This takes about two hours and needs nothing you do not already have.
The setup
Pull ten leads from last month that did not close. Random, not cherry picked, not the ones anyone remembers. Open them one at a time and read each like a story.
The six questions
1. Where did it come from? Campaign, list, referral, inbound. If you cannot tell from the record, stop and write that down. It is finding number one and it means attribution is currently impossible.
2. How long from arrival to first real attempt? Not an automated acknowledgement. A genuine attempt to reach the person. Write the number in minutes or hours.
3. What does the record say happened? Can you reconstruct the conversation from what is written, or would you have to ask someone?
4. What was promised? A callback, a quote, a timeline. Then check whether it happened.
5. Who owned it, and did that change? Ownership gaps and silent reassignments are where leads disappear.
6. Where does the story stop? The last thing that happened, and what was supposed to happen next.
Reading the result
Sort the ten by where the story stopped. The biggest pile is your leak.
Stopped before first contact. Routing and coverage. Structural, cheap to fix, and expensive to leave.
Stopped after one contact. Follow up depends on memory.
Stopped at a handoff. Something got lost between two people or two systems.
Stopped after a promise. The worst pile, because those customers remember.
No story at all. The records are too empty to reconstruct. This is a bigger finding than any individual leak: a process you cannot reconstruct is one you cannot improve, automate or defend.
What to do with it
Pick the largest pile and fix one thing in it this month. Not a platform, one thing. Then run the same audit again in six weeks with a fresh ten.
If the pile moved, you have a method. If it did not, you fixed the wrong thing, which you now know in six weeks rather than in a year.
Why this beats a dashboard
Dashboards tell you what happened in aggregate. Ten records tell you why, and the why is what you can act on. Non deterministic and human systems alike have to be examined by looking at real cases rather than by reasoning about them from summary numbers.
The part most teams find uncomfortable
It is almost never lead quality. That is the explanation everyone arrives with, and ten records usually dismantle it in the first hour.