TECHNOLOGY LEADERSHIP
Why I built a local OMS for Decathlon Japan
Choosing to build an order management system for e-commerce and wholesale, with AI-assisted development and room for further automation.
- My role
- CTO
- Status
- Career case study
- Period
- 2025–present
Career experience at Decathlon Japan. Not a KumaBits client engagement.
When I returned to Decathlon Japan as CTO, wholesale was becoming our largest channel. We also wanted to invest again in e-commerce and preserve our direct relationship with customers. My responsibility covered the technology behind both, working alongside an e-commerce leader and with a small technology team.
One requirement made the challenge tangible. A product could be in our warehouse, allocated to one business unit, while another unit needed the same item for an order. Letting that second order consume the stock could undermine a commitment elsewhere in the business. We needed the software to respect those distinctions.
The order management system, or OMS, we needed had to serve decathlon.co.jp, Rakuten and wholesale. It had to preserve the e-commerce operations we had previously handled through Next Engine, add wholesale flows and support integration with a new logistics partner.
Much of our technology environment had developed around supplying stores we operated ourselves. Our wholesale partners brought their own requirements for receiving goods. Cross-docking meant handling products arriving from production and different warehouses for onward delivery. Additional labelling and other value-added services, or VAS, meant that preparation depended on who would receive the shipment. Stock isolation, routing and preparation all had to work together.
I also wanted the OMS to support further automation. Connecting it with other applications, including our EDI flow for exchanging information with wholesale partners, and eventually with agents was part of the ambition from the start.
There were several ways to respond: adapt more of the operation to an existing tool, extend an existing solution, or build locally. Existing software could support individual requirements such as cross-docking. What mattered to me was how we could make the whole flow work and change it as wholesale developed. If important rules remained outside the application, people would have to remember and apply them at each handoff.
A local build would let us develop those flows in Japan, close to the people using them. It would also put another application on our team's list of things to maintain and understand. I was choosing responsibility for the rules and for what happened when we changed them. The part I had to convince myself about was whether we could carry that responsibility with a reduced team and a tight deadline.
AI-assisted coding helped me reach that confidence. Its support with implementation made the development effort feel achievable within those constraints. It influenced my decision to commit to a project I would also be building myself.
I chose the local build and contributed directly to it. During development, we had to make the operational decisions precise enough to implement: which stock an order could use, what instructions the warehouse needed and what should happen when a delivery arrived short. Those were questions we had to answer from our understanding of the business.
The OMS now supports our e-commerce and wholesale operations within that local scope.
One detail in the warehouse integration shows what the work involved. When a cross-dock receipt is finalised, the application can prepare an outbound draft using the quantities recorded as received and retaining the carton references. If the receipt is marked as having discrepancies, the draft carries an alert to verify quantities before shipping.
The next order is prepared from what was recorded as arriving. The discrepancy remains visible for attention. That is a small piece of behaviour, but it connects what happened in the warehouse with what we can send onward. Being involved in the build let me follow the requirement through to that level of detail.
That same discrepancy is a useful example of where I want to go next. An agent could help investigate it and prepare a proposed response for the team. A closer EDI connection could bring partner orders into the OMS with fewer manual handoffs. My intention is for those applications to use defined operations in the OMS, with the same stock and fulfilment rules applying whoever initiates the work.
For now, it is still early to judge the full impact of this work, and the OMS is one part of a wider operational effort. We are already seeing fewer incidents and delays in e-commerce shipments, while wholesale operations are beginning to ramp up toward ten times last year's level.
← Back to work