To evaluate more than 5,000 FBA companies, acquire over 100 top-rated brands, and manage over 14,000 category-leading products.
First, we used a Jupiter Notebook in SageMaker, where we did our Ad-Hoc research process, evaluated data, and built models.
Next, we collected data from multiple sources such as Amazon seller central, Walmart, Shopify, direct store sources, and proxy vendor API targets such as Rainforest, Fivetran, Rakuten, etc.
There are 3 components that come from the research process: data processing steps, models, and visualizations.
We then converted these steps into a production-grade code in GitHub Repos, deployed with the CI as Airflow data pipelines or models and visualizations in Plotly Dash.
- Provided Thrasio with a reliable data-driven foundation for decision making.
- Allowed M&A, Marketing, SP&A, and Brand Management teams to utilize a single source of Truth.
- Enabled Thrasio to proactively manage and expand its portfolio of products at scale.