Plan for the day the automation fails in the warehouse

Robots excel at repetition and struggle with the unexpected, from a cracked tote to an order that won't fit. Thomax founder Thomas Jackson explains where automation projects come unstuck, and what businesses should ask before they bring in the robots.

Every automation project in supply chain is built on a belief that the robots will work. Throughput targets, business cases and go-live dates all depend on it, which is why businesses need to better plan for the day the automation systems stop.

Thomas Jackson, founder and managing director of warehouse management software provider Thomax, has seen what happens when something falls outside what a system has been set up to handle. "There's a lot of excitement about the automation path, and the robot that will do a thousand units in an hour. What happens when it doesn't? What happens when you're doing maintenance, or when the unexpected occurs?"

These can be the "we didn't think about that" moments, and as automation becomes more accessible, more businesses are running into them. Thomas says smaller, modular systems and robots-as-a-service models have allowed businesses automate one area at a time and spread the cost, without tens of millions upfront - creating more opportunity, along with more room for error. 

Even robots have bad days

Warehouse automation thrives on repetition. Thomas points to an operation where the entire range fits four box sizes, all the same length and width with four different heights. "That is the entire variation in their business. Automated systems are great at that because every box is predictable," he explains. "The robot knows the footprint and there are only four heights to deal with, so it's doing the same task over and over."

"Robots are really good at something that's repeatable, but they're not very good at things that fall outside it," says Thomas. Damaged products are a common example: a plastic tote with a crack in it may stop an arm docking properly. When a section of conveyor goes down, the challenge becomes keeping the operation moving.

Humans handle these moments without a second thought. "If I'm a person pushing a trolley around a warehouse and that aisle's blocked, I'll just go down a different route," says Thomas. "A robot doesn't have that ability to think or respond if it's not trained for that specific situation." 

Common problems

Some of the most reoccurring challenges start with the physical product. In a mixed range, some items will never collectively fit the automation. Thomas describes an appliances and homewares warehouse where items like toasters, kettles and laptops all live in a robotic system. "You're not going to fit a 60-inch TV in a robotics tote, but it's quite normal for someone to order a PlayStation and a TV to go with it. How do you split that, recombine it and ship it together?"

Data creates a similar blind spot. "A human can look at a box and see it's a medium-sized box, and make a decision based on what's in front of them," Thomas explains. "A robot needs to know the exact specifications before it can do anything." Robotic systems thrive on data, and Thomas finds many businesses are wanting to run before they can walk - in this example, deploying automation before the clean, accurate data is in place to program it. "The system can do all these amazing things, but only if it has access to high-quality data."

Many legacy Warehouse Management Systems process work in batches every 15 minutes or every hour, a rhythm built for printing a ticket and handing it to someone to walk around. Deploying automation requires a change in this pace. "With some of the robots we work with, you have milliseconds to respond. If the instruction doesn't arrive in time, the robot sits there with nothing to do," says Thomas, pointing out that this then doesn't work for batch-based processing. The system also must hold that real-time response at peak, when volumes jump.

What to watch out for

Cloud-based systems are built for real-time events, but they bring their own dependency. "You're now reliant on your internet connection working and being low latency. If that internet connection goes down, so does the rest of your site," warns Thomas, acknowledging that a backup connection in this case is the usual answer.

Testing can be difficult to get truly accurate. Emulators and digital twins help, but Thomas says they are often 'too perfect'. "The barcode scanner will always fire, the conveyor will never jam, and it will always hand over seamlessly to the next step," he explains. Proper testing needs the physical system, which means that challenges can rear their head after go-live. 

Many buyers are also new to automation. Thomas is working with one business that is moving from a short run of conveyor to a multi-million-dollar tote-to-person system. "It's a big step for a business that hasn't automated before. We can bring the lessons from other sites and help them plan for exceptions, but they know their own operation best, and getting it right depends on combining the two."

Questions to ask before you sign

For businesses weighing up a WMS provider, Thomas's first question is whether they have done this before. "You don't know until you don't know," he says, also noting that it's important to distinguish what is needed with configuration and code for the integration. "There will always be an element of configuration, because every robotic system is different. But how much of that is going to be configuration versus writing code and building something from the beginning?" 

Thomas delivers a reminder that buyers should also ask whether the system holds real-time performance at peak, how it handles orders that span automated and manual areas, and how the operation keeps running when the automation is offline. "Whether the robots are offline for maintenance or something's gone wrong, you need to have a plan for how you'll keep trading."  

"Successful automation projects aren't defined by how well the robots perform on a good day," Thomas adds. "They're defined by how well the operation continues when something unexpected happens. The businesses that plan for those exceptions from day one are most likely to realise the value automation promises."