How StubHub International went from silent failures to proactive reliability

Overview

When data breaks, trust breaks. At StubHub International, the data team had invested heavily in dbt for modeling and testing, but had no visibility into test results. That changed with a single incident: a failing test on duplicate transactions that went unnoticed for over a week. This triggered a search for a tool that could bring observability into their dbt workflow, without vendor lock-in or extra complexity. The team found a perfect fit in Elementary.

In this case study, David, Analytics Manager at StubHub International, shares how his team implemented Elementary during a POC and immediately started catching real issues, shifting from reactive firefighting to proactive data quality.

Company Background

StubHub International is a global marketplace for live experiences, enabling fans worldwide to buy and sell tickets to sports, music, and entertainment events.

With operations spanning multiple countries, the company supports a diverse and dynamic user base, delivering seamless ticketing experiences while managing complex, high-volume data systems behind the scenes.

The Challenge

StubHub International had adopted dbt nearly three years earlier and saw clear benefits: reusable code, centralized logic, and consistent outputs across reports and analyses. The team had also implemented dbt tests to catch data quality issues—but there was a critical blind spot.

“We had dbt tests, but we didn’t see the results,” David explained. “They were buried in the Airflow logs, and we had no visibility into them.”

The test results existed—but they weren’t actionable. Most of the team’s dbt tests were configured with severity=warn, meaning they wouldn’t fail Airflow tasks or halt the DAG. This was intentional: not every issue was critical enough to justify stopping data pipelines or blocking updates to dashboards in Tableau.

But without a monitoring layer on top, this design came with a major downside—failing tests went unnoticed. The results were buried in logs, and unless someone manually reviewed them, issues slipped through. What began as a practical compromise quickly became an operational risk.

The final trigger came when a duplicate transaction bug went undetected for over a week, despite having a test in place. The test had been failing silently in the logs the whole time. That’s when the team decided: the current setup wasn’t enough. They needed something better.

“That was the moment I knew we needed better visibility. What’s the point of tests if you don’t know when they fail?”

David began searching for a solution that would provide:

• Easy visibility into dbt test results
• No vendor lock-in
• No additional tools or languages to learn
• Fast setup with low overhead

The Evaluation

The search focused on data observability solutions that “played nice with dbt.” His two non-negotiables were:

• Keep everything in code: “I wanted tests to live inside the dbt project and stay version-controlled.”
• Avoid long setups: “I didn’t want to spend months setting up a tool before getting value.”

While exploring other options, David quickly ruled them out due to added complexity and lack of native dbt integration.

He installed the open-source Elementary dbt package over a weekend.

The POC

The team proceeded with a one-month POC using Elementary Cloud, connecting it to their BigQuery environment and dbt project.

Setup took just a couple of hours. No custom integrations. No engineering overhead.

Key wins during the POC:

• Immediate detection of freshness issues: Elementary flagged outdated tables that would’ve gone unnoticed.
• Test history and performance insights: Helped optimize model runtimes and detect degradation.
• Built-in monitors: Out-of-the-box tests catch anomalies from day one—no configuration needed.
• Improved team workflows: Alerts and insights were sent directly to Slack, enabling faster triage.

The Impact

Just a few months in, the StubHub International data team is already seeing a major shift in how they work:

• From reactive to proactive: They now identify issues before business users notice them.
• Trusted by the business: Their reputation across departments continues to grow.
• No engineering bottlenecks: Setup and monitoring run smoothly without requiring scarce data engineering resources.

Why Elementary Cloud?

While the open-source dbt package provided immediate value, the team chose to go with Elementary Cloud to unlock additional benefits:

• Ease of use: “I didn’t want another integration for my team to have to manage. I wanted something that just works.”
• Out-of-the-box anomaly detection: Automated monitors, powered by ML, started catching issues immediately with zero configuration.

David also appreciated that the dbt-native approach aligned with their development philosophy and avoided vendor lock-in.

What’s Next

The team plans to expand usage of more advanced features, including dimension-level anomaly detection—an area with huge potential to catch silent failures in their fraud models and core transaction pipelines.