Mapping the Consideration Phase in Automotive Retail

Dealerships have more customer data than at any point in the history of automotive retail. A dealer can see when a shopper visits the website, which vehicle pages they view, how often they return, which marketing source brought them back, and whether they submitted a lead. Customer data platforms can connect many of these events to one customer record.

The technology has become good at reconstructing activity.

The harder question is what the customer was trying to learn during that activity.

A shopper who visits the same Toyota RAV4 five times is sending a clear signal of interest. The five visits do not explain the decision that the shopper is trying to make. The customer could be concerned about long-term reliability. They could be comparing the XLE with the Limited trim. They could be checking the vehicle history after seeing an accident record. They could be trying to understand fuel cost, insurance cost, resale value, or whether the asking price is reasonable.

Each case can produce the same VDP view in a traditional activity log.

This gap has become an important area of research for us at Carvia. We want to understand whether dealers can map the consideration phase itself. That means understanding the subjects that a customer researches while they decide whether a vehicle is right for them.

The industry has built a detailed view of customer activity

Current automotive technology provides much more customer visibility than it did several years ago.

Orbee, for example, creates an identity-resolved timeline that can include web sessions, ad clicks, CRM activity, service visits, forms, and transactions. Orbee describes the purpose of the system as understanding who a customer is and what that customer did.¹

Fullpath provides a similar customer timeline. It records dealership website activity, CRM events, marketing interactions, leads, appointments, and other events. It can also show VDP activity from Autotrader and Kelley Blue Book when the required Cox data is available. This includes vehicles from the dealer and vehicles from the wider market.²

This is significant progress. A salesperson can have a much more complete record of a customer's path through the market.

Other products add information about the customer's apparent preferences.

Foureyes can show every vehicle a known prospect viewed on a dealership website. It can also show repeat views, browsing price range, new or used preferences, and search history. Foureyes states that this search history can show preferences for trim, color, and specific features.³

Cars.com has also moved in this direction. Its Shopper Details product can provide the vehicles that a lead compared, the price range used during the search, and other dealerships that the person contacted. Cars.com positions this information as a way to give the salesperson more context before the first conversation.⁴

These products provide useful information about a customer's consideration set. A salesperson can start to see whether the customer is focused on one vehicle, several similar vehicles, or a wider category.

There is still another level of information inside that process.

Predictive systems solve a different problem

Automotive retail has also invested heavily in predictive analytics.

automotiveMastermind provides a useful example. Its Behavior Prediction Score includes separate In-Market, Vehicle, and Deal scores. These scores help a dealership determine which customers to contact, which vehicle a customer is likely to purchase, and how the customer is likely to transact.⁵

This type of system can be valuable because sales teams have limited time. Prediction helps them decide where to direct that time.

However, a prediction about the final outcome does not provide a complete record of the customer's evaluation process.

Suppose a system predicts that a customer has a high probability of buying a RAV4. A salesperson now has useful information about priority and likely vehicle choice. The prediction does not necessarily explain why the RAV4 remains under consideration, what could remove it from consideration, or which questions the customer still needs to answer.

Those questions concern the content of the decision.

We have started to think about this as a separate data layer.

Activity tells us what happened. Predictive analytics estimates what could happen next.

Consideration data describes what the customer is evaluating while the decision is still in progress.

Vehicle research now happens across many surfaces

The need for this information increases as the automotive research process becomes more distributed.

Cox Automotive's 2026 Car Buyer Journey Study surveyed 2,300 people who had purchased a new or used vehicle during the previous 12 months. Buyers reported using an average of 4.6 websites. Seventy-five percent used third-party automotive websites, 59% used dealership websites, 41% used search engines, and 26% used social media. AI sites had already reached 12% of buyers.⁶

A dealer can therefore observe only part of the research process in many cases.

Consider a customer who leaves a dealership VDP and returns two hours later. During those two hours, the customer can read owner discussions, compare another model, search for common mechanical problems, check fuel economy, ask an AI service about reliability, and review market prices.

The dealership sees the return visit.

Much of the research that caused the return is outside the dealership's field of view.

This has an important consequence. Better identity resolution can connect more events to the same person, but identity resolution cannot create information about an interaction that the dealer never observed.

The type of experience that a platform operates therefore affects the type of data that it can collect.

A website analytics system can record a page view. A marketplace can also record search filters, saved vehicles, and comparisons. A conversational system can record the specific questions that a customer asks.

Impel provides evidence for the value of this last category. The company reports that it analyzed almost 50 million AI conversations across 8,000 dealerships. Financing questions represented almost 15% of shopper requests. A conversational interface can identify this subject because the customer states the question directly.⁷

The signal becomes richer when the product takes part in the research process.

That observation has become important to our work at Carvia.

Owning the research interaction changes the data

Carvia was originally built to help answer vehicle-specific questions on dealership VDPs.

The product separates vehicle research into defined areas. Depending on the vehicle and available data, the Carvia experience can include vehicle insights, valuation, verified records, ownership history, specifications, ownership cost, and EV information.⁸

These sections have a second function that becomes important when we study customer behavior.

They give meaning to an interaction.

A conventional analytics platform can record that a customer viewed a RAV4 VDP. If that customer interacts with Carvia's ownership-cost section, Carvia can identify the subject of that interaction. If the customer then opens the vehicle history, returns to valuation, and later spends more time with buyer-fit information, those events contain information about the areas the customer chose to investigate.

Carvia already records defined interactions within the research experience. Its implementation of the Automotive Standards Council GA4 specification includes events for report navigation, carousel interaction, configuration actions, accordions, cost-view controls, and other interactions that Carvia directly operates.⁹

The important point is not the number of events that can be recorded. The value comes from knowing what those events represent.

A click on a generic page tells us that the customer clicked.

An interaction with an ownership-cost analysis tells us that ownership cost received the customer's attention at that point in the journey.

Repeated interactions can provide more context. If a customer returns to ownership cost several times, the subject could have more importance in the decision. If attention later moves to vehicle history, that change can become part of the customer record.

We have to be careful with the interpretation. Viewing reliability information does not prove that a customer has a reliability concern. Opening a valuation section does not prove that price is an objection. These are signals, not statements of fact about a person's thoughts.

However, they are more specific signals than a VDP view.

Shared Pages give us a way to follow consideration over time

This question became more important as we developed Carvia Shared Pages.

A salesperson can share a persistent Carvia vehicle research page with a customer. The customer can continue to use that page after the original dealership visit or conversation. Our customer-level reporting then gives the salesperson visibility into activity associated with that shared experience.

We initially treated much of this as engagement visibility. A salesperson could see that a customer returned to the information that they received.

The more interesting opportunity is in the structure inside those returns.

Consider a customer who receives a Shared Page for a used SUV. The customer reviews valuation on the first visit. Later that evening, they return and spend time with ownership cost and owner information. The next morning, they review vehicle history.

A basic activity timeline could show three visits.

A consideration timeline could preserve the subjects that received attention during those visits.

As more information becomes available, this record could start to show how the customer's evaluation changes over time. It could show which vehicle remains in the consideration set, which research subjects receive repeated attention, and where new areas of research appear.

This would give the salesperson useful context before the next conversation.

The salesperson does not need another alert that says a customer is "hot." The salesperson needs enough information to prepare for a useful conversation.

If ownership cost has received repeated attention, the salesperson can prepare the relevant information. If the customer has spent time with vehicle history, the salesperson can review that history before making contact. If trim information becomes important, the salesperson can prepare the differences between the available vehicles.

This can reduce the need to restart discovery each time the customer and salesperson speak.

Cars.com uses a similar principle in its explanation of Shopper Details. It states that many sales representatives begin with basic discovery because they have little more than a customer's name, contact information, and vehicle of interest. Its additional shopping data is designed to give the representative context before the conversation begins.⁴

We think vehicle research can provide another source of that context.

A consideration map needs more than a score

There is still a large amount of work between collecting these signals and producing a reliable model of customer consideration.

A useful system would need to understand the vehicles that remain under consideration, the research subjects that receive attention, the intensity of that research, and how those signals change with time.

It would also need clear limits.

Research behavior cannot tell us exactly what a person thinks. A customer can open a section for many reasons. A salesperson should therefore see evidence and patterns rather than unsupported conclusions.

The system could say that a customer reviewed ownership cost four times.

It should be much more careful before it says that ownership cost is preventing the sale.

This distinction matters because the goal is to give the salesperson better information. The goal is not to replace the customer's own explanation of their needs.

The dataset may become as important as the interface

Our current work is focused on making customer-level research activity useful to dealership teams. Over time, the underlying data could answer broader questions.

For example, we could study which research subjects become more common late in the purchase process. We could measure whether different vehicle categories produce different research patterns. We could examine which topics customers revisit before they request more information, schedule an appointment, or move to another vehicle.

Those questions require a dataset with semantic structure.

Traditional activity data records the occurrence of an event. A structured research environment can also record the subject of the event.

At sufficient scale, that could give dealers a better view of the consideration phase than they have today.

Automotive technology has already made substantial progress in identity resolution, activity tracking, and purchase prediction. Products across the market can now show detailed customer journeys and increasingly accurate signals about vehicle interest.

The next area of work is deeper inside the decision.

For Carvia, that starts with a simple research question: Can we make the customer's vehicle research visible enough that a salesperson can understand how the customer is evaluating the car before the next conversation begins?

Shared Pages are giving us a practical place to test that question.

If the answer is yes, the customer journey can become more useful than a sequence of visits and clicks. It can begin to show the information that the customer used to make the decision.

References

1. Orbee. “Behavioral Exploration.” Orbee. Accessed September 14, 2026. https://www.orbee.com/products/behavioral-exploration/.↩
2. Fullpath. “Shopper Timeline.” Fullpath Help Center. Accessed September 14, 2026. https://help.fullpath.com/hc/en-us/articles/49172101654164-Shopper-Timeline.↩
3. Foureyes. “Using a Prospect’s Profile Page to Uncover What to Say.” Foureyes Help Center, October 19, 2021. https://support.foureyes.io/en/articles/5586454-using-a-prospect-s-profile-page-to-uncover-what-to-say.↩
4. Cars.com. “How to Follow Up on Cars.com Leads with Shopper Details.” Cars Commerce, May 7, 2026. https://www.carscommerce.inc/shopper-details-lead-follow-up/.↩
5. automotiveMastermind. “automotiveMastermind Introduces Its Enhanced Behavior Prediction Score.” October 2, 2024. https://www.automotivemastermind.com/automotivemastermind-introduces-its-enhanced-behavior-prediction-score/.↩
6. Cox Automotive. “Cox Automotive Car Buyer Journey Study Finds Efficiency, Digital Tools and AI Drive Record Satisfaction.” January 13, 2026. https://www.coxautoinc.com/insights/cox-automotive-car-buyer-journey-study-finds-efficiency-digital-tools-and-ai-drive-record-satisfaction/.↩
7. Impel. “Automotive AI Financing Intelligence That Puts Real Numbers First.” Impel. Accessed September 14, 2026. https://impel.ai/blog/automotive-ai-financing-intelligence/.↩
8. Carlson, Jack. “Widget Carousel — Your Full Carvia Report, One Embed.” Carvia, January 21, 2026. https://carvia.ai/release-notes/widget-carousel.↩
9. 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.↩