Takeaway: if your lead score is one number, it measures your CRM hygiene, not your buyer.
Split "who they are" and "what they do" into separate axes - and your database starts talking.
How does your scoring work - one score or axes?
We built lead scoring for a client and almost shipped it with a fatal flaw:
a hot lead with an empty CRM card scored 48. A cold lead with a perfect card scored 40.
Same segment. Same priority for the sales team.
Results on the live base (724 leads):
26 "sleeping diamonds": perfect fit, faded contact. Invisible to a single score.
49% of active leads waited 24h+ for a first touch.
The head of sales now gets a daily grid: call / nurture / reactivate.
The expensive question now isn't writing the code. It's answering honestly: am I automating this process as-is, or rebuilding it from scratch around AI?
That second one is a lot more work and a lot less fun. Which is exactly why people skip it.
When factories first got electricity, it did almost nothing for twenty years.
They just ripped out the steam engine and dropped in one big electric motor. Same shafts running across the ceiling, same belts, same floor plan. New motor, old factory.
For most companies "adopting AI" means bolting a big motor onto the old factory. Wire a bot into a process nobody has touched in years. The box is checked. The gain isn't there.
Building an AI quote builder with a client right now. Field notes, not a success story.
They sell lighting. Every sales quote took ~2 hours by hand: find models, pull prices, calc wattage, format.
Goal: one button, 3 minutes, done.
So right now I'm running a blind test: the same brief built by 10 different models, comparing which one wins.
7 iterations and ~150 tests went into one thing: the pipeline not lying about numbers.
No results yet. Still monitoring - might be clean.
If you pay for leads or quals: who defines the metric?
If it's the party you pay, and nobody checks it against the CRM - you're paying for their honesty, not for results.
#PerformanceMarketing#Sales#CAC
The fix:
- Source of truth β CRM, not self-reports. A qual counts when QC confirms it from the call recording.
- Fraud reconciliation: cross-check rejected leads for repeats and patterns. Inflation leaves traces.
- KPI on confirmed quals, ideally on sales.