The workflow before
DEMDACO sells through more than 12,000 retail locations, and its reporting ran on an Oracle EBS R12 ERP feeding a Cognos warehouse built in 2009. The first business unit's data mart was never finished. Product data in Centric PLM and sales data in EMUN SFA sat outside the warehouse, so reports disagreed and teams reconciled them in spreadsheets. More than 20 report requests sat in the IT queue, and a new report took weeks.
What we built
We built an analytics stack on AWS and Databricks. Oracle EBS, Centric PLM and EMUN SFA data land in an S3 data lake with raw and cleansed zones, and Databricks workflows clean, standardize and apply business rules.
We finished the first business unit's data mart, then added marts for Willow Tree, Studio, Big Sky Carvers, sales, inventory and finance, with a semantic layer that defines each metric once.
A data quality framework runs validation checks on every load, and we wrote down who owns testing. Cognos now reads the new marts, so business users build their own reports.
Before and after
| Measure | Before | After |
|---|---|---|
| Data completeness | About 85% | Over 99% |
| Report request backlog | 20+ requests | Under 5 |
| Time to build a new report | Weeks | Hours |
| Data issue resolution | Days | Hours |
What's next
Clean, governed sales and inventory marts are the foundation for Level 2 work, starting with demand forecasting and an assistant that answers sales and inventory questions from the semantic layer.
