Machine Learning in Production
Real-time lead scoring turned rejected inquiries into net-new revenue worth ~1% of company top line .
One scoring engine, two live sales treatments — the client’s first real-time ML model, deployed within 9 months
One scoring engine, two live sales treatments — the client’s first real-time ML model, deployed within 9 months
- Client: National consumer referral service
- 9-month engagement
- 2 consultants
- Scope: data enrichment, modeling, use-case testing and rollout
The client runs a high-touch sales model with limited specialist capacity. Web inquiries arrive with only sparse self-reported data — no objective read on a lead’s value or ability to pay — so an entire cohort was rejected outright, including leads the client could have served profitably.
~1%
of company top line in net-new annual revenue from the lead-recovery treatment
+50%
referral conversion lift in the secondary closed-lead nurture pilot vs. control
2
live sales treatments on the client’s first real-time scoring engine
90-day referral conversion — closed-lead nurture pilot
Control
0.4%
Nurture
0.6%
Recovered leads — auto-rejected inquiries that the model scores as high-value — convert at ~40% of the core-treated rate: profitable volume under the high-touch model, from a cohort that was previously discarded.
