Your marketing budget is sorted by invoice size. Gross isn't.

Between July 9 and August 31 we merchandised 313 used units across 13 lots — 2,231 finished images. The processing bill for all of it came to $197.31.

That's 9.9 cents an image and about 63 cents per car, front to back. On a lot carrying 60 units, the entire line item is smaller than one tank of gas for the loaner.

It's also the cheapest thing on the marketing bill, which in most budget reviews means it's the first thing questioned and the first thing cut. That instinct is backwards, and the arithmetic isn't close.

Budget reviews are conducted in descending order of invoice

Watch how the meeting actually runs. Somebody prints the vendor list, sorts it by monthly cost, and starts at the top. The listing-site subscription gets defended because it's big and everybody's afraid to turn it off. The website platform gets defended because the site would break. Then the cursor reaches the small lines and the tone changes: do we really need this?

The sort order is the problem. Invoice size tells you what you spend. It tells you nothing about what the spend gates.

Line item	Typical monthly	Units it touches	Cost per unit	What it gates
Listing site subscription	$1,500–$2,500	All	~$30	Reach, if listings are complete
Paid social + search	$800–$3,000	Some	Varies wildly	Traffic to VDPs
Website platform	$300–$800	All	~$10	The page that closes
Merchandising / photo pipeline	Under $50	All	$0.63	Whether any of the above has anything to show

The bottom row is two orders of magnitude cheaper than the top and sits upstream of every other row. You can pay for reach on a listing site and still ship a car with five photos and no interior frame, at which point you've bought traffic to a page that can't convert it. The $30 line is downstream of the $0.63 line, and the $0.63 line is the one on the chopping block.

The real comparison isn't against other line items

Compare the merchandising cost to the asset instead. A used unit on the lot carries floorplan interest, depreciation, insurance and reconditioning capital — call it $30 to $40 a day, conservatively, on a mid-priced car.

At 63 cents per unit merchandised, the entire photo bill for one car is worth about twenty-five minutes of that car's carrying cost. If better merchandising moves the unit one day sooner, it paid for itself roughly fifty times over. One day. Not a week.

That's the frame the budget review never uses, because carrying cost lives on a different page of the P&L than marketing does, and nobody in the room owns both pages. Which is exactly how you get a lot that spends $2,000 a month on reach and five days per car on merchandising lag — the two numbers never meet.

We wrote about the lag itself in cars waiting on Thursday: the processing side now finishes in minutes, and units still take days to go live because the calendar never changed. The budget version of that same finding is that the constraint moved to a line item too small to appear in the conversation.

Sort by gross per dollar and the list rearranges

Try the review again with a different sort. Not "what does this cost" but "if I zeroed this line, how many units stop moving, and how much gross walks out."

The answers get uncomfortable fast, and they're specific to your lot:

• Cut merchandising. Every unit ships thinner. Not one stops listing, so nothing looks broken on Monday. Days-on-lot drifts up two weeks later and nobody connects the two events. This is the most expensive cut on the list and the one that looks the most harmless.
• Cut paid social. Traffic drops immediately and visibly. The units that were already moving keep moving. You find out within nine days whether it was doing work, which is the argument for testing a pause rather than debating it.
• Cut the listing site. Reach drops hard on aged units specifically, because that's where third-party traffic over-indexes. Fresh, well-priced cars barely notice for a month.

None of that is a recommendation — it depends on your mix, your market, and what your VDPs actually look like. The point is that the ranking by consequence looks nothing like the ranking by invoice, and only one of those two rankings is on the printout in front of you.

Want your own version of that table? We build it per lot, with your numbers, before proposing a single change — see how that works or book a call.

The number that makes the sort possible

You can't rank by gross per dollar without knowing which dollars touched which cars, which is the whole argument in attribution failing at the stock number. Without that key, every line item in the table above gets defended with a story instead of a number, and the loudest vendor wins the meeting.

It also means the denominator has to be honest. Ad spend divided by units sold is not cost per car, for reasons we laid out in what cost per car sold actually includes — it counts walk-ins and repeat customers as marketing wins and flatters every channel equally.

Two numbers, then, before any budget decision is worth making: what a unit actually cost to sell, and which spend touched that unit. Everything else in the meeting is preference.

What to do Monday morning

1. Re-sort the vendor list by cost per unit touched, not by monthly invoice. Monthly cost divided by the units it plausibly affected. Ten minutes in a spreadsheet, and the order will surprise you.
2. Put carrying cost on the same page as marketing spend. One line: what a day on the lot costs you. Every merchandising and speed decision becomes arithmetic instead of opinion once that number is visible in the room.
3. Test one pause, not three. Pick the line you're least sure about, pause it for two weeks, and watch leads per unit — not total leads. Three simultaneous pauses teach you nothing, because you can't tell which one moved the number.