# A solar call center dials the leads a model says will convert

A propensity model trained on past customers sorts leads green, yellow and red so setters stop dialing the bottom deciles first.

Category: Prospecting and first touch
Industry: Solar
Difficulty: moderate
Updated: Oct 18, 2026
URL: https://revstackai.co/use-cases/predictive-scoring-for-solar-call-center

## Problem

Rules-based scoring used a few static data points. A lead's priority never changed, and setters needed 180 connected calls per scheduled appointment.

## How AI solves it

Train a lead score on existing customers plus third-party consumer data.

Bucket leads into green, yellow and red.

Keep the same call cadence and measure appointment rate by bucket.

Refresh scores as new customers and lead behavior come in.

## Verdict

33% fewer outbound calls to hit the appointment goal.

## Tools

- [Faraday](https://faraday.ai/): Consumer data and propensity models for lead scoring.

## RevStack prompt

Written by RevStack. This is not the sourced result below.

Inputs:
- Lead fields you have
- Past win/loss export summary
- Current calls per appointment baseline

Help a solar call center manager design a test of predictive lead scoring.

Use only the inputs. Do not invent lift numbers.

Fields: {{fields}}
History: {{history}}
Baseline: {{baseline}}

Write: how to bucket leads, how to keep cadence equal across buckets, what to measure weekly, and when to call the test.

Example output, written by RevStack, not a reported result:

Bucket by score decile: top 2 green, next 2 yellow, rest red. Same 6-touch cadence for all. Measure connected calls per set appointment by bucket weekly. Call the test after each bucket has at least 50 appointments.

## Result

'the call center needed 33% fewer outbound calls to hit their scheduled appointment goal'; green group '+191% lift in conversion rate over baseline'

Vendor case study. Reported by [Momentum Solar / Faraday](https://faraday.ai/blog/solar-call-center-lead-conversion-boost).

## Setup

1. Export won and lost leads
2. Agree on a baseline (calls per appointment)
3. Run a bucketed test before changing cadence