Run the Peak-Hour Test
Warehouse automation covers conveyors, sortation, goods-to-person systems, autonomous mobile robots, vision-guided arms, and the software that schedules all of it. Most of these perform well in a quiet aisle, which is why quiet-aisle evidence is close to worthless.
The Bot Scout peak-hour test evaluates every proposed system during the busiest hour the building actually runs: full aisles, competing traffic, a charging queue, an inbound surge, and at least one item the system cannot handle. A deployment that has never met those conditions has not been tested, it has been demonstrated.
The failure mode that surprises operators is not a robot breaking. It is a robot waiting: for a blocked aisle, for a charger, for a human to clear an exception. Utilisation collapses long before reliability does, and utilisation is what the business case was built on.
| Work to remove | System category | Peak-hour question | Cost line usually missed |
|---|---|---|---|
| Long picker walking routes | Goods-to-person or AMR fleet | What happens when two routes meet in a full aisle? | Floor marking, traffic rules, charging footprint |
| Repetitive case movement | Conveyor and sortation | What is the recovery when a jam occurs mid-shift? | Maintenance access and downtime cost |
| Item picking | Vision-guided arm | What is the intervention rate on the hardest SKUs? | Perception tuning and exception staffing |
| Inventory counting | Mobile scanning or drone systems | Does counting run while the aisle is in use? | Data reconciliation with the WMS |
| Coordination and scheduling | Fleet or warehouse execution software | Who arbitrates when priorities conflict? | WMS integration and interface work |
Scope the Software Before the Hardware
The recurring integration cost is the interface between fleet software and the existing warehouse management system. It is quoted late, discovered mid-project, and it sets the schedule; the same pattern is visible in the vertically integrated approach described in the Amazon Robotics guide.
Fixed-station work belongs in a different comparison. If the job happens at one reachable place, compare arms and cobots first using the manufacturing robots guide and the cobot guide.
Ask every vendor for the exception path in writing: which items the system will refuse, how those reach a human, and what the expected intervention rate is at your SKU mix rather than a reference customer's.
- Pilot during the busiest hour, not a quiet shift.
- Measure utilisation and waiting time, not just uptime.
- Scope the WMS interface before signing for hardware.
- Get the exception path and expected intervention rate in writing.
- Budget floor preparation, charging, network coverage, and traffic rules.
Safety and Staffing in a Shared Building
A warehouse is a shared space with pedestrians, forklifts, and deadlines. OSHA notes that many robot incidents occur during non-routine work such as maintenance, setup, and testing, which in a distribution centre means the hours when a fleet is being recovered.
Staffing does not disappear, it changes shape. The roles that appear are fleet recovery, exception picking, and someone who owns the data the system produces. A plan that removes headcount without naming those roles is not finished.
Compare the industrial framing in the industrial robots hub when the same building also runs fixed automation, since the two are usually assessed by different teams.
Bottom Line
Warehouse automation succeeds when it is proven at peak, integrated with the WMS, and staffed for exceptions. Judge utilisation and waiting time rather than headline throughput.
Run every candidate system through your busiest hour before comparing throughput claims.
FAQs
What is the biggest hidden cost in warehouse automation?
Integration with the existing warehouse management system. Fleet coordination software has to exchange data with the WMS, and that interface work is routinely scoped late and drives the schedule.
How should a warehouse pilot be run?
Run it during the busiest hour with full aisles, competing traffic, a charging queue, and at least one item the system cannot handle. Measure utilisation and waiting time, not only uptime.
Does warehouse automation remove staff?
It changes the roles more than it removes them. Fleet recovery, exception handling, and data ownership are new responsibilities, and a plan that cuts headcount without naming them is incomplete.
AMRs or conveyors for a distribution centre?
Conveyors suit stable, high-volume, fixed routes. AMRs suit changing layouts and mixed flows. Compare the route stability and expected layout changes over the system's life before choosing.