Maximising Efficiency

Operational efficiency

We aim for efficiency in every process throughout our operations, from our customer-facing interfaces in the webshop, mobile and tablet applications, through to the routing software that supports our deliveries. That means adhering to our key design principles of automation, use of our own technology, and centralisation – the three foundations behind our industry-leading CFCs.

Removing several of the typical human processes from grocery retail and distribution helps us operate with product waste of 0.8% of retail revenueA – which we believe is the lowest in the industry – and our digital stock management systems enable perfect first in, first out stock rotation.



Gross Sales (Retail)

FY17: £1,429m


EBITDA (Retail)A

FY17: £79.2m


Delivery punctuality

FY17: 95.0%


CFC efficiency (UPH) (Mature)

FY17: 164


Product waste

FY17: 0.7%


Average deliveries per van

FY17: 182

Numbers on a 52 week basis


  • Delays in implementing new capacity for Ocado.
  • Failure to achieve projected improvements to the fulfilment solution, impacting ability to lower long term cost of ownership as expected


Our fulfilment solution is critical, and we are planning to roll more of it out faster than even we anticipated. Developing a new generation of bots helps us bring down long-term cost of ownership, the combination of capital and running costs. In the first half of the year, we started to take delivery of our second-generation bot. This has about 25% new components compared to our first generation.

While we've been developing this second-generation bot, which forms the majority of the fleet in Erith, we are also developing the third-generation bot, and we're currently testing this. We have concurrent teams developing new generations, because of the length of the development cycle. This is the bot we'll use for the majority of our international customers. The third-generation bot is built on 75% new components compared to the second generation, so is a significant development.

We've dramatically increased our testing capabilities to build this. In 2017 we had one test grid in Hatfield that ran ten hours a day, five days a week. We've now moved into a new facility three times larger, and have six test grids running 24/7, giving us a significant increase in our capability to design and engineer and prove new hardware.

Today, before we make a software change to the fleet running in Andover, that software will have been through over 1,000 hours of testing, which we weren't previously capable of doing. All in all, improvements mean the engineering cost on the Andover fleet is reducing dramatically, with costs per order reduced 66% during the year.

We measure efficiency within our CFCs by average units processed per labour hour (UPH). In our mature CFCs, this year the UPH figure was 164, stable on the prior year.

As we increase customer densitiy in existing geographies this brings efficiencies as the drive between our customer drops gets shorter. We continue to find opportunities to improve our drops per van (DPV) measure of delivery efficiency through improvements in our complex algorithms which calculate delivery routes and cover variables such as timing of delivery slot, house location and order volumes. Increased demand in our existing catchment areas increases the density of customer orders, so we are continuously improving the efficiency of routing.

Future focus

We are constantly looking for ways to use pioneering technology to improve our operational efficiency. We have multiple strands of both business-sponsored and more speculative research under way.

One example is the robotic picking and packing of customer orders. In grocery, this challenge is made more difficult by varying product characteristics such as the size, rigidity, and fragility of each of our 50,000+ products, coupled with the added complexity of packing into bags. Solving this challenge requires a combination of grippers that can mimic the capabilities of the human hand, advanced vision systems that recognise how to pick up different products, and advanced machine learning that can make smart decisions on the fly. Our robotic picking cell in Andover has operated for a year now as a "living lab", and will go into production in 2019.

One of the most exciting applications of pioneering technology is attaching a machine learning model to highly realistic simulations and visualisations of our business. Similar to DeepMind's simulation of the game Go, this approach will enable us to discover new optimisation strategies and solutions that might have otherwise remained hidden.

Capital efficiency

The core design principles behind our efficiency are automation, use of proprietary technology, and aggregation of scale through the use of our large CFCs. Combining these attributes, we believe we have developed the most sophisticated and operationally efficient grocery shopping and delivery solution in the world.

As we develop new CFCs we have been able to improve the capital efficiency of our operations. We have demonstrated this through continuous improvement within our mature CFCs, where we have extracted additional capacity without significant investment. The proprietary technology we use in our newer CFCs enables us to achieve an attractive return on investment, estimated at over 50% for our Erith CFC, even before any further efficiency benefits from other innovations such as robotic picking.

Our proprietary materials handling equipment (MHE) solution in Erith and Andover is modular, allowing for reduced upfront capital commitment as it can be added to as volumes grow. This also means we can build in smaller spaces if required, making our capabilities more customisable to the varying capacity requirements of our Solutions partners.


With our need to bring new capacity online, we have been delighted with our successful ramp-up at Andover, more than doubling the volume over the year. It is now running at 30,000 orders a week, almost 50% of its projected volume. We have over 600 robots on the grid, over half of the projected total.

We also opened our fourth CFC, in Erith. This is the same solution but with some enhanced components. For example, a redesigned grid that's more structurally durable, and with improved peripherals. More importantly, in its first three weeks it reached the same volume that it took us 32 weeks to achieve in Andover. We are designing our platform so that every lesson we learn at Andover, every enhancement to automation and software can be quickly transferable to every future site.

Erith is four times the size of our nearest UK competitor's largest automated facility, yet we expect to achieve 16 times the volume. Our investment will achieve lower long-term cost of ownership.

Thanks to design enhancements to our bots, we have been able to increase the weekly production of bots by 100%.

Our first two CFCs in Hatfield and Dordon continue to operate at high levels of accuracy and with improved efficiency. With engineers constantly on site to ensure what are they are well maintained, we incur minimal maintenance capex. Cash flow return from these CFCs is nearly 10% of revenue a year, illustrating the significant cash we can generate once a CFC has scaled and reached maturity.

Future focus

In 2018, we successfully applied our learnings from Andover to significantly improve our ability to deploy and ramp up our latest generation solution in Erith. As we continue on this trajectory, we expect to see further improvements in the speed of deployment allowing for reduced upfront capital commitment and shorter ROI timescales, automation enhancements that further increase throughput, and the gradual introduction of robotic picking for additional efficiency gains. As we continue to progress our own CFCs and garner new learnings along the way, we naturally pass these benefits to our Solutions partners.


Constantly enhance end-to-end technology systems

Continuously develop more capital and operationally efficient infrastructure solutions

Case Study

Bot HealthcareImproves Efficiency

In 2018, our technology teams have achieved further significant breakthroughs in applying advanced AI, machine learning, and simulation techniques to make our systems smarter and more autonomous than ever before. One area where we have seen significant progress is in our ability to predict deviations in the behaviour of bots.

With tens of thousands of bots that will run on the Ocado Smart Platform in the foreseeable future, the real time monitoring of these swarms of bots will soon be beyond human scale. To address this, all the operational and sensor data from the bots are streamed to the cloud where a machine learning based healthcare system performs powerful predictive analytics and drives preventative maintenance.

Due to the fundamental nature of our swarm of identical bots, any bot requiring maintenance can instantly and seamlessly be replaced by another one, with no loss of throughput. In the future, we will further enhance this bot healthcare by embedding machine learning processors into each bot, enabling it to perform smart self test, diagnostics and exception handling, in order to drive even greater efficiencies of scale.