, EV controls, CTAs, and communication engagement.10
That gives the interaction semantic structure.
Opening ownership cost means something different from opening vehicle history. Revisiting valuation means something different from moving into trim specifications.
The distinction sounds small at the event level. Across a customer journey, it can become meaningful.
Carvia has already been exploring this idea through Shared Pages. A customer can continue using the same vehicle research experience after the initial visit or sales interaction. Instead of recording three generic return visits, the research record can preserve which subjects received attention during each visit.
This creates the possibility of following consideration across time rather than measuring engagement only within one session.
This should complement the CDP
There is a tempting way to describe this as a limitation of CDPs.
That framing misses the more interesting opportunity.
The CDP is extremely useful once these events exist.
Suppose Carvia sends a structured consideration stream into Fullpath, Orbee, Tealium, or another customer platform. The CDP can connect that research to everything it already knows.
A repeated interest in ownership cost can be joined to CRM status, previous ownership, service history, marketing source, and eventual purchase. Research on vehicle history can be examined alongside lead submission and appointment data. A dealer group could determine whether certain research patterns occur more frequently before a sale, a vehicle switch, or customer defection.
The resulting customer profile becomes deeper without requiring the CDP to become a vehicle research product.
This is why the stack may be better represented as four layers:
Identity: Who are they?
Activity: What did they do?
Prediction: What may happen next?
Consideration: What are they evaluating about this vehicle?
The first three layers are already becoming sophisticated.
The fourth requires a source of structured research data.
There is an important limit to what these signals can tell us
More granular data can also encourage overinterpretation.
Suppose a shopper looks at valuation four times.
We can confidently record that valuation received repeated attention.
We cannot confidently say:
price_objection = trueThe shopper may think the price is excellent. They may be comparing the vehicle against another listing. They may be showing the valuation to a spouse. They may simply be curious.
The observable fact is valuable precisely because it does not require us to guess.
The same applies to reliability, vehicle history, or ownership cost.
A good consideration dataset should preserve evidence:
ownership_cost_interactions: 5
history_interactions: 3
last_research_subject: history
research_sessions: 4Higher-level systems can then determine what those signals correlate with.
The distinction matters. The objective should be better visibility into the customer's evaluation process, not an attempt to claim certainty about what the customer is thinking.
The next question for automotive customer data
Customer data platforms have made major progress in connecting fragmented dealership data.
They can increasingly tell the dealer who a shopper is, reconstruct what that shopper did, and use those signals to estimate what the shopper may do next.
The next area of opportunity may sit one level deeper.
A vehicle purchase involves a series of questions. Is this priced correctly? What will it cost to own? Is there anything in its history that matters? Does this trim have the equipment I want? Is this vehicle likely to fit how I use it?
When those questions are researched outside the dealership experience, the dealer rarely sees them.
When the research happens inside a structured environment, those questions can begin to produce data.
That opens an interesting role for Carvia in the customer data stack.
Carvia does not need to replace the CDP. The larger opportunity may be to give the CDP something it did not previously have:
a structured, first-party record of VIN-level consideration.
The CDP can then do what it already does well. Resolve the identity. Connect the journey. Build the profile. Model the behavior. Activate the data.
Carvia can help make the research inside that journey visible.
References
- Fullpath. “Shopper Timeline.” Fullpath Help Center. Accessed September 30, 2026. https://help.fullpath.com/hc/en-us/articles/49172101654164-Shopper-Timeline.↩
- Orbee. “Behavioral Exploration.” Orbee. Accessed September 30, 2026. https://www.orbee.com/products/behavioral-exploration/.↩
- Tealium. “AudienceStream CDP.” Tealium Docs. Accessed September 30, 2026. https://docs.tealium.com/server-side/getting-started/audiencestream-cdp/.↩
- Adobe. “Real-Time Customer Profile Overview.” Adobe Experience League. Accessed September 30, 2026. https://experienceleague.adobe.com/en/docs/experience-platform/profile/home.↩
- Twilio Segment. “Identity Resolution Overview.” Segment Docs. Accessed September 30, 2026. https://segment.com/docs/unify/identity-resolution/.↩
- automotiveMastermind. “automotiveMastermind Introduces Its Enhanced Behavior Prediction Score.” October 2, 2024. https://www.automotivemastermind.com/automotivemastermind-introduces-its-enhanced-behavior-prediction-score/.↩
- Twilio Segment. “Unify.” Twilio Segment. Accessed September 30, 2026. https://segment.com/product/profiles/.↩
- Adobe. “XDM ExperienceEvent Class.” Adobe Experience League. Accessed September 30, 2026. https://experienceleague.adobe.com/en/docs/experience-platform/xdm/classes/experienceevent.↩
- Carlson, Jack. “Widget Carousel — Your Full Carvia Report, One Embed.” Carvia, January 21, 2026. https://carvia.ai/release-notes/widget-carousel.↩
- Carlson, Jack. “ASC GA4 Standard v1.2 — Dealership Analytics Without a Separate Session.” Carvia, August 15, 2026. https://carvia.ai/release-notes/asc-ga4-v1-2.↩