The argument you've had eleven times
The vendor rep says your closing ratio is below their network average. Your salespeople say the leads are junk — bad numbers, tire kickers, people who already bought. The GM asks what you want to do about it, and somebody suggests cutting the source and moving the money.
Then you cut it, spend the money somewhere else, and three months later you're having the same argument about the new source.
The problem is that nobody in the room has separated two different questions:
- Were these buyers?
- Did we work them like buyers?
You cannot answer the first until you've answered the second. An execution audit answers the second, on paper, in about two hours per source. Do it before you touch spend.
Pull a real sample, not the ones anybody remembers
Take 30 leads per source from a closed month — 60 days back, so outcomes have settled. Random, not hand-picked. If the source sends fewer than 30 a month, take 60 days' worth.
Exclude nothing at the start. Duplicates, bad numbers, and out-of-market leads are findings, not exclusions. You want to know what percentage of a source's volume is unworkable, and you want that number to come out of the audit rather than out of a rep's memory.
For each lead, you're collecting five columns. That's the whole audit.
1. Time to first meaningful attempt
Not time to auto-response. Not time to "worked" status. Time from lead receipt to the first outbound call, text, or personal email from a human.
Log it in minutes. Then sort the sample and look at the shape, not the average. An average of 22 minutes can hide a distribution where half the leads got touched in four minutes and the other half sat until the next morning shift.
What you're looking for: does the slow half correlate with a time of day, a day of week, or a specific rep? Sunday-night leads at a store that opens at 9 Monday are a staffing question, not a lead-quality question.
2. Contact attempts and channel mix
Count the attempts across the full life of the lead, and count them by channel. Two calls and two texts is a different effort than four calls to the same number at 2 p.m. on consecutive days.
Also count attempt spacing. Four attempts in the first 90 minutes followed by silence is a burst, not a cadence. It reads as effort in the CRM and reads as nothing to the customer.
Here's the version that works in an internet lead audit: for each lead, write down attempts in the first 24 hours, attempts in days 2–7, and attempts after day 7. Three numbers. You'll see the pattern immediately — most stores have a fat first day and almost nothing after day three.
3. CRM context quality
This is the column most people skip, and it's the one that decides the argument.
Read the notes. For each lead, mark one of three:
- Context captured — you can tell from the notes what the customer wants, what they're driving now, their timeline, and what happened on the last contact.
- Activity only — "LVM," "called no answer," "sent email." Log entries, no information.
- Empty — nothing beyond the auto-generated entries.
A source with 40% "activity only" and 30% "empty" is not a source you can evaluate. Nobody learned anything about those customers, so nobody can tell you whether they were buyers.
This is also where a CRM follow-up audit stops being a compliance exercise. You're not checking whether reps logged something. You're checking whether the next person to touch that lead would know what to say.
Weak note: "Called, LVM. Will try again."
Usable note: "Reached her — she's in a 2019 Highlander, lease up in March, wants a third row and won't go over $650. Husband decides. Asked for Saturday 11, she said call back after she talks to him Thursday. Set task for Thursday 4 p.m."
4. Was an appointment actually asked for
Binary. Not "did the rep try to build rapport." Did a specific day and time get proposed, in writing or on a recording?
Then a second column: how many times. One ask on the first call, then nothing, is very common and it's the single biggest hidden driver of weak automotive lead conversion.
If you have call recordings, listen to the first connected call on 10 of the 30. You will hear the difference between these two endings, and the difference is worth more than any spend decision you could make this quarter:
"Okay, well, feel free to come by anytime this week, we're here till eight."
"I've got the Highlander and one other three-row that'll fit your payment. Are you better at 11 tomorrow or around 6 after work?"
5. Outcome, with a reason attached
Sold, appointment shown-not-sold, appointment no-show, contacted-no-appointment, never contacted, unworkable (bad number, wrong market, already purchased elsewhere).
"Never contacted" and "unworkable" need to be separate. Those two categories are the whole fight. If 8 of 30 were genuinely unworkable, that's a vendor conversation. If 8 of 30 were never contacted, that's a Tuesday morning conversation with three reps.
Read the grid
Now put your two sources side by side and compare the same five columns. A few patterns show up over and over:
Same execution, different outcomes. Response times match, attempt counts match, appointment asks match, and one source still converts at half the rate. Now you have a real dealership lead quality finding. Take it to the vendor with the grid, not with an opinion.
Different execution, different outcomes. The weaker source got a 40-minute median response and 1.8 attempts, the stronger one got 6 minutes and 4.2. You don't have a lead problem yet. You have an assignment or a cadence problem, and cutting spend would hide it.
Same source, different reps. This is the most common result. Two reps, same source, same volume, and one has a 15% appointment rate while the other has 4%. Nothing about spend fixes that.
Unworkable concentration. If one source sends 25% bad phone numbers and the other sends 5%, that's measurable and it's a price negotiation. Bring the count and the specific records.
Assign fixes before you touch the budget
Every row that came back weak gets a name and a change. Not "improve follow-up."
- Sunday and after-hours leads sat until morning → one rep owns the after-hours queue until 9 p.m., checked at 7:30 p.m. by the manager on duty.
- Reps averaging 1.8 attempts → cadence of 6 attempts over 10 days, mixed channels, and the manager reviews day-4 and day-8 activity every Wednesday.
- "Activity only" notes above 25% → four required fields on every contacted lead: current vehicle, need, timeline, decision-maker.
- Appointment asked once or not at all → every connected call ends with two proposed times. Pull three recordings per rep per week and score it.
Give it 45 to 60 days, then rerun the same 30-lead audit on the same source. If execution moved and conversion moved with it, you saved the spend. If execution moved and conversion didn't, now you've earned the right to cut the source — and you'll know the next source won't have the same problem.
That second audit is the part everybody skips, and it's the part that ends the argument permanently. If you're pulling call recordings and CRM notes by hand, scoring the same five columns consistently across reps is where tools like MoreSignal save the afternoon — but the grid matters more than how you build it.
The one-line rule
Never change lead spend based on a closing ratio alone. Change it based on a closing ratio plus proof that the leads were worked the same way.
Anything less, and you're just moving the argument to a new invoice.