The problem
BuyBay takes returned goods from retailers, grades them, and sells close to 98% of them on. That is a warehouse business with a software problem: the warehouse-management system set the ceiling on how fast items moved from dock to listing, on how much the operation could see, and on how much it cost to run. Growth was pushing against that ceiling, and the system had become the thing every improvement waited on.
What I decided
Replace it, in one programme, with a transport-management system and a low-code process layer alongside, rather than patch the old one. As Strategic Technology Leader I owned the business case, ran the request for information and the request for proposal, selected the platform, and was accountable for the technical implementation, the hardware, and the change management on the floor.
The low-code layer was a deliberate choice: business-process automation that the operation could change without a development cycle, inside a hybrid multi-cloud strategy, so that the warehouse system did not become the new bottleneck the day after go-live.
How it went
- Business case to go-live in six months.
- Implementation of the new warehouse and transport systems with integration into the existing landscape, and a cut-over over one weekend so the operation did not stop.
- Several iterations after go-live on the processes themselves, which is where most of the efficiency came from.
The result
| Measure | Outcome |
|---|---|
| Operating cost | Down 20% |
| Operational efficiency | Up 15% |
| Return on investment | Within 7 months of go-live |
| Observability of the value chain | Better, and the basis of a separate observability programme |
These are the figures the business tracked. They are mine to state as the person accountable for the programme, not independently audited, and the method behind the efficiency figure is the operation’s own throughput measurement before and after.
What it taught me
Replace, do not patch, when the system is the ceiling. Six months of focused replacement cost less than the slow bleed of a system everyone had learned to work around.
Put the change capacity in the operation’s hands. The low-code layer mattered more than the warehouse system itself, because it moved process improvement from the development backlog to the floor.
Measure before, or you cannot claim after. The payback figure exists because the baseline was recorded before the project started. That habit is the reason every AI claim on this site carries an evidence grade.