The Monday Morning Number

The Monday sales meeting starts the same way in too many stores. The GM walks in with the DMS report. The desk has its log. The BDC manager has the CRM dashboard, and the vendor portal says something different from all three. The first fifteen minutes go to arguing about whose number is right, and the decisions that actually matter get made on gut feel or pushed to next week. Before your store buys any AI, it needs one number everybody trusts.

The argument is the symptom

Every report in that room is technically right. The DMS counts a sale when the deal caps. The CRM counts it when a salesperson marks it sold. The vendor portal counts whatever flatters the vendor, on whatever date its tracking fired. The numbers disagree because the definitions disagree, and nobody ever wrote the definitions down.

The cost runs deeper than fifteen wasted minutes. When no number is trusted, every decision becomes a negotiation, and the manager with the best story beats the manager with the best week. Ad spend gets reallocated on anecdotes. Underperformance hides in the gaps between reports for months. And when an AI vendor shows up promising clarity, the store buys an eleventh opinion instead of a first agreement.

Picture the gut-feel version of the store for a second. The desk swears the third-party leads are junk. The BDC swears the follow-up is happening. The vendor report swears the campaign is crushing it. Somebody is wrong, and without one number, the loudest voice wins the budget. That is how stores end up paying for the same lead twice and cutting the source that was quietly closing.

That is the trap to skip. You do not have a reporting problem. You have a treaty problem, and treaties are cheaper than software.

Three definitions end the argument

The whole fix rests on three definitions, agreed in writing and signed by the GM, the desk, and the BDC manager. Not inherited by default from whatever a vendor shipped. Agreed.

What is a lead? Pick the line and hold it. A unique customer with working contact information and expressed interest in a vehicle, counted once, no matter how many forms they fill out or how many times they call. Decide now whether service-to-sales handoffs count, whether unsold showroom traffic counts, and which rule merges duplicates. There is no universally right answer. There is only your answer, written down.

What is the sale date? Contract date, delivery date, or funding date. Pick one and make every report in the store use it, even when a system default disagrees. Half the Monday argument usually lives right here, in two reports counting the same deal in different weeks.

What is a source? One list of allowed source names, owned by the store, never by vendors. Ten or fifteen values, no free text. Add a tie-breaking rule: first touch, last touch, or split. Your call, but one call. I built vendor-independent lead-source attribution at Lazare Auto Group back in 2009, with early Google Analytics and cookies, because every vendor wanted to grade its own homework. Stores fight the same fight today with fancier tools, and the defense has not changed: your definitions, applied by tracking you control.

Build it in thirty days with the tools you already pay for

You do not need new software for this. You need four weeks of discipline and the exports your current tools already produce.

Week one: write the treaty. One page. The three definitions above, plus a single name that owns the number. Not a committee, a name. Argue hard this week, because the whole point is to never have the argument again. The GM signs it. The desk signs it. The BDC signs it. Tape it to the wall if you have to.

Week two: build the table. One spreadsheet or one report, fed by exports from the CRM and DMS you already pay for. One row per customer. Columns for lead date, source from the approved list, sale date by the agreed definition, and gross if you want it. Map each system's fields to the treaty once, and document that mapping on a second page. This is an afternoon of work for whoever runs your reports, not a project.

Week three: run it in parallel. Bring the new sheet to the Monday meeting next to the old reports. Every gap between the sheet and a vendor dashboard gets explained out loud, and the explanation almost always exposes either a broken process or a vendor counting generously. Fix the process. Write down what you learn about the vendor.

Week four: make it law. The meeting runs off the number stack and nothing else. Anyone who brings a different number gets sent back to the treaty. By the second month, the fifteen-minute argument is a sixty-second readout, and the meeting is finally about what to do instead of what is true.

Total new software purchased in those thirty days: none.

Where AI fits after the number

Once one trusted number exists, AI earns its seat fast, because AI is leverage on whatever you point it at. Point it at chaos and you get faster chaos. The good news is that the store that did the thirty days above is now the easiest store in town to apply AI to.

The honest sequence looks like this. AI reads your calls and tags sources against the approved list instead of guessing. AI flags duplicate leads before they inflate the count. AI drafts the Monday readout from the table, so the meeting starts with the answer instead of the search for one. At Strolid I built an internal meeting-intelligence pipeline in TypeScript and Python for exactly this class of work, turning conversations into accountable records. And the growth there, monthly opportunities up from an average of 12 to 50 or more rooftops per month with a 40 percent MRR lift behind it, ran on numbers the whole team trusted. The trust came first. It always comes first.

What not to do is just as clear. Do not buy an AI reporting layer to reconcile tools that disagree, because reconciliation without definitions is just prettier guessing. Do not let any AI tool log its results into a private dashboard either: its numbers join the treaty table or it does not get bought. The rule runs one direction. AI feeds the number stack. It never replaces it.

The meeting is the test

You will know it worked by the sound of Monday morning. One page, no argument, decisions made before 9:15, and a team that stopped debating reality and started acting on it. Build the number first. Buy the AI second, and only the AI that feeds the number. If you want to see what this looks like wired into production systems, the work page shows what I build, and pricing shows what it takes to bring it into your store.