Last touch attribution
Last touch attribution explained: what it is and where it misleads
Last touch attribution gives 100 percent of the credit for a sale to the final recorded marketing touch before it, ignoring every touch that came earlier in the path. It is the default in most ad platforms and reporting tools, which is why it shapes more dealership budget decisions than any other model, deserved or not.
You will also see it written as last click attribution, last interaction attribution, or last-touch. It is the mirror of first touch attribution, and it carries the same structural flaw pointed the other way.
How is last touch attribution calculated?
For each delivered vehicle, find the most recent recorded touch before the sale and assign the whole unit to that channel. From there:
- Channel share equals units whose last recorded touch was that channel, divided by total units, times 100.
- Last-touch cost per sale equals that channel's spend, divided by the units it was credited with.
Most platforms add a lookback window, commonly 30, 60, or 90 days. A touch older than the window does not count as the last touch, so the window quietly changes the answer as much as the model does.
Last touch attribution explained: a worked example
The buyer path: day 1 an organic search, day 5 a third-party marketplace listing, day 12 an opened BDC email, day 19 a click on the branded paid search ad, day 21 delivery. Last touch hands the entire car to paid search. The organic search that started it books zero.
Across the month, a store spending $40,000 delivers 40 units. Last touch credits paid search 12 units, marketplace 10, email 10, SEO 6, and social 2.
Turn that into cost per sale. Paid search runs $12,000 divided by 12, or $1,000 a car. SEO is $10,000 divided by 6, or roughly $1,667. Social lands at $3,000. And CRM email, on $2,000 of spend and 10 credited units, comes out at $200 a car, apparently the best-performing channel in the store by a factor of five.
Under first touch the same 40 cars gave email zero units and social eight. Nothing about the month changed. Only the rule for handing out credit did.
The case for last touch attribution
- It matches the platforms. Ad platforms and analytics tools report this way by default, so your internal report reconciles with the invoices and the dashboards people are already looking at.
- It is the cheapest model to run. One field, no path storage, no modelling, no minimum conversion volume. A single rooftop can maintain it in a spreadsheet.
- It is the most actionable for bidding. If you are managing keyword-level bids week to week, the touch nearest the conversion is the one you can move fastest.
- It is hard to argue with on short paths. Some deals genuinely are short. A service customer clicking a service reminder and booking the same day has one touch that matters, and last touch gets that case right.
- It resists inflation from old touches. A stale click from 80 days ago cannot claim a sale, which stops long-lookback channels from hoovering up credit.
The case against last touch attribution
- It confuses harvesting with creating. Branded search, retargeting, and direct traffic are usually the customer coming back to a store they already chose. Last touch pays those channels for demand somebody else created.
- It defunds the top of the funnel. Discovery channels score near zero, so they get cut, and the pipeline they were filling thins out one or two quarters later when nobody connects the two events.
- The real last touch is often untracked. The phone call to the BDC, the text from the salesperson, the second showroom visit. The last thing your system saw is rarely the last thing that happened.
- Every walled garden claims it. Each large ad platform measures conversions in its own system with its own window, so more than one of them will report the same car. Added up, platform-reported sales routinely exceed the units the store actually delivered.
- The 100 percent is a convention. There is no evidence in the data that the final click did all the work. The model would report the same confident figure if that click had changed nothing at all.
How last touch differs from first touch
Both models take a purchase that involved four channels over three weeks and hand the whole thing to one of them. First touch rewards the opener, last touch rewards the closer, and neither is measuring cause. They are two different tie-break rules applied to the same unresolved question.
The practical move is to run both and read the gap. High on last touch and low on first touch means a harvesting channel: useful, and not something to scale on the assumption it generates demand. High on first and low on last means the opposite. Weak on both is the channel that deserves the hard conversation.
If the disagreement between the two is large enough to change a budget decision, that is the point at which a multi-touch or time-decay model earns its cost. Marketing attribution covers those, along with why privacy changes and platform reporting have made click-level credit less dependable than it used to be.
