I worked on the order management system at The Pokémon Company International, where inventory had to stay in sync across OMS, 3PLs, and middleware integrations.
What initially appeared as an overselling bug turned out to be a distributed systems challenge. Multiple 3pls, Inventory updates were delayed across systems, and during high-traffic events, queue backlogs amplified inconsistencies.
The Problem
Inventory accuracy depended on loosely coupled external systems:
- Multiple 3PLs with inconsistent data quality and update patterns
- High order volume creating queue backlogs during peak events
- Limited control over the OMS (Fluent Commerce)
This led to inventory being available in one system while already consumed in another, resulting in overselling.
The Approach
A single “fix” wasn’t realistic.
Inventory in this architecture is inherently eventually consistent, while user expectations are real-time. With external dependencies and limited control over the OMS, the goal shifted from perfect accuracy to mitigation, control, and visibility.
What We Built
- Leveraged Fluent control groups to dynamically restrict inventory exposure during high-risk scenarios
- Built an internal tool to simplify complex configurations, like control groups, inventory upserts, enabling faster operational response
- Analyzed cross-system inventory and order flows to identify high-risk mismatch points
- Introduced operational safeguards to reduce overselling during peak load and sync delays
Impact
- Reduced overselling during high-traffic events
- Improved response time for inventory-related issues
- Enabled non-engineering teams to act proactively
Key Takeaway
In distributed commerce systems, designing for control, visibility, and operational flexibility is more effective than chasing perfect inventory syncs, especially when relying on external platforms like Fluent Commerce.