Every listing site sold you the same car three times

A dealer group sells four hundred units in a month. The marketplaces report six hundred and twenty leads converted, the manufacturer's campaign dashboard claims a hundred and ninety, the group's own paid search reports two hundred and forty, and the general manager is looking at a set of reports that between them have sold roughly two and a half times the actual inventory. Nobody has misreported anything.

Every vendor in automotive marketing counts the buyer they touched. Only the DMS counts the car that left the forecourt.

Automotive is the category where this problem is most acute, because the shopping journey is unusually long, unusually multi-platform, and ends in a physical building where the last three touches leave no digital trace at all.

Why the automotive journey defeats tracking

A vehicle buyer typically researches across several marketplaces, manufacturer sites and review platforms over weeks, then contacts one or two dealerships, then visits in person. Most of that journey happens on platforms you do not own, and the final, decisive part happens in a showroom.

This means digital tracking sees the middle of the journey and neither end. It does not see the months of consideration on third-party sites, and it does not see the test drive, the trade-in valuation and the finance conversation that actually close the sale. What it does see is the enquiry, which sits between them and is easily over-credited.

The consequence is that any measurement built purely on digital signals will systematically favour whatever platform happened to be the last click before an enquiry form, which in this category is very often a marketplace whose role was to display inventory the shopper had already decided to look at.

That is not an argument against marketplaces, which genuinely generate demand as well as intercepting it. It is an argument that their own reporting cannot tell you which of the two they did.

The DMS is the only honest denominator

The dealer management system records every unit sold exactly once, with a customer, a date and a value. Matching those records against your lead sources produces a channel table that sums to reality rather than to the sum of your vendors' claims.

This is the whole method and it needs no new technology. Export closed deals with phone, email, sale date and gross. Export marketplace leads, website enquiries, call records and showroom intake going back further than your longest shopping cycle. Normalise the phone numbers on both sides. Match.

The export is usually the hard part, and it is hard for organisational rather than technical reasons: the DMS is owned by operations, the lead sources by marketing, and nobody has previously had reason to put them in the same place. It is a week of coordination, not a project.

The eight hundred and seventy claimed units compress into four hundred real ones. Every source is smaller than it reported and the ranking has changed: the marketplace with the largest claim is not the largest contributor of matched revenue.

Walk-ins are not unmeasurable

A large share of walk-ins have already appeared in your data as a website visit, a marketplace enquiry or a phone call weeks earlier. Matching on phone and email recovers many of them, and the genuinely sourceless remainder is a real and useful number.

The assumption that walk-ins are unmeasurable is one of the most expensive in the category, because it exempts a large slice of revenue from any analysis at all. In practice, most people who walk onto a forecourt have been researching for weeks and have left a trace somewhere, and the reason it is not found is that nobody looked.

What this requires is intake discipline: capturing a phone number or email at first contact, and asking the source question consistently. Neither is popular with sales staff, and both are the difference between a measurable business and one operating on impressions.

The genuinely sourceless remainder is worth reporting rather than hiding. In most groups it is a meaningful figure representing reputation, location and repeat custom, and knowing whether it is growing or shrinking is a strategic signal that no campaign report contains.

Rooftop variation, which is usually the biggest finding

Close rates and response times vary far more between rooftops in a group than between marketing channels. The group average conceals it, and closing the gap between the best and worst sites is usually worth more than any media reallocation.

The pattern repeats across groups: two sites with similar traffic, similar inventory and similar brands, converting at rates that differ by a factor of two. The difference is almost never marketing. It is how fast enquiries are answered, whether evening leads are worked the next morning, and whether anybody follows up a second time.

Attribution by rooftop makes this visible in a way that lead volume reporting cannot, because it connects each site's behaviour to closed gross rather than to activity. That is the number that gets attention from a group operations director.

The OEM co-op question

Manufacturer co-operative funds come with reporting requirements and a strong implicit claim about their own effectiveness. Matched revenue lets a group evaluate that claim on its own data rather than accepting the campaign dashboard.

This is delicate, because co-op funding is a commercial relationship as much as a marketing one and the money is welcome regardless. The useful position is not to challenge the programme but to know what it is actually producing, which puts the group in a better place in every conversation about tiered spending and local match requirements.

What matched data can establish is how much closed gross traces to leads generated during co-op activity. What it cannot establish is how much of that would have happened anyway, since manufacturer campaigns run continuously and there is rarely a comparable region with the activity switched off.

Being clear about that distinction protects the analysis. A group claiming that co-op spend produced a specific incremental return will be challenged and will not be able to defend it; a group reporting matched gross by source, with the causal question left open, is on solid ground.

Trade-ins, finance and the value question

Vehicle gross is only part of the transaction value. Finance and insurance income, service plans and trade-in margin vary systematically by source, so ranking channels on vehicle price alone can rank them wrongly.

Marketplace shoppers are typically price-focused by construction, since the platform's core function is price comparison. Buyers arriving through brand search or referral frequently take more finance and more added products. A channel comparison on unit gross alone will therefore understate the sources producing the most profitable customers.

The remedy is to export total transaction profit rather than vehicle gross where the DMS can produce it. It is one more column and it changes conclusions often enough to be worth insisting on.

The used and new distinction

New and used vehicles have different shopping journeys, different platforms and different margins, and blending them produces a channel ranking that describes neither accurately.

Used shopping is more marketplace-driven, more price-sensitive and more geographically flexible; new is more manufacturer-driven and more brand-loyal. A group running one attribution report across both will find that marketplaces dominate, largely because used volume dominates, and will draw a conclusion about new vehicle marketing that the data does not support.

Splitting the analysis costs nothing, since the DMS already records the distinction. It is one of the few refinements in this article that is purely free.

CloseRev reports on any dimension present in your export, so rooftop, new versus used, and vehicle line all come from columns your DMS already produces, without separate configuration per site.

The service department nobody attributes

Service and parts revenue is substantial, recurring, and almost never included in marketing attribution. It is also where a large share of a group's profit sits, and where repeat contact makes matching unusually reliable.

The omission is historical rather than principled. Marketing budgets in automotive have traditionally been aimed at vehicle sales, so the reporting followed, and service marketing has been treated as an operational cost rather than as an investment with a measurable return. Meanwhile the service department is generating a steady stream of transactions against customers you already have identifiers for.

Matching here is easier than on the vehicle side, because the customer is usually already in your systems from the original purchase, and the interval between marketing and transaction is short. A service reminder campaign and the resulting bookings can be matched within weeks rather than months.

It also changes how the vehicle sale should be valued. A customer acquired through one channel who then services with you for five years is worth considerably more than the same vehicle gross acquired through a channel whose customers never return, and the only way to see that difference is to include service revenue in the matched history.

Group versus rooftop budgets

Most groups run some marketing centrally and some at rooftop level, and the two are rarely evaluated on the same basis. Matched revenue lets you compare them directly for the first time, which is usually uncomfortable in both directions.

Central spend tends to be evaluated on reach and cost efficiency because it is bought at scale; rooftop spend tends to be evaluated on whether the general manager felt it worked. Neither is a measurement, and the two are never compared, which means the split between them has usually been set by history rather than by evidence.

A matched revenue view by source and by rooftop makes the comparison possible, and the finding is rarely that one is uniformly better. More commonly, central spend performs well in some markets and poorly in others, and a rooftop with strong local activity is being charged for central campaigns that reach a catchment it does not serve.

That is a governance conversation as much as an analytical one, and it goes considerably better when the numbers arrive before anybody has taken a position. Producing the analysis quietly and sharing it with rooftop management before it reaches a group meeting is worth the delay.

What to do about the sourceless remainder

After matching, a real share of units will have no traceable source. Treat that figure as a measurement of your intake discipline and your reputation, and watch its direction rather than trying to eliminate it.

Some of it is genuine: customers returning after five years, personal recommendations, people who drove past. That portion is a real asset and its growth is a signal worth tracking, particularly in groups investing in local reputation and customer experience rather than in media.

The rest is intake failure, and it is addressable. Enquiries taken without a phone number, source questions left blank, walk-ins logged after the fact from memory. The share of your unknown bucket that is intake failure rather than genuine word of mouth can be estimated by looking at how many unknown sales have any contact record at all, and it is usually higher than management expects.

What you should not do is pressure the number downward as a target. An unknown bucket driven to near zero by staff filling in a mandatory field with whatever gets past validation is worse than a large honest one, because it looks like data.

Where to start

Take one month of DMS closed deals and three months of lead records, match them, and look at the difference between claimed and matched by source. That single table changes most groups' view of their marketing within an afternoon.

Start with one month rather than a year, because the objective at this stage is to establish whether the match works and what the data quality looks like, not to produce a definitive analysis. Phone number formatting between a DMS and a marketplace feed will be the first obstacle, and it is much easier to diagnose on a small file.

Then extend the lead window backwards until the matched share stops improving. That point tells you your real shopping cycle, which is a number most groups quote from instinct and have never measured.

One practical warning about that extension. The matched share will keep improving for a while and then flatten, and the temptation is to keep reaching back because the number is still moving. Stop when the improvement per additional month becomes marginal, because beyond that point you are adding data volume and processing cost for an accuracy gain nobody will notice in a decision.

Your vendors will keep sending reports that add up to more cars than you have. The reconciliation has to come from your side, and the file you need is already in the DMS.