Business Intelligence Engineer II · AWS Marketplace · In Progress

Predicting renewal risk before it becomes churn

Propensity modeling · Predictive scoring · Proactive decision support

Problem

When an AWS Marketplace Private Offer is approaching expiration, the teams responsible for renewals have no reliable way to know which customers are likely to walk away and which will renew on their own. Prioritization defaults to deal size and subjective judgment — which means at-risk customers get missed, and time gets spent on customers who would have renewed without any intervention.

The cost is asymmetric: a missed at-risk renewal is far more expensive than an unnecessary outreach. But without a signal, every expiring offer looks the same.

Goal

Build a predictive model that scores each expiring offer with a propensity-to-renew score, giving teams a reliable, data-driven signal to prioritize their time and intervene where it actually matters.

Approach

Rather than improving renewal reporting after the fact, I shifted the problem upstream: from tracking what happened to predicting what will happen. The model scores each expiring offer on a 0–1 scale, where 0 indicates very low renewal likelihood and 1 indicates high likelihood, enabling teams to focus effort on the customers most at risk.

Why This Matters

This project represents a meaningful shift in how I think about data products. Dashboards answer "what happened." Predictive models answer "what should I do next." The goal isn't better reporting. It's better decisions, made earlier, with less manual effort.

It's the same principle behind the AI narrative system I built for Marketplace reporting, and behind Lu at Akaani: the most valuable systems don't wait to be asked. They surface the right signal at the right moment.

Status

Currently in development. Model design and feature engineering are underway. Results and impact metrics will be updated as the project progresses.