Dexcom: Automating Backstage Catalog Completeness with Roadie

When Dexcom set out to improve the completeness of their software catalog in Backstage, they weren’t just solving a technical problem – they were tackling one of the most persistent blockers to platform maturity. Like many organizations, Dexcom had a sprawling GitHub organization with hundreds of repositories, but not all were onboarded into Backstage. That made it difficult to understand software ownership, run meaningful checks, or drive improvements at scale. After a concerted push, catalog completeness improved from 60% to over 95%, bringing nearly all of their IT repos into view.

From manual overhead to intelligent automation

At the center of this transformation has been Natalie Brooks, a platform engineer at Dexcom, who took a hands-on approach to solving the problem. Working within the IT organization within Dexcom, Natalie found herself repeatedly chasing down ownership info and trying to figure out which IT teams were responsible for long-abandoned or unclear repositories. The existing process was frustrating, unsustainable and didn’t scale.

Rather than relying on individual engineers to create and populate catalog-info.yaml files for the various repositories, Natalie wrote a Python script to automate catalog ingestion. The script pulled data from the GitHub API to surface repo metadata – contributors, admins, last committers, and more. It opened each repo in the browser so she could make a quick visual judgment when needed. She created a simple interactive flow in the terminal where she could select the most likely owner, set lifecycle and system metadata, and confirm the description. If none was provided, the script defaulted to using the repo name.

Once the input was validated, the script would generate a catalog-info.yaml using a simple template and inject it directly into the repo. It committed the file straight to the default branch, with CI actions disabled to avoid triggering unnecessary pipelines.

It was either herd cats or automate it,” she said. “I was going into every single repo trying to copy-paste the same info or track people down. I figured I could either do this a hundred times, or make it easy on myself and do it once, well.

The results came quickly: what started as a side project led to a systematic onboarding of nearly every repo in the IT org within a matter of weeks.

A top-down approach to onboarding

Rather than waiting for individual teams to take action, Dexcom took a top-down approach, a strategy that echoes guidance shared in Roadie’s adoption journey framework. However, the change wasn’t done without due consideration of including the broader organization – Natalie posted a banner in GitHub alerting teams to the incoming catalog-info.yaml files, with a clear note that they wouldn’t affect production code. She coordinated the rollout carefully, using her internal ownership of the GitHub org to make changes without disruption.

The result: catalog completeness surged past 95%, giving Dexcom a clear, organization-wide view of their software landscape and unlocking the full potential of their developer portal.

Preventing drift with the Scaffolder

To prevent regressions, Dexcom now enforces repo creation through Roadie’s Scaffolder. Repositories can no longer be created through GitHub directly. Instead, every new repo creation must be done via a software template that requires users to define ownership, resource type, and lifecycle up front, all of which is captured in an appropriate catalog-info.yaml. This ensures all new projects are compliant from day one – no more retroactive fixes.

Unlocking governance with Tech Insights

Dexcom had already been using Tech Insights to track software standards across the organization, but the dramatic improvement in catalog completeness has made those checks far more meaningful. With nearly every repository onboarded, scorecards now provide a full and fair view of team-by-team performance.

KPIs like repo safety and hygiene are reported directly to leadership, and team-level reporting has introduced a new level of accountability.

If a team shows up in a KPI meeting because they’re dragging down our safety score, that drives change,” said Natalie. “And the best part is they go into Roadie to fix it.

Building a foundation for automation and improvement

Dexcom is now turning their focus to remediation. Natalie plans to create scaffolder templates that help teams fix issues automatically, like enabling branch protection or adding READMEs, to further improve standards across the org.

By combining catalog automation, template enforcement, and governance via Tech Insights, Dexcom has created a platform strategy that’s scalable, measurable, and deeply pragmatic. It didn’t require mass buy-in upfront. Just a smart script, some well-placed nudges, and a strong desire to remove friction from the developer experience.