Physical automation: the gap between pilot and production
Why robotic pilots look convincing but scaling hits a wall - and how to think about this before making an investment decision.
By 2025, the industrial robotics and warehouse automation market has grown to a point where a physical automation pilot has become accessible for mid-size businesses. Prices for cobots, mobile robots, and machine vision systems have come down. Providers are ready to run demonstrations and fast pilots.
The problem I see regularly: the pilot works well, but the transition to full production operation turns out to be more expensive, more complex, and longer than expected. This is not accidental and is not the fault of a specific provider - it is a systemic gap that is important to understand before making an investment decision.
Why a pilot almost always looks better
A pilot runs in controlled conditions. Operators are motivated to show their best result. The integrator gives maximum attention to a single installation. Edge cases are handled manually or excluded from the demonstration.
In full production the picture is different:
- equipment runs around the clock and wear accumulates;
- the product changes - new SKUs, new packaging, non-standard items;
- the operators on shift are not the ones who participated in the pilot;
- technical problems occur during shifts without technical support on site;
- integration with ERP, WMS, or other systems creates dependencies that did not exist in the pilot.
Each of these factors exists separately. Together they explain why a system's actual throughput in the first year is often 20-40% below its pilot performance.
Where the gaps specifically appear
Object variety. Robotic grasping works well with uniform, predictable items. When the product range expands, packaging is non-standard, or items are deformed, the error rate rises. This is what pilots typically do not test across the full range of variations.
Maintenance. Who services the equipment? Are specialised technicians required? What is the support contract, and what happens if the system stops during working hours? In the pilot, the integrator is always nearby. In production operation, they are not.
Retraining when things change. When the product or process changes - how quickly does the system adapt? Who does this, and how much does it cost? This question is rarely discussed before the contract is signed.
Integration with operational systems. An autonomous robot without integration into inventory management or a warehouse management system creates parallel record-keeping. That is a new source of errors and manual work.
What to check before the investment decision
A few questions I recommend asking the provider and yourself before signing a contract:
- Is it possible to visit a live installation at another client - not a specially prepared one, but in normal operational life?
- What is the documented performance at 12 months after launch, not at the point of handover?
- Who services the system if it stops at night or on a weekend?
- What happens when a new SKU is introduced or packaging changes - how quickly does the system adapt, and who does it?
- What is the total cost of ownership over three years, including maintenance, updates, and adaptations?
If there are no clear answers to these questions, that is a signal. Not necessarily to abandon the project - but definitely to get the answers before, not after.
Physical automation works. But the gap between pilot and production operation is real, and it is measured in money. The right time to think about it is before signing, not after the first year of operation.