The vote nobody puts on the stakeholder map
Case study · Warehouse robotics · Packiyo, Rabot, Rufus Labs · 2023 to 2025
The wearable scanner shaved two seconds off every pick. It also dug into a forearm by hour six of a ten-hour shift.
Nobody in the room where we approved it had worn one for ten hours.
Packiyo is the system of record for a warehouse. Selling it to a growing third-party logistics operator means promising that the rest of the building will work with it, and the rest of the building is hardware: computer vision watching the pack bench and flagging a wrong item before the box is taped shut, scanners that move the pick interface off a cart-mounted terminal and onto the worker’s arm.
I ran those partnerships as Product Director. Integration requirements and deployment specs, across three engineering teams who had never shipped anything together.
The build went to plan. The approvals did not, over and over, and it took me most of a year to understand that this was not bad luck. A warehouse robotics deployment needs four independent yes votes. Each of the four is answering a different question, and none of them is the question you prepared for.
Here they are in the order I learned to be afraid of them.
The CFO, who you expect
Every automation vendor quotes productivity gains against an idealized baseline. The CFO’s baseline is not idealized. It is a building where two people call out sick the week before Thanksgiving.
Nothing here is won with a better slide. It is won with a model built from the customer’s own operational data, the bad case already in it, walked through by me rather than by the vendor whose product is under evaluation. Something like half the credibility comes from volunteering the downside before anyone thinks to ask for it. The other half comes from the number being small enough to believe.
The technical lead, who asks a different question than you rehearsed
Not does it work. How does it fail.
What happens when the vision system’s confidence score drops below threshold. When a wearable loses connection mid-pick. When the WMS and the scanner disagree about what is in a tote.
Every one of those has an answer. The answers have to already exist when the question is asked out loud, because a technical lead who watches you improvise has decided something about you that they will never say in the meeting, and the deal cools three weeks later for reasons nobody puts in writing.
The operations manager, whose entire year is eight weeks
Peak.
A change that adds throughput in March is worth nothing to this person if it adds any uncertainty in November. Their instinct is that new systems fail exactly when you need them most, and after enough deployments I stopped arguing with it, because it is true.
What worked was giving up on persuasion entirely. Put the cutover well clear of peak. Hand them a rollback they can execute themselves, at 2am, without calling us. The rollback is almost never used. Handing it over is the point.
The picker, who is not in the room
No budget line, no seat at the table, and more dead deployments than the other three combined.
A device that is uncomfortable at hour six does not generate a complaint. It generates a workaround. It gets left on the charger. It develops an intermittent fault that nobody can reproduce. Six weeks later the utilization data is bad, and the story that reaches the CFO is that the technology underperformed.
Adoption on the floor is not a change-management workstream that runs alongside the deployment. It is the deployment.
What changed in how I worked
The floor-level vote went onto the stakeholder map first instead of last, which sounds like a small reordering and functioned as a different job. It meant the pilot got scoped around a shift, not a workflow diagram. It meant asking a vendor for wear-time data they did not have and watching how they responded to not having it.
Two phrases from Packiyo’s sales leadership stuck, and both are about the same failure. “Time kills deals.” Every week a promised capability slips, confidence compounds downward, and the loss is never attributed to the slip. And “no science projects” -- adaptability sounds like a feature right up until it becomes an invitation to build something bespoke and impossible to sell twice.
The integrations shipped. Roughly 15% improvement in client operational productivity where the systems were actually adopted, and that qualifier is carrying real weight. What came out of it that I still use: a way to scope a WMS configuration against one specific customer’s workflow rather than against a feature matrix, and a business-case format a CFO can evaluate without translating it first.
The scanner is a good product. So is the vision system. Both are still running in the buildings where somebody thought to ask the fourth person.
Trevor MacDermid was Product Director at Packiyo from 2023 to 2025, owning technical partnerships with robotics and IoT providers and leading solution design for warehouse and 3PL deployments.