korab eland
← all case studies
case study in flight

Lead scoring: fixed tiers to a decile model

Stopping hundreds of low-value calls held conversion steady and earned the mandate to replace fixed tiers with a decile model built on evidence.

role
Led strategy + evidence
team
Admissions, marketing, paid media, data engineering
stack
SQL · CRM · ML pipeline
span
2026
status
in flight

We were calling our lowest-tier leads hundreds of times for every application they produced. When we stopped calling them, conversion held. Applications arrived on their own timeline, and the team got real capacity back. One analysis changed how the business thought about effort.

It also bought the mandate for something better. Fixed tiers treat every lead in a bucket the same. A decile model scores each lead on expected conversion probability, so call strategy, nurture sequencing and media spend can follow the evidence instead of the rulebook. I ran alignment sessions with six stakeholders across admissions, marketing, paid media and data engineering, synthesised where they converged, and took the model to senior leaders. They endorsed v1.

Then the groundwork. I audited the lead signals we actually had and wrote the analysis query myself: around 600k leads with point-in-time features. I set an interpretability requirement so the team could trust the scores, and briefed data engineering. The model build is scheduled behind a data-platform migration, and the whole workstream is documented to run without me as the dependency.

outcome

Call effort on lowest tier, per application hundreds → 0

Before/after comparison of outbound call attempts against the lowest-tier lead bucket, per resulting application, once calling was stopped for that tier.

Dataset assembled for model build ~600k leads

Approximate row count of the lead dataset with point-in-time features, assembled from the analysis query written to scope the model build.