Most dealership marketing ROI reports answer a question nobody asked in a form nobody can check. They print one number per channel and imply that number is cars sold. It is not, and the agency that built the PDF knows it is not.
The honest answer to channel ROI has two layers. One layer is what the ad platforms report: spend, impressions, clicks and conversions, per channel and per month. The other is what only the CRM and DMS can report: units sold and gross. A report that blends them is a story, not a calculation.
This is how to build the stack correctly, which monthly KPIs belong in it, and where each layer has to say "this is where my view ends."
How Do I Calculate Dealership Marketing ROI by Channel?
Start with the only data every channel hands you in the same shape: what the platform itself reports. For each channel, pull spend by month, impressions, clicks and platform-reported conversions from the dealer's own ad account. That is the left half of the equation, and it is auditable, because the dealer owns the account and can open the same screen.
Then divide, and stop. Cost per click, cost per platform-reported conversion and conversion volume by month are all legitimate channel metrics. They tell you what each dollar bought inside that platform's own measurement.
Now bring in the right half separately. Units sold and gross profit live in the CRM and DMS. Take the month's sold units and gross from those systems, and set them beside the channel table as a second table. Do not paste them into the channel rows unless your records can tie a specific sale to a specific click.
Most stores cannot do that tie reliably, and pretending otherwise is where the damage starts. Total marketing spend against total gross is a defensible store-level return. Spend per channel against platform conversions is a defensible channel-level efficiency read. Spend per channel against sold units is only defensible where a record connects the two.
That is the whole method. Two tables, labeled by their source, with a clearly marked seam between them.
Why Does the Agency PDF Blend Two Different Kinds of Number?
Because the blend flatters the agency. A platform conversion is a form fill, a call or a page event. A sale is a signed deal. Between them sit follow-up speed, desk process, inventory, price and a customer who may have shopped three stores. The agency controls the first and none of the second.

So the PDF shows a cost per lead, adds a closing ratio borrowed from somewhere, and prints an implied cost per sale. The arithmetic is clean. The provenance is not. Nobody in the room can say which sold units that figure refers to.
The same PDF usually settles on whichever channel looks best this month. A budget split fixed weeks earlier then gets defended with a retail claim nobody can trace. We made the related argument about rigid allocation in why a channel split set once cannot keep up with a month: the decision gets worse when the evidence behind it is blurred.
The loser here is the agency PDF that implies sold-unit attribution it cannot show. It is not the dealer's fault for trusting it. The format was designed to read as proof.
What Marketing KPIs Should Car Dealerships Track Monthly?
Split the list by who can verify each item. A short list beats a long dashboard.
Platform-reported, per channel, per month:
- Spend, against the approved monthly budget.
- Impressions and clicks, which show whether the channel is reaching anyone.
- Platform-reported conversions, labeled as platform-attributed.
- Cost per click and cost per platform-reported conversion.
- Month-over-month movement in each, so a drift is visible before it becomes a quarter.
CRM and DMS, store-level:
- Leads received and how long the first human response took.
- Appointments set and shown.
- Units sold, new and used, and gross.
- Lead source as the CRM recorded it, treated as a record of what was typed, not a verdict.
Notice what is not on the list: any single figure that mixes the halves. Total marketing spend divided by units sold is a fine store-level number for the GM. It belongs on the CRM side of the seam, computed from the store's own books.
One more item belongs on the list, and almost nobody prints it: response time. Paid media gets blamed for leads that died in follow-up, a point we made in naming the minute a lead waited before blaming the ad. If the CRM side shows slow replies, a weak channel ROI may be a desk problem and not an ad problem. The platform tables cannot tell you that. Only the CRM can.
Where Does the CRM Take Over, and What Should It Print?
The handoff happens at the lead. Everything before it, the auction, the click and the conversion event, is the platform's record. Everything after it, the response, the appointment, the deal and the gross, is the dealership's record.

A good report says so in plain words, on the page. "Conversions shown are platform-reported. Retail units and gross are not visible to this report and come from your CRM and DMS." That sentence costs nothing and protects the reader from drawing a conclusion the data cannot carry.
On the CRM side, the report should print the numbers the store can count from its own record: leads, first human response, appointments, sold deals, gross. We covered the response figure specifically in the one number a CRM demo never prints. If the CRM cannot show how long a lead waited, the dealer cannot separate a bad ad from a slow reply.
Where a store wants to tie the halves, it can do it deliberately. Match a lead's recorded source and timestamp to a sold deal, in the CRM, and label the result as a match from the store's own records. That is a documented link. An agency extrapolation is not.
What Does an Honest Channel ROI Report Refuse to Say?
It refuses four things.
First, it refuses to say a channel "sold" cars. A channel produced platform-reported conversions. Cars are sold by the store.
Second, it refuses to print gross it cannot see. An ad platform has no idea what a deal earned, and a report built from platform data should say so rather than estimate it.
Third, it refuses to hide missing data. If a window is incomplete, the report should state that, not print a confident document over a gap.
Fourth, it refuses to touch the account while it reports. A report that also changes budgets is grading its own work. Reporting and acting should be separate, with the recommendation written down so the dealer can read it before anything moves.
That last point matters for another reason. AI answer engines and human GMs both favor documents that state their source, their date and their limits. We wrote about what makes a page citable in a page with a date, a source and a check. A channel ROI report built the same way is both more trusted in the meeting and easier to defend a year later.
How AUTONOMi Solves This
AEGIS dealer reports are built from the advertising platforms' own reported daily data over the window the dealer asks for: spend, impressions, clicks and conversions, per channel and per month.✓ Nov 6 A dealer asks in their own words, through chat, email, text or phone, for something like a three-month review or a document to show the GM, and AEGIS researches the account, writes the analysis and emails a designed PDF with a download link.✓ Nov 6
The report states what it cannot see: it does not read the dealer's CRM or DMS, so it reports advertising performance and platform-attributed conversions, never retail units sold or gross. The seam is printed on the page, not left for the reader to discover.
The report makes no changes to any account. It reports and recommends, and other teams act.✓ Nov 6 The narrative, findings and recommendations are written by AEGIS, while the code owns layout and the disclosure rails that refuse to print a confident document over incomplete data.✓ Nov 6
When AEGIS does act on a dealer's written instruction, such as moving budget between channels it runs, AXIOM records each move on the dealer's hash-chained decision trail.✓ Nov 6 That keeps the reporting layer and the acting layer apart, and keeps both on a record the dealer can read. And because the ad accounts are dealer-owned, with AEGIS operating on delegated access the dealer can revoke, every figure in the report can be checked against the dealer's own platform screens.
For the other half of the equation, AUTONOMi's own CRM, PRISM, prints the morning numbers from the store's own record: leads and first human response, appointments, deals sold and gross.
PRISM Reports read every figure from the store's own customer record and timeline, and nothing in them is estimated. Rates, such as the share of leads answered within an hour, are calculated from those counted events.
A store that keeps another CRM brings its own numbers to the second table. The method does not change.Where Does Channel ROI Reporting Go From Here?
The standard is moving toward documents that name their source and their limit. A GM who has been burned by one blended PDF will ask the next vendor a simple question: which of these numbers did the platform report, and which did my own system record? A report that cannot answer in one sentence has already lost the room.
The practical step is small. Pull last month's spend and conversions by channel from your own ad accounts. Pull sold units and gross from your CRM and DMS. Put them in two tables with a visible seam, and compare them to the report your agency sent. The gaps you find are the claims nobody could trace. If you want a platform-sourced report with the seam already printed, sign up and ask AEGIS for one.



