Applied AI · Full-Stack· GTM Lead Scorer · gtm-lead-scorer.onrender.com

GTM Lead Scorer

Open the tool ↗
What It Does

A lead-scoring and routing system that turns raw inbound-lead data into a ranked, explainable action queue for sales reps: who to work, how hard, and why. It scores a single lead or a whole uploaded CSV, and degrades gracefully on the messy exports reps actually paste in — bad rows are flagged and still ranked, never dropped.

Every score is explained with each trait’s historical close rate, so a rep can see exactly what pushed a lead up or down. A manager can tune the tier cutoffs to match team capacity, deciding how many leads land in each action bucket.

How It Works

It’s an intake-only logistic model, forward-validated so it scores leads the way a rep sees them at the moment they arrive — nothing that would only be known later leaks into the score.

I built it against a real, confidential B2B sales funnel. This public version runs on synthetic data generated to match that funnel’s exact format, so the model, the tool, and the approach are the real ones and only the underlying data is a stand-in.

It’s live at gtm-lead-scorer.onrender.com.

GTM Lead Scorer ranked queue: every lead scored, tiered, and ordered by win chance
The ranked lead queue — every lead scored, tiered, and ordered by win chance; unscoreable rows flagged, not dropped
GTM Lead Scorer single scored lead with each trait's historical close rate
A single scored lead — each trait’s contribution shown against its historical close rate
Stack
scikit-learn
Logistic regression scoring model, forward-validated on intake-only features so it scores a lead the way a rep sees it on arrival.
pandas
Parses single leads and whole uploaded CSVs; handles the messy real-world exports, flagging bad rows and still ranking them rather than dropping them.
Flask
Backend and web tool: the ranked action queue, per-lead score explanations, and the manager view for tuning tier cutoffs.
Python
End-to-end — data generation matched to the source funnel’s format, feature handling, scoring, and the close-rate explanations behind each score.
Open the tool ↗ ← All work