Finding the Developer Bottleneck Worth Fixing

Our internal developer tools organization was in greenfield territory. We knew we wanted to help developers code faster, but not where the biggest pain points were. With limited bandwidth and high expectations, spending time in the wrong place was a real risk.

The challenge

Build a clear, defensible way to decide where to invest: one that combined hard usage data with what developers were actually saying, and that leadership could act on.

What I did

Set a quantitative baseline. I pulled usage across internal wikis, tool pages and download pages to see monthly active visits and traffic by property.
Asked one ROI question of every option. If we invest here, does the developer experience improve per employee?
Pressure-tested the numbers with sentiment. I stacked the data against the volume and tone of Slack comments, documentation questions and feedback on download pages.
Looked for the overlap. The clearest opportunities were where high traffic met high frustration.
Presented a simple story. Here’s what the data shows, here’s what developers are saying, and here’s where time spent will return the most.

One theme kept coming up: application upgrades, and how much time engineers lost updating large numbers of code files. That led to an AI-assisted upgrade tool, and I supported the product strategy decisions around its launch.

What changed

Leadership got a repeatable method for investment decisions in a fast-moving space, replacing gut-feel prioritization. The upgrade work grew into what became AWS Transform. Amazon has publicly credited AI-assisted upgrades with roughly $260 million in annual savings, with more than 10,000 internal applications and microservices updated.

What I took away

Traffic tells you where people are. Frustration tells you where they’re stuck. You need both to know where to dig.