Dealership AI ROI Calculator

The math behind an AI purchase, written so you can run it on your own numbers before a vendor runs it for you.

Every AI vendor in the automotive space has a spreadsheet that shows a return, and every one of them is arithmetically correct. The trick is not in the math, it is in the inputs.

Two moves do almost all the work. The first is an assumed lift, usually presented as a range with the top of the range in bold, sourced to a case study you cannot inspect. The second is an attribution rule that credits the tool with every unit it touched, regardless of whether that unit was going to close anyway. Put an optimistic lift next to a generous attribution rule and the model returns a number that would make a hedge fund blush.

The fix is not to distrust vendors. It is to supply your own inputs and to insist that only one of the four numbers in the model is an assumption. Volume, conversion and gross should be measured from your own systems. Lift is the only estimate, and it should be treated as a range you are testing, not a figure you are accepting.

If you take nothing else from this page, take this: a projection is a hypothesis, and the only way to close it out is a measured trial. Our page on evaluating an automotive AI vendor covers the diligence side.

Get these from your own reporting, not from memory and not from an average someone quoted at a twenty group.

Opportunity volume. The number of addressable opportunities per month that the tool would actually touch. Be strict. If the tool only handles inbound sales calls, your number is inbound sales calls, not total leads. Most stores overstate this by a factor of two because they count the whole funnel.

Current conversion. The percentage of those opportunities that convert at the step the tool affects. For an answering tool that is usually appointment set. For a follow up tool it is often reengagement after a stalled thread. Match the metric to the tool, not to the metric that flatters the tool.

Expected lift. The percentage improvement in that conversion. This is the only assumption, and it should be entered as a low, likely and high case. If the vendor will not give you a low case, use zero.

Average gross per unit. Front and back combined, on the vehicle class the tool actually influences. Use a twelve month average rather than last month, and use your own number, not a regional figure.

The core calculation is short enough to run on a napkin.

Incremental units per month equals opportunity volume, times current conversion, times expected lift, times downstream close rate on the converted step.

That last term is the one most calculators skip. If a tool sets more appointments, the money does not arrive at the appointment, it arrives when a percentage of those appointments show and a percentage of those shows buy. So an appointment lift has to be multiplied by your show rate and your closing rate on shown appointments before it becomes a unit.

Incremental gross per month equals incremental units times average gross per unit.

Monthly ROI equals incremental gross minus total monthly cost, divided by total monthly cost.

Total monthly cost is not the subscription. It is subscription plus implementation amortized over twelve months, plus integration or telephony charges, plus the internal hours your team spends configuring and supervising it, priced at a real wage. A tool that costs $1,200 a month and consumes eight hours a week of a manager's attention is not a $1,200 tool.

Run the whole thing three times, at your low, likely and high lift. The spread between the low case and the high case tells you how much of the decision is evidence and how much is hope.

The numbers below are illustrative only. They are not our results, not a customer's results, and not a benchmark. Replace every one of them with your own figures before you make a decision.

Working it through: 400 opportunities at a 20% set rate is 80 appointments. A 10% lift adds 8 appointments. Half of those show, so 4. Forty percent of those close, so 1.6 units. At $2,500 gross that is $4,000 of incremental gross against $1,500 of cost.

Now run the low case at a 3% lift instead of 10%. That produces roughly half a unit and about $1,200 of gross, which is less than the cost. The tool is a clear yes in one scenario and a clear no in the other, and the only difference is an assumption. That is the whole point of the exercise.

Five of them, in the order they usually surprise people.

Supervision. Any tool that talks to customers needs someone reading what it said. Budget a named person and real hours, especially in the first sixty days. Tools that are never audited drift, and nobody notices until a customer complains.

Telephony and messaging. Per minute, per message and per number charges are usually billed separately and scale with success. A tool that works costs more to run than a tool that does not.

Integration and rework. Getting the tool talking to your CRM, your phone system and your inventory. Then the second round, six weeks later, when you discover the routing rules do not match how your store actually works.

Opportunity cannibalization. If the tool handles conversations your team would have handled, you have not added gross, you have moved labor. Real if your staff is not at capacity. Ignore it only if you can show your people are turning work away.

The exit. Contract length, notice period, and what you get back. Ask for the export scope in writing before you sign, not when you want to leave. AI BDC software and the staffing comparison both cover this tradeoff.

Three specific failure modes account for most bad AI purchases in this industry.

Attribution by touch. The tool sent a text to a customer, the customer bought, the tool claims the unit. That is not incrementality, it is proximity. The only clean answer is a holdout: run the tool on half your flow and not the other half, matched by source and time, and compare.

Baseline drift. You install the tool in March and compare to February. March is a better month in most markets. Compare to the same period last year as well, and be suspicious when both comparisons disagree.

Counting activity as outcome. Messages sent, calls answered and conversations handled are activity metrics. They tell you the tool is running. They tell you nothing about whether it produced anything. Insist that the reporting you buy shows appointments, shows and units, not volume of contact.

The discipline is easier than it sounds. Pick the outcome metric before you start, write it on the whiteboard, and refuse to let the review meeting drift to activity when the outcome is flat.

A thirty to sixty day measured trial beats any spreadsheet, and it costs less than a year of the wrong tool.

Set the baseline first, over at least four weeks, on the exact metric the tool claims to move. Then split your flow. Half through the tool, half through your current process, assigned by a rule that does not let anyone cherry pick, such as odd and even lead identifiers. Keep everything else constant, including scripts and staffing.

At the end, compare the two halves rather than before and after, because a split comparison removes seasonality and market movement from the result. Then compute the ROI with your measured lift instead of the assumed one, and make the decision on that number.

Two practical notes. Give it long enough that the sample is not noise; a store doing 400 opportunities a month needs at least a month per arm. And do not let the vendor pick which half they get. If a tool only works on the leads someone hand selected, it does not work.

We cannot guarantee a result from any trial, ours included, because outcomes depend on your market, your staffing and your pricing. What a split test gives you is a number you can defend in a meeting.

Naming specific tools rather than describing benefits, because you cannot model a benefit.

Call recording with transcription turns a twelve minute call into something a manager reads in thirty seconds, which is what makes coaching cheap enough to actually happen. Automations, task automation and follow up processes let you build a cadence once and let it run, which is where most stalled opportunity value sits. Drip campaigns with a drip editor cover the long tail. Reminders and appointments keep the calendar honest.

On the phone side, AutoMail provides IVR, call routing, forwarding and number management, so calls get answered and directed instead of ringing out. Voicemail drop covers outbound at volume. RCS with automatic SMS fallback gives branded messages on supported handsets without excluding anyone.

Document AI reads documents and scans VINs, which removes typing from intake. An email validator and a phone validator clean contact data before a campaign runs, which protects both deliverability and your team's hours.

What we do not sell is a conversational AI voice agent that negotiates an appointment in natural speech. Model that separately if you need it. Our pricing starts at $199 a month for CRM Only, month to month, on the pricing page.

What lift number should I use if I have no data?

Run three cases: zero, a modest figure the vendor will commit to in writing as their low case, and their headline number. If the decision is only positive at the headline number, you are not buying a tool, you are buying an assumption.

Should the calculation use front gross or total gross?

Total gross, front and back, on the vehicle class the tool actually influences. Use a twelve month average from your own reporting rather than last month or a regional figure someone quoted.

How long should a trial run before I trust the number?

Long enough that each arm of the split has a meaningful sample. For a store handling a few hundred opportunities a month, plan on at least four weeks per arm, and prefer a split test to a before and after comparison.

Does LeadLocate publish ROI figures for its own tools?

No. We do not publish lift percentages, close rates or customer counts, because we would be asking you to trust a number you cannot audit. Run the same split test on us that you would run on anyone else.

Is an AI tool worth it for a small independent store?

It depends entirely on your opportunity volume. Below a certain volume the fixed monthly cost cannot be recovered no matter how good the tool is. Run the math at your real numbers before you sit through the demo.

What is the most common mistake in these calculations?

Forgetting to multiply an appointment lift by the show rate and the close rate. Extra appointments are not units. Skipping those two terms typically overstates the return by more than half.

We will walk your volume, conversion and gross through the model with you and tell you honestly if the answer is no. Month to month, no long term contract.