What it does, and why the number it gives you is different from every other tool.
Carmano looks at a car somebody is selling privately, works out what a company like Carvana or CarMax will actually pay for it, and then tells you the highest price you can pay and still make money.
That last number is the whole product. Everything else is arithmetic in service of it.
Say you find a car. The seller wants $3,000. You check and Carvana will pay $4,000 for it.
That looks like a thousand dollars of profit. It is not. It is a loss, and here is where the money actually goes:
| What Carvana offers | $4,000 |
| Their price isn't final until they inspect it — 6% | −$240 |
| Driving there and back (230 miles, plus your time) | −$418 |
| Repairs you can't see in the photos | −$450 |
| Title and registration | −$145 |
| The minimum profit that makes it worth doing at all | −$600 |
| Most you can pay for this car | $2,147 |
So paying the seller's $3,000 loses about $253. It felt like a good deal and it was a bad one. That gap is what costs people money in this business, and closing it is the entire point of the tool.
Same car, same asking price, but eight miles away instead of 115: the ceiling goes up to $2,494. That is $347 of difference created by nothing except where the car is parked.
And if the car were sitting outside your front door, the ceiling still only reaches $2,520 — because meeting the seller, looking the car over and doing the paperwork costs an hour no matter what. There is no such thing as a free pickup.
On this car, distance and unseen repairs move the number more than the spread does.
There are tools that already do part of this. They cost around $80 to $100 a month. Every one of them estimates what a car is worth using their own software, trained on what other people are asking for similar cars.
An asking price is an opinion. A Carvana offer is a company telling you the exact amount of money they will wire you.
This matters more than it sounds. The most common complaint about those tools, in their own app store reviews, is that their software flags beaten-up cars as great deals — because a wrecked car and a clean car look the same in a listing. That problem cannot really be fixed by guessing harder. It goes away entirely when you ask the buyer instead.
The obvious way to estimate what a buyer pays is a percentage of the car's value. Say 78%. That works fine on expensive cars and falls apart on cheap ones.
The reason: before a buyer makes a single dollar, they spend roughly $2,900 per car on reconditioning, auction fees and holding costs. That $2,900 is nearly the same whether the car is worth $30,000 or $3,000.
So the percentage they keep is not a fixed number at all. A flat 78% would have told us this Altima was worth $5,616 when the real offer was $4,000 — over $1,600 of money that does not exist, and it invents that money precisely at the cheap end, which is where this business actually operates.
Carmano models the fixed cost separately. On the one car we have a real offer for, it predicts $3,960 against an actual $4,000.
Some of this is worth saying plainly, because a few of these were choices to give something up:
The engine works and is tested — every figure in this document is produced by the software itself, not typed in by hand. Open it on your phone:
The home page is now a live dashboard. Give it a zip code and it scans real for-sale listings nearby — licensed data, Craigslist and Autotrader and KBB, never scraped — and ranks the ones worth a second look, each with a full VIN decode and an estimated instant-offer number.
Found something yourself instead — Facebook Marketplace included? There's a box for that on the same page. Paste in what the listing says and it fills in the year, make, model, price, mileage and VIN on its own; you just check it over and fill in anything it missed.
What's still typed in by hand, on purpose: the real Carvana or CarMax number. See "what it deliberately will not do" above for why that one stays manual — and log it once you have it. That's the single most valuable thing you can put into this tool, more than any amount of automation would be.
Every cost in that first table is currently an estimate: the $450 of unseen repairs, the $145 of paperwork, the 6% inspection haircut. They are reasonable estimates, but they are not measured.
They get replaced by real numbers the moment we do real deals. The tool already records what it predicted next to what actually happened, so it gets more accurate with every car — and that record is the part a competitor cannot buy.
The first one just landed: the F-150 you found, messaged the seller on, and got a real $14,000 Carvana quote for. Carmano's own math, run on nothing but the ask and that quote, landed on the same call you made — pursue it, and don't go past about $11,400. That's not a demo number. It's the first real row in the record.