← Selected work
Case study / Applied machine learning

ZIP-code lead prioritization

A regression system using public demographic data to help a home-services call center prioritize incoming leads.

Production context
9+ years in production
Production context
Home services
Production context
Operational prioritization
The challenge

The operating problem.

Granular outcome data was limited, but the operation still needed a defensible way to prioritize scarce call-center attention.

The engineering

What was built.

A Tobit-regression model combined American Community Survey signals with the available operating data and was integrated into the lead workflow.

Outcomes

What the system made possible.

  • A practical ranking signal despite imperfect source data
  • A model simple enough to understand and operate
  • Long-running production use across changing business conditions
Have a related problem?

Talk directly with an engineer.

Bring the messy version. We will help identify the smallest credible path to production.

Talk to an engineer