// Cost per car sold ·
What "cost per car sold" actually includes — and why your number is wrong
Ad spend divided by units is a comforting fraction, not a cost. Here is what belongs in the real number, and why it usually lands two to three times higher.
Daniel Melo · 4 min read
Ask a dealer what he spent on marketing last month and you get an answer in four seconds. Ask what it cost to sell one car and you get a pause, then the same number divided by units.
Those aren't the same thing, and the distance between them is where the money has been hiding.
The comfortable fraction
Ad spend ÷ units sold is easy to compute, which is exactly why it survives. It also quietly assumes two things that aren't true.
That marketing produced every sale. It didn't. Some of those units went to walk-ins, repeat customers, and the guy who drove past on a Saturday. Counting them makes marketing look cheaper than it is.
That the cost of a sale is only the ad. It isn't. A car that took 61 days to move cost you floorplan the whole time, and it took that long partly because the listing had four dark phone photos.
What actually belongs in the number
| Line | Usually counted? | Why it belongs |
|---|---|---|
| Paid media | yes | the only one nobody forgets |
| Photography and listing production | rarely | a 1% VDP wastes every click you bought |
| Lead handling | almost never | the lead that waits 13 hours is spend with no return |
| Days on lot | never | floorplan is a real cost of a slow sale |
| Attributed units only | no | crediting walk-ins to marketing flatters the number |
Build it honestly and it usually lands two to three times the comfortable figure. That reads like bad news for about a day. Then it becomes the first number you can manage, because you can't cut a cost you've never counted.
Why the honest number is the useful one
Here's the practical difference. A lot running $6,000 a month and moving 40 units says "$150 a car" and feels fine. Rebuild it — only attributed units, with production and lead handling and the extra days carried in — and it's $380. Now the questions change:
- Is it cheaper to buy more clicks, or to fix the VDP that converts at 1%?
- Is the next hire a salesperson, or the thing that answers leads in 60 seconds?
- Which channel is actually producing sales, not sessions?
None of those questions are answerable from the comfortable fraction. All of them are obvious from the honest one.
That's also why the number only works if you can trace a sale back to a click. Without that, you're allocating on vibes — which is the attribution problem, and it's the piece most lots skip. The systems we run exist to make that one number trustworthy.
What to do Monday morning
- Pull last month. Media spend, plus what you paid for photos and listing work, plus whatever the lead handling costs you in salary hours.
- Count only attributed units. If you can't attribute, mark it unknown — don't credit it to marketing to make the ratio prettier.
- Add days on lot. Multiply your average floorplan day rate by the days above your target. That's a marketing cost, even though it never shows up on a marketing invoice.
Divide. Write the number down. It's uglier than the old one and it's the only one worth defending a budget with.
// Straight answers
CPA usually stops at the lead. Cost per car sold goes to the sale and carries production and time. A lot can have a great CPA and a terrible cost per car sold — that's exactly the case where the leads are cheap and nobody closes them.
Keep them in a separate bucket labeled unknown, and work on shrinking it. Don't split them proportionally into channels — that invents precision you don't have and hides the real problem, which is the missing attribution.
When the delay is caused by the listing, yes. A car that sits 14 extra days because of four dark photos cost you those days. If you'd rather keep floorplan separate in the P&L, fine — but track it next to this number, not in another report nobody opens.
There isn't a benchmark worth quoting, because it moves with price band and market. The useful comparison is your own number, month over month, computed the same way. A number you can trust and trend beats an industry average you can't verify.