Saving $30M/year with ML-powered self-service
The Red Hat Customer Portal served over 1 million users seeking technical support. Response times were slow, service margins were low, and the team was spending heavily on human support. I was brought in to find a better way.
Background
Users faced multi-week turnaround times for technical issue resolution. The cost of maintaining a large human support team was unsustainable, and satisfaction scores reflected it. The portal had the information users needed, but they couldn't find it on their own.
My role
I led user research to identify where the self-service experience broke down, then ran a vision sprint to define a new approach. I implemented a machine learning algorithm to surface relevant solutions earlier in the support journey, redesigned the UI for self-resolution, and built an A/B testing framework that was later adopted across all Red Hat web teams. I also became the company SME for Pendo (user behavior analytics) and built data pipelines in Python, SQL, and JavaScript.
Outcomes
- ·Saved Red Hat $30M annually through reduced support overhead
- ·Cut 200,000+ support tickets per month (2.4M per year)
- ·25 thank-you emails from users in the first release week
- ·Won the ASP Top Ten Support Websites award in 2019
- ·A/B testing framework adopted across all Red Hat web teams
$30M saved annually
Key outcome
Next project
Lisa AI Leasing Assistant→