Peter MacLeod spoke to Ocado’s Andy Ingram-Tedd to listen to how cutting-edge dwell digital twins take away the guesswork from warehouse operations.
Ocado Clever Automation (OIA) has by no means been shy about scale. However in my dialog with Andy Ingram-Tedd, VP of Superior Expertise, he makes the purpose that scale shouldn’t be the story. The story is what you do with it. After practically 25 years at Ocado, he has watched the corporate develop from a decent early workforce to a world organisation with hundreds of individuals, and he’s nonetheless struck by the identical inner power that drove the primary deliveries.
“It simply by no means slows down,” he says. “There’s all the time one thing occurring, there’s all the time some new journey, there’s all the time some new mission.”
That tempo issues as a result of it shapes how OIA, the Ocado Group division that takes its know-how to prospects worldwide, thinks about automation deployment. Ingram-Tedd is candid a few acquainted false impression: that robotics is solely the substitution of individuals with machines. His view is {that a} extra correct approach to see it’s as methods design, and the interaction between people, software program and {hardware}.
“Lots of people all the time ask me, you’re creating robots, you’re placing folks out of enterprise,” he says. “However we’ve bought extra folks that we make use of right this moment than we ever had. We’re doing extra, and we’re turning into extra environment friendly.”
Simulation, he provides, is the self-discipline that forces you to deal with that interplay significantly.
Simulation vs. Digital Twins
If there may be one thread Ingram-Tedd needs readers to remove, it’s the distinction between simulation and digital twins, and why the 2 are sometimes muddled. Simulation, in his definition, is a predictive mannequin used earlier than one thing exists within the bodily world. A digital twin solely turns into a digital twin as soon as the warehouse is constructed and working, as a result of it’s constantly aligned to actuality utilizing precise operational knowledge.
Simulation is what you attain for when spreadsheets fail. Fundamental processes could be approximated with time and movement assumptions. However when you search excessive throughput and excessive utilisation throughout many shifting components, you want discrete occasion simulation, modelling numerous actions with begin factors, finish factors, course of occasions and guidelines.
“We actually do imply a discrete occasion simulation,” he says. “There are many issues occurring. They’ve a begin level, they’ve an finish level. You possibly can’t simply calculate that on a spreadsheet.”
Ocado’s personal definitions are simple. Simulation is used earlier than a system is constructed. You load assumptions, together with orders, inventory, structure, speeds and guidelines, you then run what-if eventualities to see outcomes and dangers. The questions are sensible: will this design work, what dimension ought to it’s, the place are the weak factors. A digital twin, in contrast, is a digital illustration of an actual bodily system that stays aligned to the dwell operation utilizing operational knowledge. Its worth is choice assist throughout operation, together with testing modifications safely and understanding what occurs in case you change one thing right this moment or tomorrow.
Eradicating Guess-Work
Ingram-Tedd emphasises that simulation shouldn’t be about your finest day. It must be about your worst day. Which may imply modelling downtime, late inbound autos, or labour gaps, both individually or together. “We’re operators of our personal gear,” he tells me. “We’re not guessing. We all know what the dangerous issues can occur. They’ve occurred to us within the final 25 years!”
As soon as a website is dwell, the inputs are not assumptions. They’re measurements. You possibly can take knowledge from the true warehouse, feed it into the mannequin, and check configuration modifications, from merchandise placement methods to outload timing, decide speeds and useful resource utilisation. The objective is steady enchancment, pushed by proof quite than intuition.
I ask why does OIA construct its personal simulation instruments. Ingram-Tedd argues that third-party packages are helpful, however inadequate for modelling the complexity of Ocado’s grid-based system, the place software program determines the place and when to retailer, retrieve and sequence inventory, whereas bots navigate above dense storage. “We don’t use a 3rd social gathering and there’s a very vital motive for that,” he says. “There isn’t an off-the-shelf simulation bundle that may try this.”
Ocado has developed its personal simulation functionality in-house since 2008. A key level, in his telling, is that the software program powering simulation is similar to the software program that powers the manufacturing website. That tighter hyperlink between mannequin and actuality, he says, helps higher design selections and extra confidence earlier than capital is dedicated.
Simply as importantly, simulation is finish to finish. It doesn’t cease at bot motion. It extends to conveyors, pallets, autos, folks and robotic decide, as a result of optimisation solely is smart on the degree of the entire ecosystem.
“True optimisation solely occurs while you put all of the subsystems and also you mannequin all of them collectively,” he says. “Integration brings complexity, and simulation helps you perceive the knock-on impact of each design alternative.”
Infrastructural Optimisation
The sensible worth is that simulation turns design questions into testable eventualities. One slide instance is the connection between bot numbers and achievable throughput. Run a spread of instances in parallel and you may plot the place diminishing returns start, figuring out a candy spot past which further funding yields little profit.
That very same strategy applies to select stations. OIA’s stations are modular, and simulation can discover how structure modifications have an effect on each throughput and the way in which an operator performs. The goal is to keep away from paying for human time whereas permitting the station to underfeed the operator with work.
In a single demonstration clip mentioned within the interview, Ingram-Tedd references a choosing efficiency determine that he is aware of will sound implausible to many readers: 1,072 models per hour on a station. He’s fast to caveat that it’s not a sustained working promise. Constructing a system round peak human efficiency dangers waste if folks can’t keep it, and drives pointless funding in upstream sources. A extra smart working goal could be 600 to 700 models per hour, he suggests, nonetheless effectively past frequent trade expectations.
What usually breaks automation shouldn’t be the typical case, however the exception: odd merchandise, awkward displays, or uncommon failure modes that also happen steadily at excessive quantity. In robotics and automation engineering these are often known as nook instances, uncommon or excessive conditions outdoors regular working circumstances that should nonetheless be dealt with safely and reliably. “You possibly can’t have robots like this in a dwell website until they will do nook instances,” believes Ingram-Tedd.
Future Seems to be Shiny
Past grocery, OIA is making use of its platform to different verticals. Ingram-Tedd highlights a significant undertaking with McKesson in Canada – not but dwell on the time of the interview, however not far off – which he describes as a big system in Montreal designed to boost productiveness whereas enhancing traceability, accountability and safety. He argues that pharma distribution shares many traits with grocery, however with stricter compliance necessities, significantly round batch and lot traceability. He hints at vital productiveness good points, whereas noting there are customer-specific variations that stay confidential.
He additionally dropped at my consideration that mutual exclusivity has ended within the majority of markets the place Ocado operates its grocery know-how with companions, opening the door for Ocado to carry its advanced product again to a few of the world’s most developed e-commerce markets after a interval of exclusivity agreements.
In direction of the tip of the interview, Ingram-Tedd briefly referenced a brand new choosing know-how deliberate for introduction in 2026, which he characterises as a major step-change. Logistics Enterprise was given an early have a look at the idea, however particulars stay beneath wraps forward of public launch at MODEX within the Spring. We hope to return to this in a future version, as soon as OIA is prepared to discuss it in full.
For now, his message is constant. Whether or not the query is what number of bots to deploy, how one can design a decide station, or how one can combine the subsequent wave of automation, the differentiator shouldn’t be a single robotic. It’s the functionality to mannequin complicated methods precisely, be taught from actual operations, and hold enhancing.
