Why we built it
I kept watching the same scene at high-volume dealers. The best closer in the company drives forty minutes to sit with a homeowner who was never going to buy, while a ready-to-sign kitchen remodel gets the newest rep on the roster, because the routing decision was made from a zip code and a property record.
The painful part: the company already had the answer. A twenty minute phone call created that appointment, and the homeowner spent it explaining exactly what they wanted, how soon, who decides, and what they were worried about. Then the call ended and every one of those signals evaporated. The score never heard the conversation.
So we built lead scoring that listens to it.
What does third-party scoring miss?
Demographic, property, and credit-band data has real value, and Craft uses it. The ceiling is that it describes the household and says nothing about the moment. It is the same data every competitor can buy, it looks identical before and after your best qualification call, and it cannot tell a motivated buyer from a price-shopper who lives next door to one.
Scoring on third-party data alone
- Static: the score is the same before and after the qualification call
- Commodity: every vendor sells the same records to every competitor
- Blind to intent: describes the household, misses the moment
- Unexplainable: a number arrives with no reasons a manager can check
Scoring on the conversation
- Live: the score moves while the caller is still on the line
- Proprietary: your conversations are data nobody else can buy
- Intent-rich: project scope, urgency, decision-makers, budget signals, objections
- Explainable: every score ships with the reasons behind it
How conversational lead scoring works
Craft trains a custom model for your company, on your outcomes. The inputs come from three places: the conversations themselves, third party consumer and property enrichment, and the closed-loop results of every lead you have already run. More than half of the model's inputs come from the conversation itself.
Because the conversation layer runs live, the score does too. A lead that mentions a second opinion scheduled for Thursday, both homeowners on the call, and a permit already pulled stops looking like every other three-bed in the zip code before your agent hangs up.
Danielle Foster
44% close on similar opportunities · 6 mi away
The score decides. Assignment acts.
A great score that routes to the wrong rep is a spreadsheet curiosity. Craft assigns each lead to the rep most likely to win it, weighing geography and drive time, capacity, lead value, and each rep's track record on that job type. When a stronger lead lands, the day gets rebuilt, and AI agents reschedule the displaced appointments so nobody falls through the crack.
Managers see the reasons behind every match. When the system puts your closer on the Hendersons instead of the Parkers, it tells you why, and you can overrule it.
Marcus Rivera → The NaylorsWalk-in bath · Thu 9:00 AM$14,100Why this assignment
Half the model's inputs come from the conversation itself.
Which means the model gets smarter with every call your team takes, and no competitor can buy their way to it.
What it changes for your team
Your best closers spend their afternoons in living rooms that can actually buy. New reps get leads that match their level while they ramp. The marketing budget stops subsidizing drive time to appointments the phone call already disqualified. And because every score is explainable, the Monday argument about lead distribution gets replaced by a screen everyone can read.
Teams run Craft's scores in parallel with their existing scoring from day one, so the model proves itself against the incumbent on your own closed deals before you change anything.
Why this approach wins
Every scoring vendor starts from the same purchasable records, so scores built only on those records converge on the same answers. Your conversations are the one input your competitors cannot buy, and they compound: every call adds training data, every closed job sharpens the model, and the gap widens with volume.
That is the bet we made. Score the moment, and let everyone else keep scoring the household.
Common questions
Do we have to replace our existing scoring?
No. Many companies keep their incumbent scoring and add Craft as the conversation layer, running both in parallel until the results settle the question.
Where do the conversation signals come from?
From the calls Craft already captures and scores in your contact center, and from recorded in-home appointments. If Craft coaches your conversations today, the scoring model is already fed.