The short answer: most of what your store sells, you cannot trace. In 2025 Autotrader published a white paper — produced by Perficient using Clarivoy data across roughly 875,000 vehicle sales — that found only about 8% of sales were traceable in the dealer's CRM. The other 92% never submitted an inquiry that became a lead. They researched, compared, decided, and arrived. Your CRM has no origin record for them, because there was never a lead to hang one on.
The same research puts the average buyer's journey at around 62 touchpoints over roughly 95 days in market, against about two that the average dealer tracks — call it 3.5% of the trip. You're grading a two-hour movie off a three-second clip, and then reallocating six figures of ad spend based on what you think you saw.
Worth knowing where that research comes from: Autotrader is a marketplace with its own case to make about multi-touch measurement. The underlying sample is large and the direction is consistent with what every operator already feels — but read the number as what it is, a marketplace-funded study, and hold your own sources to the same standard you'd hold theirs.
What "attribution" actually means — three different questions
Most attribution arguments at a dealership are two people answering different questions. There are three, and they are not interchangeable:
| Layer | Question it answers | Where it usually lives | What it can't tell you |
|---|---|---|---|
| Traffic | Where did the visit come from? | Web analytics | Whether the visit was ever a buyer |
| Lead | Where did the lead come from? | CRM source field | The other 60 touches, or the gross |
| Sale | What touched the person who signed, and what did it make? | Nowhere, in most stores | — this is the gap |
The gap between the second row and the third is where the money is. A source can produce a pile of leads and no cars. Another can produce a handful that close at twice the rate on twice the gross. On a lead-count report those look identical — or the bad one looks better.
Why the CRM's source field is lying to you
Set aside the sales that never became leads for a moment. Even the ones that did get sourced badly, for three reasons that compound.
- It's typed by a human who's guessing. A shopper who saw your Facebook ad in March, searched your store by name in April, browsed the site four times, then walked in on a Saturday gets logged as "Walk-In." Or "Lives in Area." Or "Manual Add." The rep isn't lying — they genuinely don't know, and the dropdown demands an answer.
- It's single-value. One field, one answer, one journey flattened into one word. Nobody buys a $40,000 vehicle off one touch. The field structurally cannot hold the truth.
- It's last-touch by default. Whatever happened most recently gets the credit — which over-credits the bottom of the funnel (branded search, direct traffic, the site they used to look up the store they'd already chosen) and starves whatever created the demand in the first place.
Put those together and you get the quiet disaster: you defund the channels that actually work. The awareness spend that started the journey shows zero sales, so it gets cut. Six months later your branded search volume is falling and nobody can explain why.
Worth separating the two problems, because they have different fixes. This page is about the model — one last-touch field can't hold a multi-touch journey. Underneath it sits a plainer data-quality problem: the value in that field is often just wrong, typed from memory days late. That one you fix by moving the capture point, and it has its own playbook — why your CRM's lead source is wrong, and how to clean it up.
The walk-in is not sourceless
The most expensive myth in the store is that a walk-in has no source. Nobody wakes up and drives to a dealership at random. That walk-in searched, scrolled, compared, and probably visited your site more than once — you simply didn't capture the moment of contact, so the journey got amputated at the door and stamped "Drive-by." This is the same population as that 92%: buyers who were never a lead, so they were never a source.
The fix isn't a better dropdown. It's capturing the up at the moment it happens — on the floor, at check-in, in seconds — so there's a real record with a real timestamp to reconcile against everything digital you already know. A walk-in captured properly stops being a dead end and becomes the last touch in a journey you can finally see end to end. That's the whole argument of the walk-in tracking guide, and it's the highest-leverage hole to close first.
What closed-loop attribution actually requires
"Multi-touch attribution" gets said a lot and delivered rarely. Strip the vendor language and it comes down to four hard requirements:
- One identity across sources. The same human is a website session, a form fill, a phone-up, a third-party lead and a walk-in. If those are five records, you have five sources and zero journeys. Everything else depends on this, and it's the part most tooling quietly skips.
- Every touch kept, not just the last one. Store the sequence — impression, click, session, lead, appointment, visit, deal. You can argue about credit models later; you can't recover touches you threw away.
- A join to the sold unit. Attribution that stops at the lead is a vanity metric. It has to reach the deal — the stock number, the delivery, the customer who actually signed.
- Gross attached, not units. Two sources sell ten cars each. One averages $2,900 front and back; the other $1,400 with three chargebacks. Units say they're equal. The cheque book says otherwise.
Miss any one of the four and you're back to storytelling.
An honest caveat: no system resolves every shopper. Some buyers really do arrive with no recoverable trail, and anyone who promises you 100% is selling. The goal isn't perfect attribution — it's moving from 8% traceable to most-of-it traceable, which is more than enough to change where the money goes.
Third-party leads are one slice of this
If you're specifically working out whether CarGurus or AutoTrader is earning its invoice, that's this same machinery pointed at one line item — and there's a full guide on it. The trap is evaluating those platforms with a different, usually more generous, method than everything else: cost-per-lead for the listing sites, gut feel for the rest. Every source, owned and paid, gets measured the same way — cost in, gross out — or the comparison is theatre.
What to do first
You don't need a data science team. In order:
- Stop treating the CRM source field as truth. It's an input, not a finding. Reconcile it against what you can actually observe.
- Capture the walk-in at the door. It's the biggest hole and the cheapest to close.
- Join your sold list to your lead list. Even a monthly manual reconcile — every delivery matched back to a lead record — will tell you more than a year of source reports.
- Report gross by source, not leads by source. The moment you do, the rankings change. That change is the finding.
- Then fix the clock. Attribution tells you which sources deserve more money; response time decides whether you keep what you paid for.
None of this is glamorous — it's plumbing. But it's the plumbing that decides where next quarter's ad budget goes, and right now that decision is being made by a dropdown a busy rep guessed at on a Saturday. If the whole category is new to you, start with what dealer intelligence software actually is.
Questions dealers ask
- Why are 92% of car sales untraceable?
- The figure comes from a 2025 Autotrader white paper produced by Perficient using Clarivoy data across roughly 875,000 vehicle sales: only about 8% of those sales were traceable in the dealer's CRM. The reason is simpler than most people assume — the other 92% were buyers who never submitted an inquiry that turned into a lead. They researched, browsed and compared, then showed up or called, so the CRM has no origin record for them at all. It isn't that the source was recorded wrong; it's that there was never a lead to record a source against.
- What's the difference between lead attribution and sales attribution?
- Lead attribution answers 'where did this lead come from' and lives in your CRM. Sales attribution answers 'what touched the person who actually signed, and what gross did that deal make.' They routinely disagree. A source can generate lots of leads that never close, while another generates few leads that close at a high rate on strong gross — and on a lead-count report the first one looks like the winner. Only sales attribution connects marketing spend to money.
- Can you attribute a walk-in to a marketing source?
- Often, yes — if you capture the walk-in at the moment it happens instead of reconstructing it later. Nobody drives to a dealership at random; that shopper almost certainly searched, browsed and compared first. When the up is logged on the floor in real time with a timestamp, it can be reconciled against the digital touches already on record, and the visit becomes the last step in a visible journey rather than a dead end labelled 'Drive-by.' It won't resolve every shopper, but it converts the single biggest blind spot in the store into something you can measure.
