A retailer that runs on data
INSIDER was founded in 2017 in São Paulo by Yuri Gricheno and Carol Matsuse around a simple idea: apparel built on technology, design, and sustainability. Today the brand serves more than 1.2 million customers across more than 50 countries, and data underpins nearly every part of the operation, from revenue tracking and customer behavior to marketing performance and leadership decisions.
As the company scaled, the role of data expanded. What began as a supporting function became the foundation for day-to-day operations and long-term growth. The data team supports this by enabling teams across the business to access and use trusted metrics.
A business moving faster than its platform could follow
Before moving to the dbt platform, INSIDER's data transformation workflows ran on dbt Core through an external consulting firm. The external team handled orchestration, pipeline execution, and platform maintenance, but as INSIDER's growth accelerated, the platform's response time couldn't keep pace.
"When something broke, we couldn't investigate it ourselves. We had to rely on the external team to understand what happened and tell us how to fix it. That made it harder to move quickly and harder to build knowledge within the team," says Rafael Bernardes.
During critical periods like Black Friday, the team needed near real-time data to track performance and react fast. The external operating model added steps between the problem and the fix, slowing response times when they mattered most. Limited cost visibility compounded the issue: overall spend was visible, but the team couldn't trace which models or workloads were driving it, so inefficiencies went unresolved.
A change in the external provider landscape then introduced a fixed migration deadline, turning a growing limitation into an immediate decision point.
A migration nobody noticed
With Fivetran + dbt Labs, INSIDER brought its data transformation workflows in-house using the dbt platform. A small internal team led the migration, taking ownership of pipelines, models, orchestration, monitoring, and debugging, all consolidated into a single system they could operate directly.
One of the most immediate changes was visibility and control. Engineers can now investigate failures and deploy fixes without waiting for external support.
They also gained visibility into costs they couldn't view before. Model-level cost tracking and pipeline execution data let the team trace exactly which workloads drive warehouse spend, so inefficiencies that previously went unresolved are now identifiable and fixable.
The migration also changed how the team operates day to day. INSIDER introduced an on-call process and shared ownership of the platform, enabling analytics engineers to diagnose and resolve issues directly as part of their normal responsibilities.
The team completed the migration and validated the new environment by running production workloads in parallel. Business users didn't notice the switch, confirming the platform was ready before the legacy environment was retired.
"A year ago, we were completely dependent on an externally operated setup. To look back and see that we made the migration in two months and now run it ourselves is a real achievement." - Rafael Bernardes, data manager, INSIDER Store
What ownership actually changed
Instead of rebuilding entire upstream pipelines to test a change, engineers now validate only the models they've modified, cutting CI feedback time.
"Before, our CI could take 30 minutes or even hours because we had to rebuild entire upstream pipelines to validate a change. With dbt platform, we only validate the models we've modified, and most checks now run in around 30 seconds," says Bernardes.
That speed change also reduced compute costs. Fewer queries per deploy meant lower warehouse spend on every cycle. The cost visibility the team gained through model-level tracking had an even larger impact. Once the team could trace which workloads were driving spend, they identified and cut inefficiencies that reduced BigQuery costs by approximately R$10k-25k per week during the period analyzed.
The migration also changed how the team thinks about operating the platform. Rather than depending on external support or hiring a dedicated platform engineering function, analytics engineers now share responsibility for monitoring, troubleshooting, and improving the platform.
From trusted metrics to a trusted foundation for AI
INSIDER plans to deepen its use of the Semantic Layer to standardize metric definitions across the business, so teams outside of data can work from the same trusted numbers without depending on the data team to pull them.
The team is continuing to strengthen its data governance practices, including data classification, access management, and monitoring. As more teams rely on data to make decisions and power customer-facing experiences, that governance foundation is what makes the next set of use cases possible.
INSIDER is increasingly using trusted data to support recommendation and personalization use cases, and the team is building the governance and context required for future AI applications.
The team is evolving from building data primarily for human analysis to creating a trusted data foundation that also supports AI applications, and dbt is an important part of that foundation.
