A coaching and decision engine for the half of the deal nobody built software for.
More deals, structured better and closed faster, with more parts and product on each, and customers who come back. Analysis nobody on the floor has time to do, handed to the people who make the call.
It runs on your own hardware, in your building, beside the systems you already have.
$600K to $1.2M
recaptured a year at a hundred-unit store
A hundred-unit store means a hundred units a month. Every figure on this page is projected, not measured.
Every recommendation goes to a person. Nothing is submitted, approved, or signed by software.
We are signing launch dealers in Portland, Los Angeles, St. Louis, and Chicago.
The current wave of dealership software is chat widgets, lead responders, automated callers, appointment setters. A shopper fills in a form on a listing site and software answers faster than a person would, books a time, and sends the reminders.
Some of it is good. All of it is receptionist work: the cheapest part of the process, the easiest thing to demo, and the part with the most vendors already in it.
None of it touches the desk, where the deal gets built. None of it touches finance and insurance. None of it touches what the store earns on a car, or what the store pays its people for selling one.
This is not a better version of those tools. It starts where they stop.
The reasonable assumption is that this must already be solved. Here is the stack a modern store runs, and the gap each piece leaves.
| System | What it does | What it does not |
|---|---|---|
| DMS | System of record. Inventory, accounting, repair orders, deal jackets. | Does not structure a deal or find the best version of one. |
| CRM | Tracks leads and customers, schedules the follow-up. | Does not know what is in a deal or what it pays. |
| Desking tool | Works up payments on a chosen vehicle. | One car at a time, and blind to where an approval has room. |
| F&I menu | Presents protection products in the box. | Presents them at hour four, to an exhausted customer. |
| Lead providers | Send shoppers who filled in a form. | Nothing at all for the shopper who never fills one in. |
Every one of those is bought, paid for, and working. The deal still gets built by hand, on a desk, by someone with eleven minutes.
One deal
to the people who sold it
Illustrative figures.
Carniche works the half of the deal that starts when the customer is interested. It sits beside the systems you already run and does the analysis nobody on the floor has time to do.
Traffic count says how many arrived. It says nothing about what they wanted, and nothing at all about the ones who left without speaking to a salesperson. They are the largest category of lost revenue in a dealership and the only one with no report attached to it.
A shopper connects to your guest wifi and scans a sticker to see the car's details. There is nothing to download and no account to create. The store gets a profile it was never going to get from a conversation that never happened.
The ones who leave without engaging are not a lost sale. They are a prospect the store never knew it had.
Ask a customer on the lot what they want to pay and you get a shrug. Not because they do not know. Because saying a number out loud to someone who works on commission feels like handing over the only card you are holding.
Asked on the lot
We're just looking.
Answered on a screen
Illustrative figures.
The difference is control. Their number, on their own time, with nobody reading their face and nothing being negotiated yet. Seven answers in under a minute, because there is no advantage to be lost by giving them.
And some questions are simply embarrassing to answer out loud to a stranger. What someone can afford. What they still owe on the car outside. What they were turned down for last year. Nobody wants to say those across a desk to a person who is watching them say it. Almost everyone will type them into a form that is not looking back.
That is the customer's own information, volunteered rather than pried loose, before anyone worked a figure. It is the most valuable input in the building and almost no store collects it.
One couple, one lot, one afternoon. Nothing in the first version went wrong. It is a normal, competent Saturday.
Signed at 4:55. A mini. Everyone is tired, nobody is thrilled.
Signed at 3:05. Two hours less, four hundred more.
The protection she agreed to on her own phone that morning is already in the deal by the time she reaches the finance office.
Illustrative figures.
On the customer's own phone, over your guest wifi. No app, no login, no account to create and no password to forget. It carries your store's name. They scan a window sticker, a Monroney, a VIN, a QR code, or a stock number, and that car lands in their own list. The system gives them a code instead of asking for a name, and it asks the harder questions while they browse. It works in the service drive too.
The same screens the customer sees, plus everything the customer does not. Add to the record as you learn it. Scan more inventory worth showing. Note what they did not like on the drive, which is the part nobody writes down. Carry the marque in full, what else is close, and how to build an alternative when the first choice will not fit. A salesperson never leaves the table to find something out, and never hands a customer to the desk with a blank profile attached.
Desk, F&I, and the office. Structure inside the approval, with alternates and the real gross on each. Three stock numbers at once. Payout off the same record that produced the gross. Pay plan modeling against twelve real months. Who is performing, and who is being handed the easy half of the floor. Stocking signal from what the lot is telling you. This is the only seat that sees all of it, and it is the seat where the decisions get made.
One piece of hardware, in your back office, on your network. It stays in your building and serves your store and nothing else. The AI runs on the box. Access by PIN or a single-use link, issued per person. It arrives preloaded with marque information, specifications, comparison data, and product rules.
Nothing here replaces a system you already run. The three screens are new, the box is new, and everything else stays exactly as it is.
None of these are new and none of them are anybody's fault. They are what happens when the back half of a deal is built by hand, under time pressure, by people with other customers waiting.
Six departments under one roof, each with its own numbers, most of them learning what the others did after the fact. One record of what happened changes what every one of them can do.
| Team | What they get |
|---|---|
| Sales managers | A new hire producing in week one instead of month four, and a floor that can answer any question about any car without leaving the customer. |
| Inventory managers | What was scanned and never sold, by segment and price band. Demand expressed by people who never filled in a form. |
| Parts managers | Accessories priced as a payment at the table, offered to the four people who want them rather than to everyone. |
| Service managers | Higher dealer-add volume, and a real introduction to the advisor at the time of sale instead of a stranger at day ninety. |
| Store operators | The GM and the controller: who is genuinely performing, what a pay plan change would do, and where gross leaks. |
| F&I and the desk | The best structure inside the approval the moment it lands, with alternates and real gross on each. |
Week one against month four is projected from operating experience on real floors, not measured from a live installation.
Lands on: sales managers and every new hire
A new salesperson normally spends a quarter learning on live customers, and the deals lost during that quarter are the real cost of hiring. They are not lost to bad attitude. They are lost to not knowing yet.
The knowledge is in the room, not in their head.
The cost of ramp is not the training. It is the deals a new person loses while learning on live customers.
Lands on: everyone standing in front of a customer
Nothing is a guess. Every question a customer asks has an answer standing by, so nobody has to say the sentence that cools a room: let me go find out.
They want the performance car and do not have the performance car budget. Today that is a lost afternoon and a customer who leaves rather than settle.
She wants the sport model. It does not fit.
| The sport trim, as built by the factory | Built to match, standard trim with parts added |
|---|---|
| tuned suspension | performance suspension kit |
| 19 inch wheels | 19 inch wheel and tire package |
| sport exhaust note | cat-back exhaust |
| badge and trim | rolled into the payment |
| $812 a month | $716 a month |
Illustrative figures.
The customer does not experience any of this as software. They experience a salesperson who listened to what they wanted and knew how to get them there.
Lands on: whoever buys the metal
Traffic count tells you how many people came. It has never told you what they came for, and it has never once told you what they came for and could not find.
Three-row SUVs under $45,000, last month
Nineteen of those shoppers scanned nothing else on the lot.
Illustrative figures.
That is not a report. It is a buying instruction.
Lands on: parts and service
Dealer adds live nowhere. Tint, mats, exhaust, paint protection, a hitch. Everyone agrees they are margin, nobody owns them, so they get mentioned late or not at all. They are also the easiest thing in the building to sell at exactly one moment: when the number is a payment rather than a price.
The advisor sees one line: wanted this, declined for payment, not for interest.
Illustrative figures.
Nothing was lost at the desk. It was scheduled.
Lands on: the GM and the controller
Two salespeople finish the month on the same unit count. One converted hard ups off the lot. One was handed the easy half of the floor for a year. You have suspected it and you have never been able to prove it, so you have never been able to coach it.
Same month, same unit count
| Person | Units | What they were handed |
|---|---|---|
| Salesperson A | 12 | cold lot traffic, no appointment |
| Salesperson B | 12 | referrals and repeat, already sold |
Illustrative figures.
Until the ups are logged, the unit count says they are the same person.
Coaching and compensation decisions that survive a conversation with ownership, because the evidence exists before the argument starts.
Lands on: the desk, finance, and whoever signs
The lender approval is the real boundary of a deal. Everything worth doing happens inside it, and finding the best version by hand takes time a busy desk does not have on the third deal of a Saturday.
The moment an approval lands, the maximized structure is already built, with alternates, each one clearing the lender and each one carrying its real gross.
One approval
| Stock | Age | Payment | Gross |
|---|---|---|---|
| 4471 | aged 94 days | $648 a month | $2,910 |
| 4602 | aged 12 days | $651 a month | $1,740 |
| 4388 | aged 61 days | $639 a month | $2,240 |
Three cars she would be happy with, within twelve dollars. The desk picks. The system never does.
Illustrative figures.
The customer does not feel steered. They were shown their options clearly, because they were.
It computes. It does not decide. The arithmetic nobody has time for, done in full, with the options ranked and the consequence of each attached. The call stays with the person who is accountable for it, and nothing is ever submitted, approved, or signed by software.
| The problem | What handles it |
|---|---|
| The shopper who never talks to anyone | Scanning on their own phone, on your guest wifi |
| Gross the desk did not have time to find | Decision engine |
| The customer who leaves to think about it | Three stock numbers against one approval |
| Product raised too late to sell | Priced as a payment, at the table |
| A new hire who takes a quarter to produce | Rapid ramp |
| A salesperson who cannot answer the question | Sales certainty |
| The customer who wanted the sport model | Built to match, from parts |
| Aged units nobody can explain buying | Inventory insight |
| Pay disputes on the fifteenth | Payout off the same record as the gross |
| A service drive that meets strangers | The want list, and an advisor introduced at sale |
The $600,000 counts one thing: gross recovered on deals that were already going to happen. It is the decision engine, and almost nothing else.
Five of the six things this system does contribute nothing to that number, and several of them are worth more over a few years than the number is in one.
| Result | Where the value lands |
|---|---|
| Recaptured gross | On the deal. Counted, and the only one we put a figure on. |
| Faster ramp | Hiring cost and turnover, not per-unit gross. |
| Stocking you can defend | Floorplan cost and turn. |
| Customers who come back | Across years of ownership, in the service drive. |
| Pay nobody argues with | Retention of the people who are good at this. |
| Compounding | Every month it runs, on everything above. |
Each of those is something a store buys separately today. We have not put a number on any of them, because we cannot defend one yet. The $600,000 is the part we can.
On day one the system knows the marque, the inventory, the approval rules, and the product catalog, and it does the arithmetic faster than anyone on the floor can. On day three hundred and sixty-five it holds a record of every deal the store wrote, every customer who walked the lot, every structure that was tried, and every product that was declined and why.
That record cannot be bought and it cannot be imported. It can only be accumulated, one real month at a time, which means the store that starts first is permanently a year ahead of the store that starts later.
This is architecture, not policy. There is no promise to break because there is no path out.
Nothing in it can send your data out. It is built that way, not promised.
It runs on your hardware, in your building. It serves your store and nothing else. It reads a nightly export and writes to nothing but itself.
The only thing that ever leaves is a count of stock numbers. No customer, no deal, no gross attached to it.
Their deal views, the menus, and the number on their phone carry your name. Carniche appears on none of it.
Worth saying plainly rather than hiding behind the word platform. The math is never guessed and never will be: payments, gross, tax, and commission are calculated. The language, the matching, the explanations, and the insight are model work, and it runs locally on the box.
A conversational assistant can be added on top, either with a hardware expansion or by allowing the box to reach an outside model. That is a choice a store makes deliberately, and it is off unless you turn it on.
Because the AI runs on the box, an outage at any vendor is not an outage for you.
We read. We never write.
It only reads. A nightly export goes one way, into the box. Nothing is ever written back into your DMS, your CRM, or anything else you run.
Which means there is nothing to unwind. No record of yours was changed. No data was migrated anywhere it has to be migrated back from. No process got rebuilt around us.
Before anything switches on we reconcile your trailing twelve months, gross front and back, product penetration, days to close, and closing ratio, and hand it back to you. It is yours whether we go further or not.
Marque information, specifications, comparison data, and product rules are on it before it reaches the building. There is nothing for you to assemble beforehand and nothing for anyone to go and find. Your nightly export carries the rest.
Training and support material is online and on demand after that, for whoever starts next month. Concierge support is available for stores that would rather have a person than a page.
Out in an afternoon. Take the box out and your store is exactly where it was, because nothing of yours was ever anywhere else.
Lock-in requires entanglement, and there is none. The only reason to keep it is that it is working, which is the only reason we want you to.
A hundred-unit store means a hundred units a month.
Every figure on this page is projected. The per-unit gross, the recapture, the ratio above, and each illustrative deal are drawn from thirty years of operating experience on real floors. They are not measured results from a live installation, because there is not one yet. That is precisely what the launch dealers are for.
Recaptured gross is one result of six, and it is the only one we put a figure on.
Per car sold, it costs less than the plate frame, the keychain, and the full tank you hand over at delivery.
$5,000 covers everything to get live: the hardware, the branded app, the integration work, enablement, review, training the floor, and the first ninety days with no service fee. Because who likes to pay for ramp.
| Inception, once | $5,000 |
|---|---|
| Service, per quarter | $3,500 |
| Retail, per unit sold | $100 |
| Standard, per unit sold | $50 |
| Launch dealer, per unit sold | $20 |
The quarterly service fee begins after the ninety days covered by inception. Billed quarterly, in arrears.
Rolling, with a ninety day exit for either party. No multi-year commitment.
Discounting is by the published card only, with no rep discretion. Nobody gets a better number by asking. The three rows above are the whole card.
Launch dealers lock their rate for the life of the store. Not for a term, not until renewal. It is per rooftop, and it does not extend to other stores in a group added later.
We are not going to pretend the launch rate is a coincidence. It is set low on purpose, because what we need out of the first stores is not revenue. It is twelve months of honest numbers from a real floor.
The word guarantee appears here only to say there is not one. What there is instead is a ninety day exit either party and a box you can take out in an afternoon.
Three people, roughly thirty years each, in different seats.
Between them: the floor, the desk, the bays, the office, the channel, and marque level.
Software the floor will not use produces garbage inputs, and garbage inputs degrade the whole record. All three of us train dealership personnel for a living. Training is in the inception fee, in person, on site, by the people who built it.
One of us took a German marque's accessory and lifestyle program to a store floor, and the first month of sales equalled the previous thirty-six combined, at one hundred percent floor compliance.
That is a measured result about getting a floor to adopt a program. It is not a Carniche result. There is no live installation yet.
Three names are deliberately not on this page. One partner's participation is still being structured, and none of us goes on paper until all three have signed off. Every name is available in conversation.
There is no version of this that is specific to one rooftop, one marque, or one region. Every dealership in the country builds the back half of its deals by hand, and there are roughly sixteen thousand franchised stores in the United States and several times that many independents.
The same four hours on a Saturday. The same mini nobody can explain. The same argument on the fifteenth. The same customer who wanted the sport model and walked.
The launch rate is a land grab and it is meant to look like one. Early stores hold it for the life of the rooftop. What the launch dealers buy is not margin. It is twelve months of measured results from real floors, which is the only thing that makes every conversation after it a short one.
These are our home markets. Being close enough to be in the building matters more in the first year than it will later.
No integration with your DMS. No change to a vendor relationship. No multi-year commitment. No exclusivity from you of any kind.
Tell us about the store. One of us will get in touch to set up a call.
Half an hour, and your last twelve months. Nothing to sign.
One of us will get in touch to set up a time. You have not been added to anything.
