Promotion Approval Workflow
An Airflow-orchestrated approval pipeline with database-backed tracking that cut team effort 80% and about $200K a month in leakage.
- −80% team effort
- ~$200K monthly saving
- Python
- Apache Airflow
- SQL
- REST APIs
Problem
Promotions went through a manual approval process, and promotion-duration limits — how long a discount was allowed to stay active — weren’t systematically enforced. That combination meant approvals were slow and inconsistent, and promotions could run longer than intended without anyone catching it until the impact had already been felt.
Constraints
The workflow had to fit into systems that already existed rather than replace them: approvals, pricing, and promotion data all lived in other tools, so the new workflow had to integrate with them over their existing REST APIs rather than becoming a new system of record that everything else had to migrate to.
Design
Apache Airflow DAGs orchestrate the approval workflow end to end, calling out to the existing systems’ APIs at each step rather than duplicating their data. Workflow state — where a given promotion request is in the approval process — is tracked in a database, which also feeds dashboards so approvers and stakeholders can see the state of pending and active promotions without chasing status by hand. That same tracking layer enforces controls around promotion-duration limits, so a promotion can’t silently keep running past its approved window.
Outcome
The workflow cut the team’s manual effort on approvals by 80% and reduced promotion-related leakage — spend lost to promotions that ran longer or looser than intended — by about $200K a month.