Main content

Ocado CEO - here comes the next generation of robots in retail

Stuart Lauchlan Profile picture for user slauchlan July 11, 2018
Ocado’s tech story keeps getting stronger, even if it comes at a cost to the bottom line.

Ocado Andover CFC - robot warehouse 370px
Yesterday Dennis took another look at the plight of Marks & Spencer (M&S) as yet more dire warnings emerged from the retailer’s AGM. One of the (many) issues that M&S has is that, despite having a thriving grocery business, it hasn’t a clue what to do about providing an online grocery service.

It’s been suggested before and it should be suggested again - look at a partnership with Ocado perhaps? The firm now has relationships with Morrison’s in the UK, Bon Preu and Groupe Casino in France, Sobeys in Canada, ICA in Sweden, and - newly announced - Kroger in the U.S.

That’s still nowhere Ocado has looked to be at this point, but CEO Tim Steiner is confident that there’s now a momentum behind third parties signing up for Ocado’s proprietary tech - and that that momentum is only going to accelerate.

That’s good news, if true, but it does come at a cost. The firm’s just announced half year group revenues of £799.9 millions, up 12.1% year-on-year, with £63.3 million of that coming from tech partnership deals. But ongoing investment in the tech that powers those deals took Ocado to a £9 million loss against a profit of £7.7 million this time last year.

But it’s case of invest to innovate, says Steiner, who also points to non-financial implications arising if his faster growth prediction follows through:

As a group now, we need to go faster. What does going faster mean in practice? We need to write code faster. We need to bring new capacity online faster. We need to build Ocado Solutions capabilities faster. We need to improve our fulfillment solutions faster. And we need to innovate for the future faster. The reason that we need to go faster is that we are going to roll out more facilities faster than we have envisaged. We are going to have tens of thousands of our robots live faster than we had envisaged. And so, we need to improve at an increasing pace.

On the writing code angle, he reveals:

Interestingly, we had one of the Big Four come in and do a review of our IT, our technology and engineering area recently. They wrote Ocado as one of the best in Europe for agile code writing. We write good code. We have been writing good code for quite a period of time now but we want to go faster. Even Greg Clark, the Business Secretary, is mentioning us alongside people like Amazon and Rolls Royce in terms of our talent in AI.


How much a testimonial from Clark means is open to debate, but what is clear is that Ocado has been on an upward growth path. The Morrisons deal was signed in 2013 at which time there were 300 people working in Ocado Technology. The firm is now looking to add the same number of staffers in the second half of this year, topping up its current headcount of 1,000.

There are some significant executive appointments and reshuffles underway. Chief Technology Officer Paul Clarke is to become Group CTO with a remit to set a tech path reaching out as far as 20 years hence. James Matthews, a veteran of Ocado Technology, is now managing engineering as CEO of the division. And there’s a new role of Chief Product Officer in the form of Simon Thompson, formerly one of the main drivers behind Morrisons online efforts and then Chief Digital Officer at HSBC.

A stat that Steiner is fond of is to point out that Ocado boasts operations that are 4 times the size of Tesco’s largest automated facility, but set to deliver 16 times the volume. Bringing new capacity into play is vital, he argues, and this is where Ocado’s robotics play comes into focus. It’s reckoned that the retailer. has over 600 robots on its grid - and they’re about to get an upgrade with a second generation of bots starting to come online. Steiner explains:

These bots are intended to run to the same physics model as the previous generation, but they’re intended to bring down the long-term cost of ownership, the combination of capital and running cost.

Now, when we’re developing bots at the moment, think about how a company like Apple might develop an iPhone. You have concurrent teams developing new generations because the development cycle is longer than the period of time in which you want to roll out a new product. So, while we’ve been developing the second generation robot that we are now taking delivery of, there is the majority of the fleet in Erith going forward and we are also developing the third generation robot. And we’re currently testing subsystems of that third generation, which is a robot that we will roll out to the majority of our international customers

To give you an idea of kind of scale of change, Generation 2 has about 25% new components compared to Generation 1. Already it’s proving hugely valuable to us in terms of some benefits in its engineering ownership cost, so we can change components on it 80% faster than on its predecessor. Generation 3, the next robot, is built on 75% new components versus Generation 2, so a significant change.

The scale of change means ramping up investment in testing, he adds:

A year ago, we had one test grid in our test facility in Hatfield that we ran 10 hours a day five days a week. During the half, we’ve moved into a new facility next door that’s over three times the size. We’ve now got six test grids on that. We are running them 24/7, giving us a significant increase in our capabilities to design, engineer, and prove the new hardware. Today, before we make a software change to the fleet running in Andover, that software will have been through over a 1,000 bot hours of testing in our test facilities, something that we weren’t capable of doing beforehand.

Decreased capital costs is one of the main objectives in all this and Steiner points to success to date in this respect:

The engineering cost on our fleet in Andover is reducing dramatically. In the half, our engineering cost per order reduced 66% from the start of the half to the end of the half. So, the trends are all very positive. We are spending money developing robotic picking and we have our first robotic picking cell in Andover, currently not being used as part of the live production, but there to show our clients how it works. We are exploring opportunities in immediacy, so we can address more of the spending in this market and in the global market.

AI of course

Then there is of course the inevitable AI story to tell. Steiner makes another bold prediction - spend on AI and machine learning will outstrip that for warehousing and robotics. He explains:

We are using machine learning and AI in some areas of the live business today. So, the type of examples, the places that we would be using it would be, we’ve been improving our forecasting, which is seeing relatively notable improvements using some of those new technologies over what we were doing historically. We’ve been using it, as we’ve mentioned before, in the call center. That’s been reducing the amount of email volumes by characterizing emails for us. We’re starting to train it for responses to live chat, to emails and actually to voice as well, using the skills that we developed in AI for language cognition that we’ve been using in things like Alexa. So, those same skill sets will actually translate into future call center technologies.

We’ve been using machine learning to analyze some of the behavior of the bots to identify if there’s some condensation on oil spillage or track deviation. So, we use machine learning to analyze things like the torque on wheels to see if there’s anything out there that could advise us that there’s an issue on our grids, for example, because we need to build scalable solutions. We’re using machine learning to do diagnostics on the robots. The robot tells us that it’s not feeling healthy and it wants to come in. So rather than it breaks down on the grid, we want it to come in and tell us it’s not healthy and actually tell us what’s not right with it, by diagnosing its own behavior and telling us what that probably means has gone wrong with it.

It’s technologically-impressive stuff and there does seem to be more attention drawn to it by the essential third party partners to justify this effort. What else lies down the track is less certain, concludes Steiner:

What our software does today, what our software can do tomorrow, are obviously two slightly different things. As we develop all the software in-house, our ability to modify, enhance and change is very significant. We do expect in the future to allow ourselves or our customers to plug into a variety of different delivery options, whether those are the 3.5 ton vans that we use here in Europe or whether they are motorbikes being driven by our staff, someone else’s staff or crowd sourced, or bicycles, or a little robot or whatever else it might be. Our system is engineered to talk to all of those possibilities.

My take

Ocado remains a fascinating tech story. If Steiner is correct in his expectations, it could become a more interesting online grocery story. Come on M&S - pick up the phone!


A grey colored placeholder image