All case studies
Brightmoor Retail Group, Retail and CPG
Allocation decisions that reach the store floor
Demand signals now land in the replenishment system overnight instead of in a Monday morning slide.
- on one demand signal
- 600 stores
- Platform
- Apache Kafka, Databricks, dbt, Qlik
The situation
Demand signals existed but arrived as a Monday morning report. By the time an allocation decision was made, the week it described was already half gone.
What we did
- Moved point-of-sale ingestion from a nightly batch to a streaming pipeline, so store-level sales are queryable within minutes.
- Published a single demand signal as a contract that the replenishment system consumes directly, instead of a report a person reads and retypes.
- Gave category managers an override path with an audit trail, because a forecast that cannot be argued with does not get used.
Where it landed
Allocation now runs overnight against the current week. The reporting pack still exists, but it describes what happened rather than driving what happens next.