Ocado Full Year Results Presentation. There's a lot to celebrate in these results, but also what you're going to hear is there's a lot of work that we still need to do. And it's only proper we should start with the successes. First, you know, we've, we've seen a 25% increase in the live modules, average live modules, during the course of the year. The technology works, and it works well. It's delivering at or exceeding targets wherever we, we, we apply it. We've added three more live CFCs in U.K., Canada, Japan, so we've extended the geographic spread, and they're, they're ramping up to target. Customers love our online grocery experience that our clients are delivering in their local markets, and the NPS scores are consistently very high. You look at the Re:Imagined technologies, we're rolling those out.
This is the first year they've been commercially available, and they're already being deployed in the U.K. and in Sweden. We've expanded the partner success activity, some theme you'll see coming through quite extensively during this presentation, you know, helping our clients to grow, to achieve long-term profitability. I mean, we really do know online grocery, and we are out there helping these clients to get the most out of the Ocado Smart Platform and to find out better ways of filling their capacity and building new capabilities. Of course, it's the first year that we have secured a contract outside of grocery, with the Ocado Intelligent Automation, of course, the first contract being McKesson in Canada, which is, they're the leading pharmaceutical distributor in the U.S. Ocado Retail is now on a good growth trajectory.
It's adjusted to the post-COVID world and delivering on the Perfect Execution Program that was introduced by Hannah Gibson, the CEO. And as you'll hear, we have made considerable strides on cash flow, and we can reiterate our target of becoming cash flow positive in the midterm. But there's work to do. You know, repeat orders of new modules are just not at the pace that they need to be. You know, our partners' success will determine our success. And so it's encouraging to see the extent to which our teams are working very closely and very collaboratively with our client teams, really to think how to deliver on the potential, how to deliver the growth of the platform and the investment they've made in us.
There's been a lot of progress during this year, but there's a lot that we still have to do and we must do, and you'll hear about that during the course of the presentation. This year, coming year, we have to focus still on the partner success. We have to continue to roll out Re:Imagined technologies. We have to get more efficiency, longevity out of our other technologies, and we need to go out and find more opportunities for the Ocado Intelligent Automation. But from what I know of this company so far, I am totally convinced that the intelligence, the resilience, the creativity of the Ocado people will deliver on the huge potential in terms of value add and the technology as a whole.
When you hear this, I hope you come away thinking, as we do, that we look forward to the future with, with great confidence together. So over to you, Tim.
Thanks, Rick. Okay. Where are we? Okay. So the key messages for today are that we do continue to make financial, operational, and strategic progress. So we've had a strong year in that sense. The success, as Rick said, for our partners is our key focus right now. We want our partners to fill up their buildings, to draw down more modules, and to buy more buildings. And critically, they need to be successful with the capacity they've already downloaded. We continue to innovate, and our past innovations, our current innovations, and the pipeline of innovations are driving and will drive more opportunities and attractive returns for us and for our partners. We have got good financial discipline across the business, and Stephen obviously will talk to the numbers, but as Rick said, the cash burn for the year was down substantially ahead of what we had targeted.
We have an attractive business model operating in multiple segments, and I always think it's just so much easier to understand our business when you focus on each of those segments individually, so that's what I'm going to do today. In terms of the financial progress, we've seen strong revenue growth across all segments, positive adjusted EBITDA on improving cash flow across all businesses, and Ocado Retail returning to profitability, in line with guidance. We made operational progress. We've seen a significant growth in live modules. We have seen increasing productivity in terms of Ocado's productivity on delivering those live modules every day for our clients. We've seen cost efficiencies, and we've started the deployment. In particular, OGRP is operating at quite a pace and a growing pace in Luton. It's quite something to see if you get the opportunity.
We have rolled out AFL a little bit here in the U.K., but we're doing 100% of the loading in our client's facility in Stockholm with AFL, and we expect to see more rollout of both of those products during the course of 2024. Strategic progress, massive focus on partner success, and obviously, John's going to talk more about that, and I don't want to steal his thunder. We did sign our first Ocado Intelligent Automation deal with McKesson. We settled the AutoStore litigation, so a topic that we got to talk about for a couple of years quite extensively was not mentioned on my media call this morning for the first time. We announced this morning the 13th OSP partner signing with Panda in Saudi Arabia.
We also opened this year three new sites, and we now have over 1,000 stores live on our in-store fulfillment platform. So looking at the financial progress in summary and some numbers, I'm just going to head straight to the change column, right? Group revenue was up 10%. Adjusted EBITDA improved by GBP 126 million from a loss of GBP 74 million to a profit of GBP 52 million. Our loss before tax improved by GBP 53 million, and the underlying cash flow improved by GBP 356 million. So that was all on the back of a strong trading performance and a very rigorous focus on costs. Stephen has it all very tightly locked away.
So these are the 13 world-leading retailers currently on our platform, and between them, they have more than GBP 250 billion of annual sales in grocery that we would look to see a portion of that moving online in the next few years. If you just took 20% of that online, you'd be talking about GBP 50 billion, and you'd need to roll out north of 600 modules to achieve that level of volume. So let's have a little bit of a look at the technology solutions in a bit more detail. If we move on here, hopefully you can all see that in better size than I can look in the other direction.
What you can see here is the 232 announced modules, but you can see the breakdown here in terms of the blue being the live modules currently generating revenue, up from 99 at year-end 2022, up to 111 at this year-end. The green, the 49 modules, are modules that can be drawn down in existing live sites, right? So that's where we've built a seven-module site in somewhere or other, and the client's using three or four. The remaining three or four are in that 49. They also are able to be drawn down with significantly reduced Capex because the grid's already installed. Usually, in these early versions, we installed more of the kit up front, so the machines that take the customer boxes out of the hive boxes, they're already installed. A number of the pick stations are installed. We probably need some more robots.
We have those robots ready to install in those sites. So that's what the 49 are. They are no new sites needed. We can grow by 49 modules. The 40 are modules that are in sites currently under construction, not yet being paid for. So they would include things like Sydney, Melbourne, Phoenix, Calgary, Busan, for example, right? So that's where you get the 40. The 32 are announced sites by our clients that are not yet actually a live program. They've announced them, but they haven't come to us and said, "We found the site." So they would be places like, Miami and New York and Los Angeles where our partner in the U.S. has said and signed for something but actually has paused, as you know, while we get the eight sites that are live working better.
Then we'll restart those programs, but that's what the 32 represents. It's easy to see that if we take the 160, the 200 of those modules, the 111, the 49, and the 40, no new site announcements, no new projects, no new finding sites, we can get to 200 modules like that. Obviously, what we want to do is get to 100s of modules, as I was speaking about before. We're confident that module growth will accelerate as the partner success delivers. We expect the Re:imagined and the AI-driven enhancements to actually allow us to go back into these sites and say, "They're not actually 232 modules. Without you needing to build more as a client, we're going to take those up to 250, 260 modules." What I mean there is clients said, "I want a six-module site." We designed with them a building.
They built the building. We put the kit in. They've currently got six modules in there. They might be using two, three, or four. We're coming then as they grow, we'll then sell them a fifth. We'll then sell them a sixth. Actually, we're now able to say to them, "You know that six-module site? We've worked out through our intelligence how to turn it into a seven -module site. So if you want to draw down another module and pay us for another one, we can turn that into a seven-module site." So our expectation is that there are more than 250, 260 modules in those sites that we're referring to there. Okay. So what does that look like in terms of moving forward to 2024, 2025, 2026? So what we try to show here is the dark gray are new committed openings.
So the dark gray is where we will get to by just opening Phoenix, opening in 2024, 2025, opening Charlotte, opening Sydney, opening Melbourne, etc. So the new site openings take us to around 120 and 130, right? The lighter shaded gray is how many of those other 60 modules get drawn down in that timeframe. And our view is that it is a kind of a J curve. More of them will get taken down in 2025 and will get taken down in 2024. And so the way I look at that at the moment is a handful of those sites are close to or bumping up towards their capacity that they've currently drawn down, and I would expect to see some drawdowns from some of those sites in 2024.
The key mission for partner success is not only to help accelerate the growth in those sites so they draw down some of those 60 modules, but is to get the other sites up to that level so that in 2025, they then need to draw down more modules. The sites in the U.K., with the exception of one, are operating at their full design module capacity. So as Ocado Retail grows, there are no drawdowns until we get to the point that they have filled up the design capacity, and they then ask us to sell them the excess capacity that we can generate in those sites. So hopefully, that makes some sense to sticking there for a moment, a few more parts there.
There is another accelerant that is going to hit us in the beginning of 2025, and I'm going to talk about it on the next slide, not quite yet, which is called Swift Router, which we talked about at the Re:imagined conference. And the combination of partner success, helping our partners to get better at every aspect of running their business where it's different to sustain a grocery business together with Swift Router, which will dramatically increase the addressable market from these warehouses, is what's going to drive that growth. The two last points I want to make is the drawdown of those remaining modules is much cheaper in CapEx. We spoke about that already.
The second one is that future sites are also at lower CapEx, but future sites also can be we have worked out how to build them now so that we can put a lot less CapEx in up front and allow our partners to start sites with less Modules. So most of our sites have started with between about 35% and60% of their max capacity at go-live. We could now, going forwards, allow our partners to start at lower numbers, and then you'll see more constant drawdown because as they're growing, they'll be drawing down. At the moment, a partner might have taken a seven-Module site. They might have started at four. They initially, they're using one. Then they're using two. Then they're using three. Until they're using four, we're not going to see them draw down five.
In the future, a partner might buy a six-module or 10-module site and start with one, and then they can draw down as they're going because we've changed the way we can phase the CapEx through Re:imagined and site designs into those warehouses to allow us to do that with no susceptibility on our side to their ramp, right? Zero. So it doesn't matter that we build a 10-module site and we go in at one. That one module is a profitable module for us, and then we can add them on as we go, right? Let's move on to the next one. So what is Swift Router? And we know we've spoken about it before, but let me just re-emphasize what we're doing here. Today, we do route completion, route finalization.
We work out which customers are going on which route we're sending out, and we complete that, and then we put that bunch of work into the warehouse. We do that because historically, warehouses work as a kind of downstream process. You start here. You work your way, and then eventually, the order pops out the other side and needs to leave. You can't have the orders finished and not ready to go on a van because they would build up and block the warehouse. But when we moved to OSP grids, we've broken that challenge. We can store completed orders in the OSP grid, but the software still required us to send a completed order of a completed route into the warehouse for processing.
Some of the substantial work that we're doing this year is breaking that link so that our warehouse can work on orders that have not even been completed, let alone assigned to a specific route, okay? Now, what that then allows us to do, and the core part of the functionality, is it allows us to be picking and be ready to load stuff. So a route that we want to leave at, let's say, 10:00 o'clock., we could now, at 9:46 AM, decide which orders are going to go on that route that's leaving at 10:00 o'clock. They could be orders that were completed yesterday, and they could be orders from customers who only came in our shop 15 minutes ago.
We can process them in 15 minutes, then decide to put them on the same van, load them onto a van at 10:00 o'clock, and be dropping them at a customer's house at 10:30 AM, okay? That's what we're going to be technically capable of doing. Now, that comes with a challenge. Some of our clients already have a new thing to manage. They have to manage routes. They have to manage driver shifts. They have to manage understanding whether they're offering availability or not offering availability live, and that's quite complex. So this year, we are also developing AI copilots, and we're doing that because this is going to get significantly more complex without it when you have continuously dispatching routes all through the day that can put any orders on, whether those orders were placed yesterday, three days ago, or half an hour before.
And you want to reserve the front part of those routes for the orders that you expect you're going to take in the hour or two hours before the route goes out to enable those, that capacity on each van, to be doing short lead-time deliveries, under-four-hour deliveries. You want to be able to use part of the rest of that route to be able to do the orders that you took earlier in the day for later the same day dispatch. And you also want to use it for orders that you took the day before. And you want to be doing clever things to do with pricing of those slots. Youi want to be doing clever things to do with which ones are going on which routes.
You want to be advising the clients what they need to be doing in terms of shift patterns for the next day, the next week, the next month, in terms of learning from the data as to what the customers want when they want these slots and all this type of stuff. There's a lot of data there, and we need to be very clever with it. We need to help our partners to run this thing really, really, really simply. So building the AI copilots to run Swift Router with continuous dispatch. Why are we doing all of this? We're doing all of this because we can go from facilities that are 90+% today to day for tomorrow. We can move those facilities to something that looks more like 1/3 under four hours, 1/3 same day, and 1/3 next day.
That's more like what the average facility can look like in 2025 versus what it looks like today. That's a very, very significant change. What's amazing is that we can create that level of earlier lead-time deliveries but without compromising on the operating hours of the warehouse, without compromising on the operational efficiencies of the warehouses or the operational efficiencies of the routes that we operate, okay? So the ability to do 22 drops on a route type of thing, right? And so with those economics, we can dramatically grow the market. We think this is a very important and significant growth enabler in all of our markets. But let's be aware, there are some markets in the world that emerge differently to others, and they emerged with more expensive services that had one significant benefit being that they delivered earlier.
Otherwise, they were just more expensive. And they trained customers to expect earlier deliveries. And therefore, in those markets, we expect this change to more than double the addressable market. And this change is coming live at the end of this year, beginning of, end of 2024, beginning of 2025. This will be live and available to all our customers across all our markets. Right. So overall, our platform continues to scale. So if we look at the history here, pre-OSP, the platform is scaling at 17%. We had just the warehouses here in the U.K. We turned 2 warehouses live in that five-year, six-year period, from 2012 to 2017 inclusive. And in the next six-year period, we've now ended up with 26 sites globally. So we opened 24 sites, I think, unless that includes the sites that were opened pre-2012, but something along those lines.
We have taken up the capacity of the live sites and the drawn modules. So the drawn modules in the live sites are now at a capacity of GBP 8 billion. And of the 200 modules of the sites that we showed before, not including the announced but not drawn ones, 232, would take us up to GBP 20 billion of capacity. We spoke before about if 20% of our current clients got their business online, we'd need GBP 50 billion. But you can see where we're trying to go to. And you can see in that slide, sorry, the split was the green and the blue was the split between the U.K. and international. Right. John, over to you.
Thank you very much. And good morning, everybody. Partner success, which I'm going to talk about this morning, what is it, and why are we doing it?
Partner success is just a team of very experienced and specialist resources in Ocado Solutions. It is 100% focused on helping our partners to grow quickly and grow profitably. We've appointed a really experienced team of business leaders to lead Ocado Solutions in our three regions. You can see those on the screen with the number of partners. Each of those, importantly, now has resources dedicated specifically to each of our partners so that we can work as closely as we can with those partners on a day-to-day basis. And you see along the bottom, we've got central resources. That is a bit of an understatement. Many of these people are very experienced managers in Ocado. Very, very often, they've got really great experience of Ocado Retail and Ocado Logistics going back years and some of them decades.
They've got really good understanding of the planning, the controls, the operations, the tools that have made Ocado Retail so successful. We've also included there in those central resources; we've recruited a number of deep subject matter experts in areas like e-commerce, marketing, customer acquisition and retention, retail media, from outside the company. Finally, we also have a very effective analytics function, where we can identify with our partners what the best opportunities are. I come on board in September. I have visited every country now where we have operations, and I should share a few observations with you. Firstly, all the CFCs that I have visited, the grids, the bots, the peripherals, our latest automation, it operates absolutely as it was designed to and as it's meant to. Our technology does work really well.
Tim mentioned we talked a lot historically about Re:imagined technologies. These are being rolled out now much more widely to improve efficiency, to improve safety, and to lower cost on partner sites. The most transformational technology, I think, is going to be OGRP. We expect that when it is rolled out, that's going to save several million per year at each facility. And it'll lower our partners' exposure to expensive labor costs and also to staff churn. ISF, we've got quite a number of partners using OSP in store or dark store environments, and that might be used to service less densely populated areas, or to create demand in advance of opening a CFC. And in some other areas, it might be used in lower labor cost countries, w hatever, our partners are using this to really great effect now.
Our ISF solution has been rolled out now to more than 1,000 stores. Ocado is very strongly aligned now around that clear objective, which is to help our partners accelerate profitable growth. Very importantly, from my point of view, if you look at the last bullet on this chart, our partners are fully engaged with us. We're sharing best practice data on productivity, on customer satisfaction, on finances, efficiency, operations, organizational design. And we're doing that very openly and very collaboratively and comprehensively with them. What are we focused on today? Firstly, customer acquisition. So revenue growth here, customer acquisition, retention, driving basket size. Our partners have made this investment with us with the specific goal of taking market share in e-commerce and providing an unbeatable customer experience in that online grocery retail space.
We've got a dedicated team within our partner success team, focused on supporting our partners to grow and also to maximize income from sources that might not be available offline, such as retail media. Secondly, the second box here, operational efficiency. To get the most out of our CFCs, it needs good planning. It needs disciplined execution and dynamic integration with marketing. Our most successful operators are also maximizing the proportion of direct inbound deliveries. And they're constantly looking at reducing waste. Last-mile operations account for a large part of the cost base. Careful management of slots and route optimization makes a really material difference to the P&L. And thirdly, product development, as Tim touched on here. I won't say too much about it today, but we are working with our partners. And this is a two-way process.
We need to understand what developments they want us to make to help them gain market share. And also, we need to make sure that we are taking them through the latest capabilities that our tech team has developed to make sure that they get the best use out of the platform. The significant investment that we make in the platform that I'm sure Stephen's going to touch on, that is a really significant factor for our partners because they know they've invested in an industry-leading solution, and they want that to remain ahead of the market. And fourthly here, just under toolkit, we have got very powerful analytics and really good tools. We're continuing to develop a playbook, which allows us to share critical success factors with our partners, from organizational design through to specific operating methods and reports.
I did just want to touch on, if you look at this chart here, the five boxes on this chart, they represent five sort of categories of the P&L. When we're going through these with the partners, the reason we've anchored it on the P&L is it gives them a really clear view of the materiality of the levers that they can pull. And for each KPI and each P&L line now, we've got a really good view, which we share with them, of what good and great performance looks like based on best practice across our 22 live sites. Let me give you a flavor of the way, that we've been going about partner success by just touching on a few of the projects that we've done.
We worked with two of our partners on a project to analyze the checkout walk that will improve conversion and optimize the basket size. Customers want to be able to checkout quickly and simply, but there is also a value in allowing customers to keep the basket open as late as possible. So we're able to do A/B testing to assess the impact of different proposed changes on customer behavior. In another assignment, we worked with a group of partners to measure the cost of customer acquisition, we call CPA, and to identify the conversion of customers at each stage of the customer journey to make sure that we haven't got customers dropping out at a certain point. And that helps partners understand how much to invest in marketing overall, how much to invest in customer acquisition, and how much to invest in retention.
In another project, our team did an in-depth review of the marketing for a specific customer banner. That identified opportunities to drive performance of weaker landing pages and to improve search capabilities. In that, we worked with the partner's agency to review spend and traffic by channel. If you don't get it right, waste is a very significant cost for our partners. Products approaching their sell-by date can be cleared by flash sales. Our demand forecasting system can instantly impact future order quantities to build in learning. Project that I'm referring to here, waste charges at the sites were reduced on an annualized basis by several million pounds per year. We've been requested by a number of partners to review their organizational design, which might come as a surprise, but we have.
That's it that helps us share with them our view of what critical roles and responsibilities they need to adopt to make a real success out of these facilities. Lately, we've completed a project to review a partner's detailed shift patterns to help them improve productivity and slot availability. I mentioned in-store fulfillment. We've consulted several partners on using our ISF tools. We do detailed store audits to optimize the pick walk and improve customer fulfillment rates. Finally, as they grow, most of our partners are focused on improving last-mile productivity. As I said before, that's fundamental to achieving good economics.
Our last-mile team has been helping show our partners how to monitor slot availability and to optimize route fill, sharing methodologies, and determining which geographical areas it's most advantageous to serve, using our OSP functionality to re-zone and rebalance between the service offering and route fill. I hope that just gives an idea. It gives a sort of glimpse of the type of work that we've been doing to accelerate our partners' growth and improve their profitability. Thank you.
Thanks, John. Okay, just a quick, quick few words on OIA. Why are we in this space? We believe that our solution has clear competitive benefits in the wider automated storage and retrieval sector. We're able to offer a unique proposition because we're an operator ourselves with over 20 years' experience of running a wide variety of material handling equipment here in the U.K. So we're in a unique position to compare shuttles to overhead storage systems to conveyors to cranes to everything else because we've, we've owned and operated every one of them. And we've done so with probably the most ruthless focus on whole-site efficiency of any MHE operator in the world. So our experience is something that is massively sought after. We have engaged over the last 10 years in extensive R&D investment.
We have an extremely robust roadmap for automating further processes and all the processes across the fulfillment value chain. We're using in our systems proven expertise that creates very sophisticated dense storage and orchestration algorithms that allow us to create very high throughputs and drive very high levels of efficiency while maintaining full inventory traceability and also very, very safe inventory. It's very hard to take something out of grid that you're not supposed to, and the system doesn't want to allow you to. Our systems are more space-efficient than shuttle systems, over 20% more efficient than shuttle systems. We are able to offer the tallest available grids in the market. Our systems have the highest throughput and productivity. Our whole-site efficiency is over 300 units per labor hour today in what we can offer.
But also, the amount of density of bots, the bot control systems, the single-space bots, and the acceleration, de-acceleration, hoist speeds of our bots means that we can offer the highest throughputs, and higher throughputs than anything else out there. And because of Re:Imagined, we are price competitive. The new 600-series bots and grids obviously allow us to be price competitive, along with OGRP, where that needs it should be deployed on the top of the grids. And also, some new peripherals that we're rolling out that are significantly cheaper than the historical pick stations, etc. As you're aware, we signed the first OIA deal with McKesson in Canada, in November 2023. They are the largest pharmaceutical distributor in North America. And we have a healthy pipeline of OIA deals that is building at the moment. The OIA team are very, very busy.
We also recently showed ourselves to the world for the first time in a trade show in Las Vegas in the last few weeks and are at the MODEX show in Atlanta in the next few weeks. We'll be in Birmingham as well this year. So the first time we've kind of publicly shown our wares very, very prominently on large stands at public events, drawing even more potential customers to us. Up to now, McKesson and the pipeline has largely been built by inbound inquiry. We're very much now actually heading out and going to be building up the pipeline ourselves. So what is the size of the opportunity? I t's various industry estimates around the scale of the automation investment that needs to be made in the next, whether it's 15, 20, 25 years.
It looks like a total market of over GBP 1 trillion that needs to be spent to automate storage and retrieval-type facilities and warehousing facilities. We knock that down into an addressable market that we think is, in the economies of the world that we think that we should be targeting. When we draw that down into what is currently the serviceable market based on the product that we have today, but that actual 130 out of the 450 will be growing as we add more products into the mix and are capable of addressing more of the needs. If you kind of take that market and divide it by whether you think this market's going to automate in 10 years, 15 years, 20, or 25 years, you can then see the type of the scale of opportunity.
I think that there was more than enough space for another player, with our skill set in this market. And so there are billions of pounds of annual revenue opportunities to share among the players in this market. We would obviously like, over time, to become the largest. The size of an opportunity, clear paths to expanding it as our technology develops. Ocado Retail, let's move on to that segment. The Perfect Execution Program has been delivering results, including a stronger customer proposition. We are focused on the three bars there, unbeatable choice. We're now up to back up to 50,000 SKUs as we were before COVID. We now have around 90% of the M&S food range available on Ocado.com. And we've got new international ranges, including the Polish range, popular in my household.
The unrivaled service, we are over 99% items delivered as promised, 95% on-time delivery, and a great doorstep experience. Getting the basics right in online grocery is the most critical thing. It's amazing how I've been in this industry now for 24 years and at how few players there are in the world that can actually deliver what they promise to a customer on their doorstep and on time. It is something that our platform has always enabled. It's something that enables our clients to have high EMPS scores, sorry, NPS scores, I mean, because doing it right is doing it well. We are reassuringly good value. Our average sales price has increased significantly below market inflation over the last year as we have been investing into the proposition, and our pricing.
We relaunched the Ocado Price Promise to guarantee that our baskets are as cheap as shopping at the largest grocery retailer in the U.K. We have had thousands of product price drops. I know inflation has been a question recently. Inflation has been dropping again. Our ASP is still positive increase year-on-year, but it's very marginally down quarter-on-quarter, largely due to the investments that we have made in price to drive growth, future growth. Our customer proposition's back to where it should be. We have a positive NPS. The strong proposition is driving customer growth. You can see here where we were in terms of active customers in 2019 at 796,000.
The moves that we made in 2020 to deal with the massive excess of demand, both from our existing loyal customers and from other customers, meant that we actually ended up in 2020 with less active customers because we restricted access to the shop. We restricted even the people who had access to the shop to the amount of times that they could shop with us. It's obviously relevant to another conversation that we're having. But you can see that those customer numbers then growing back to 832, 942, 998, continual growth of active customer numbers, 9% growth in the mature customer base. Sometimes you can do different marketing activities. So in 2020, we were doing broader marketing activities that brought in higher numbers but of less quality.
What we've tried to do this year is refine back to get back to the lower CPAs that we were familiar with in 2019, going back to basics again or back to the future. And that has enabled us to have, whilst what looks like a 5.9% growth in active customers, actually a 9% growth in what we would consider to be the mature customer growth, which is where you really want the customers to be. We have materially improved the retention to fifth. We always say once a customer gets to fifth shop, they, you know, largely shop with us, you know, until they leave the country in one way or another, as well as an improved first-quarter value, which obviously, again, helps to quickly recoup the investment in customer acquisition the faster they shop with you. We have done that at lower costs.
So we have reduced the marketing spend, including vouchering, back down to historical levels around 2.6%. That's very familiar kind of territory, if you look back in the pre-COVID years from the increased marketing that we had in 2022. Our market share online is increasing, is up 0.7 percentage points to 13.5%. And so we are gaining market share on our competitors online. And we have confidence in our ongoing momentum. Sorry. We have optimized our CFC footprint, as you're aware, during the year, closing our oldest warehouse, Hatfield. Hatfield was open for, I think, 22 years. We opened Luton and have seen the fastest ramp in, probably the history of automated, single-pick facilities. And we've also returned to positive volume growth, meaning that today, we are using greater than 75% of CFC utilization. I will just be clear. That's greater than 75% of the design and drawn-down CFC utilization.
It is not greater than 75% of what we will drive that number to. So we will drive 10%, 20%, or 30% more capacity out of those U.K. warehouses in the future. But in terms of Hannah running a business that is paying for more capacity than she's using, she's up to the 70%, oh, ahead of the 75% level now. Productivity is growing. You can see the productivity there over a number of years. We're up to 199 in the fourth quarter of 2023. The exit rate of Purfleet was 225. It's been operating ahead of that. 230-something, I think, might even have hit about 240.
So the new warehouses are all operating quite meaningfully ahead of their design productivity prior to the impacts of Auto Frame Load and OGRP going into those warehouses, which means that whereas we were offering sites with a 330 productivity to our clients globally, with Re:imagined, that number will be progressing forwards. And we have got up to 21.5 drops per van route, which is per eight-hour shift, basically, back in 2023. So overall, improving capacity utilization, improving operational efficiency, underpinning the path to a high mid-single-digit EBITDA margin in the midterm. Lots of jargon’s in that sentence. Okay. So in summary, our Perfect Execution Program is delivering results. Our differentiated proposition is driving growth and growth in market share.
We are making the most of our OSP model and will, looking forward at Ocado Retail, to migrating completely onto OSP and then significantly looking forward to the Swift Router opportunities that we're going to see in 2025. We're improving our capacity utilization. We're operating efficiency underpins our path to get there. So all good. Moving on to Stephen.
Thank you, Tim. Hi, everybody. Okay. I'm going to be quite snappy with these numbers to allow a good set of time for Q&A. I'm going to start with our Technology Solutions business, which and the financial model for that business, which is a key driver of the profit and cash flow improvement over the year. But this will be increasingly the case. So let me just highlight just two or three key things here. A battle-tested IP portfolio in the top left, proven that with the case against AutoStore and that resounding victory there, sticky recurring cash flows. This is a business that has recurring cash flows that are very sticky, as our customers, around the world grow their own customers from our technology. We also have strong pipeline visibility.
We have very good visibility over the next 2 years to 3 years of specific sites that are going live. We also have very good visibility around how quickly we're ramping up into existing sites. And therefore, as Tim mentioned earlier on, the modules that are going to be drawn down on existing sites today, those 49 modules still to be drawn down on. So really good visibility. Top right, some clear value drivers. We have index link fees. We're protected from inflation. Our fees rise as inflation. So we're delivering a constant, if not improving, margin. Very high operating leverage as well in this business. You'll see the power of the business model as profits and cash flows flow through as an increasing number of modules go live. Incredibly strong cash conversion and resilient cash generation reflecting those recurring cash flows. And then returns on capital.
We highlighted Purfleet, a 22% return on capital employed at the seminars that we did in 2022. Post Re:Imagined, that number's going to exceed 40%. More on that to come in future presentations along these lines. Looking at Fiscal 2023, we delivered or beat all of our guidance. I'm not going to go through these numbers individually. This is a habit that we plan and hope to get into, delivering on or at least beating guidance. Again, looking across our EBITDA and revenue performance across all of our segments, positive adjusted EBITDA in all of our business segments for the first time. Again, I'd just point out a couple of numbers here. The improving trend in Ocado Retail, moving from a loss-making situation the first half into nicely profitable in the second half. That's an improving trajectory that's continuing. A couple of other things just to point out.
Finance expense staying flat at around GBP 62 million or so. The FX losses driven by basically the translation of our dollar balances on our balance sheet, our dollar amounts on the balance sheet. Adjusting items, we've got a 24 positive number there of adjusting items. Here's the breakdown. Top right, the AutoStore settlement that you read about earlier this year, the change in the IFRS 13 fair value of contingent consideration. That's the Marks & Spencer earnout, and how we've written that down to the GBP 28 million that we had for the full year. The ORL UK Network Capacity Review, principally two components here, the Hatfield cessation of operations that we talked about earlier in the year, and also an optimization of our strategy for Zooms, particularly focused on London.
Organizational restructuring of the headcount reductions to deliver the savings that you're going to see in the group support costs. Litigation costs is the tail end of the AutoStore settlement, legal costs. Moving on, please, to the next slide. Thank you. Technology Solutions. So here's that powerful business model that I talked about. Live modules, we've highlighted that in blue there. That's the kind of key row. That's the key driver of recurring revenues and cash flows for us. So there's the average that Tim mentioned a little earlier, 105 average live modules. Last year, it was 84. That number will grow as more modules go live as we open more sites and we grow into existing sites. It's going to grow materially over the next three or four years. And a key driver of that delivery of cash flow positive.
Recurring revenue I've just talked about there. Non-recurring revenue in that number is the released to the P&L account of prior upfront cash that's been received that's sitting on our balance sheet right now, on the line in the contract liabilities balance. Very high contribution margin that I mentioned, turning 70% of that revenue into contribution. Technology costs, going up slightly, particularly as we're rolling out our Re:Imagined innovations that Tim talked about.
And support costs coming down quite materially, 8% down despite investing a further GBP 8 million in partner success to support John and his team, deliver success for our partners. 91% EBITDA flow through from revenue growth, again, a reinforcement of that operational leverage message that I shared a little earlier. Next slide, please. Yield per module, growing to GBP 3.5 million from GBP 3 million last year. Again, that's, the indexation is a key driver of that.
I'll just highlight also the increasing mix of OSP sites within that. As more OSP sites go live, we've ceased operations at Hatfield, the average fee goes up accordingly. Post Re:Imagined rollout, we're expanding with that growing capacity as well, we're expecting to deliver OSP recurring fees of 5.5% of live sales capacity. So an improving number for yield per module. Next slide, please. OSP direct operating costs, you'll see the improvement first half, 2021, second half 2022, second half 2023, 1.65%. Our midterm target remains at 1.5% there on the far right. And you can see the drivers of that.
That accommodates, of course, the impact of new site mix and new automation that are going to be headwinds, but through to utilization, but also the efficiencies that we continue to deliver in this cost base. We're going to be driving a reducing operating cost and still on track, well on track, in fact, for 1.25% of operating costs in the long term. Next slide, please. Technology investment is peaking. I know there might be a sense of a bit of déjà vu over or around that statement, but it really is. We're very much holding to our mid-term target of GBP 240 million. And this is the peak of the, as Tim alluded to, again, in his presentation, of those Re:Imagined rollouts. ICA in Sweden have taken automated frame load. We're expecting others to follow either with AFL or On-Grid Robotic Pick.
You'll read more on that over as the year progresses into the year after that. Next slide, please. Support costs are coming down while investing in the customer. I talked around that reduction. How have we delivered it? Efficiencies from investment in enterprise platforms. The HR team, the finance team have invested in their systems to deliver, ultimately, efficiencies that will be durable efficiencies, very quick payback from that. There's more of that we can do, actually. Ocado is a business that has grown fast, and has properly invested in the customer rather than the back office over the years. I think now that we have a bit of a healthier sort of, let's say, liquidity and cash flow generation, improving cash flow generation, investing in that space to deliver efficiencies is the smart and wise thing to do. There's much more to come, in fact.
I think there's a lot of duplication in our business that we can eliminate. I think we'll continue to find efficiencies. Next slide, please. Ocado Logistics, a very reliable generator of around GBP 30 million of EBITDA for us, a cost pass-through business. Customers Ocado Retail, Morrisons. It's a very nice business to have in their portfolio. Its key job is to bring down its unit costs because if it brings down its unit costs, Ocado Retail and Morrisons make more money. That's what we like. Next slide, please. Ocado Retail, Tim walked through the key dynamics here, but 11% growth in the quarter in Q4 for Ocado Retail. I'll just call that out. That's the improving trajectory that I talked about a little earlier. Moving on to the next slide, please. Here's how we get to the EBITDA number.
GBP 10 million EBITDA, again, noting the second half performance versus the first half performance, an improving trend. Marketing costs are down materially last year, but we have put more money into vouchering as well, which is further higher up on the P&L account. Support costs are showing quite a big increase here every year. But we shouldn't forget that we had a significant incentive accrual release in fiscal 2022 that benefited that GBP 83 million support cost number that we see there. Next slide, please. So moving on to cash flow. And again, so our underlying cash flow, we beat by some way the guidance that we gave, a GBP 356 million improvement driven, of course, by the lower CapEx, with less sites going live and under construction than there have been previously, but also a GBP 128 million improvement in EBITDA.
A really good improvement in cash flow, core focus for us, turning cash flow positive in that medium term in the next 2 years to 3 years, absolutely critical for us and a real important priority for Ocado. Very much on track to do that, driven by the visibility that we have of those site openings and module go-live trajectory. Next slide, please. There's the reduced CapEx. Site CapEx is the one that's moved down most materially.
And again, the explanations are up there at the top. I wouldn't call out anything other than that, actually, on this slide. And the next one. So we're on track to be cash flow positive in the medium term, growing recurring fees I've talked about, the operating costs, tech technology investment, and group support costs will come down. Watch this space on that. And our CapEx unit costs continue to decline as well. Next slide.
Steadily improving cash flows. Now, what we highlight here is the underlying cash flow pre-growth Capex going from GBP -399 million to GBP -203 million. We expect that number to be well over 200 by the midterm. And that will therefore be able to fund the levels of CFC Capex spend, the typical levels of spend within the last two or three years. Clearly, that's a key variable in the overall net cash flow positive number. A big order for CFCs going live, let's say 20 CFCs going live in a year, would drive material Capex spend, but we would be a very good outcome. We wouldn't mind that outcome. We'd manage through it. But I'm just trying to make the point that cash flow positive will very much depend on the number of CFCs and sites going live.
Notwithstanding that, though, the underlying number, well over GBP 200 million by the midterm. Next slide, please. So there's a summary. Strong momentum, delivery of our all of our guidance, either online or beating it. KPIs continue to progress across all of our segments. Underlying cash outflows have just run through. Pre-growth Capex up GBP 196 million, well positioned to deliver our midterm targets. And then Fiscal 2024 key priorities, financial discipline. Really made, I think, a big effort at the whole leadership team to get a really strong focus on costs, on budget management. There's still a lot of room for improvement, but it's, I think it's something that's becoming part of the conversation now that, that we can that is actually helping us think harder about the choices that we make. Operational execution, partner success, absolutely critical.
We need our partners to succeed and get attractive returns from their investments. Then strategic growth. Well, there you go, win new partners, non-grocery contract wins. Ocado Intelligent Automation, good pipeline of business, is now turning that pipeline into new contract wins. Guidance for next year, Technology Solutions, 15% to 20% growth in revenue, greater than 10% EBITDA margin, logistics stable. There's that GBP 30 million number for EBITDA. Ocado Retail, mid-high single-digit growth and a 2.5% underlying EBITDA margin. But calling out underlying is after excluding the GBP 33 million of Hatfield fees that continue to be charged in full by Technology Solutions to Ocado Retail as contracted. That's an important point to bear in mind. Then Ocado Group, underlying cash flow will improve by around GBP 100 million and CapEx will be around GBP 475 million. That's my financial summary. Then we'll hand over for Q&A. Hello.
Oh, there we go. There you go.
Thank you. Hi, it's Sean Stewart from JPMorgan. I've got three questions. The first one is on Swift Router. It's sort of, you're very excited. Could you just explain a bit more where you want to take this? So what are these additional orders that you basically then prioritize? I'm coming from a very basic view that if I log in my order tonight, I get it tomorrow as well. So, if you can just, I understand it's a basic question, but since I have you, I would like to, to understand it a bit more. Then the second question for John, on partner success. Clearly sounds very important.
Given the developments we had in the past at some of your clients, it's absolutely necessary and clearly the right thing to do. But could you to talk a little bit more about the UPH developments, at least anecdotally, are they really catching up with Ocado Retail? I think that has been the problem in the past, that the UPH rates were just not really coming through. I understand you can't give precise numbers, but I think it would be helpful to give a bit more anecdotally. And then the third, maybe for Stephen, on CapEx, I think it's very good that you show the chart ex-growth CapEx. But given where we are and all this discussion about balance sheet, isn't it just maybe a thought process to just say, the clients bear the CapEx first?
Yeah, it comes down to financial math effectively. But why is it that Ocado has to bear the Capex mostly upfront, when your clients are maybe, as of today, in a different financial position? Yeah, that would be interesting. Thank you.
Thank you. And thanks nicely to have split the questions up. So on Swift Router, if you look at some of the markets that there are where there has been a lot of short lead time orders, something like 40% of the demand can come in the next few hours, you know, it is currently being served by customers in the kind of sub-four hour type segment.
Now, if you suddenly said to those customers, "Here's a GBP 5, $ 5, EUR 5 cheaper service that delivers today for tomorrow," some customers will migrate, but some of them still would love to have it in the next four hours. The point about Swift Router is it enables our clients, the users of our platform, to be able to use the full operational efficiency of the warehouse, to pick overnight the orders that were placed yesterday for today, but to use the daytime picking for orders that are placed just before they're picked and loaded onto a van and shipped out. So we move from running kind of a morning delivery routes that go out in quite a narrow window, and then we have no dispatch.
Then they come back, and we do an afternoon dispatch that goes out for the next eight-hour shift and move towards having more continuous dispatches going out through the day so that there are always the opportunity for customers who want to pay probably a small premium, but that will be worked out by the retailer and the algorithms advising them, to say, "Well, what, you know, it's 10:30 AM. I'll place an order now. I'll be out of the shop at 11:00 AM. If I'm out of the shop at 11:00 AM., it can go onto a van being dispatched between 11:00 AM. and 12:00 P.M. And if I manage to get one of the first two or three customers on that route, I could have that shopping at, at, I don't know, 1:00 PM. or 2:00 PM. or whatever it might be," right?
So it's how can I get that shopping really quickly to a customer's home while still having all the efficiencies of the large, centralized, automated, optimized routes at the same time? Now, running if you can, there are different variables here. So we can't give an exact number because if you run it with if it's a very large site and you still, despite the increasing demand, need to run spoke sites, you've got to put that time gap in, the transport from the main site to the spoke site. But if you're looking at running it in the, you know, 1.5 hours type drive time around the warehouse, you, the increase in density can more than offset the inability to come back mid-route.
So you can start doing what we've done for years in one of our central London spokes that has drops per 8 hours of 27, 28, or something like that. We've run mixes of 6-hour shifts and 10-hour shifts to average 40 hours rather than 8-hour shifts. And then on the 6-hour shifts, you can unload a whole van. On the 10-hour shift, you go out twice, and you do two 5s. So if you start running shorter shifts, then you've got more of the opportunity to deliver from when they were loaded to when they were finished. The other types of opportunities it will allow us is today, we ask some customers to leave the shop by 6:00 PM for the morning deliveries so that we can start picking them at 7:00 AM.
We are going to be able to, with this software, to pick your order at 7:00 PM. but still to allow you to come in and add items. And if you want to add an item at 12:00 AM, then we can take you know, one of the robots can pick up one of your boxes that was already completed, take it back in, drop it through a pick station or a robotic pick arm, add the extra items to it. If it happens that you had an overnight order, and it happens to be optimized to go out on a van, and the van's going to be loaded in 15 minutes at 10:00 AM, then at 9:45 AM., we can send you a push message that says, "We're about to load the van.
Are you sure you would you like us to add any last-minute items?" Last-minute items add sales. They add basket size, and they add significant profitability because that van's already driving past your home. So we are very excited about what, what, what the combination of this tech technology can bring. As I say, it will have more impact initially in markets where there is currently short lead time demand that we can't that we don't serve from or our clients can't serve from the warehouse. It'll have the most impact there. But we think it will also have impact in markets like the U.K. where, whilst we've grown up on today for tomorrow, I think customers will be super excited to realize they could shop at this time in the morning and have the food for lunchtime or for dinner.
Yeah, UPH.
So, the good thing is we have many of our facilities certainly, if you look at the more mature facilities, those have been open longer. They're really going up the UPH curve very effectively. So, I think we have a high degree of confidence that the large majority of facilities are going to go up a fairly similar curve in terms of UPH. And if you look at the operational aspects of partner success, we've got a very, very experienced team there who've done this in the U.K., can go out and help our partners to figure that through. Two other points, though. One is it's not just UPH from our partner's profitability perspective. Labor rates are very high in a lot of the economies that we're in. Clearly, that makes our model more attractive long term.
The way in which they can address labor rates being higher than in the U.K., for example, is just by more use of automation. Auto Frame Load, auto-bagging, OGRP are all going to reduce reliance on labor rates. But yes, I think the UPH curve is well-trodden, and we'd be pretty confident that the facilities are all going to get to decent, you know, good productivity.
Then finally, the question on the Capex and maybe the clients picking up the bill there for that. Well, first of all, we have explored this. We've got no plans to change our business model as it stands today, though. So we have a sort of, as you know, a licensed model, which we think works very well with these sort of attractive, you know, financials that I talked through.
At the same time, we shouldn't forget that, around a third of the Capex that we put down in the CFC is funded for by our partners upfront. So and then what we now have with the Mark VI grid and the 600 bot is an opportunity to have more of a modular approach to our Capex building sites to effectively become, self-funding as we progress with, with future Module rollouts as well in our sites. So I think those two things together will, again, reduce the burden on, on Capex that we're putting into our client sites.
Thank you for two questions.
Great. That's perfect because I have two. Tintin Stormont from Deutsche Numis. On partner success, could you remind us of the charging model, that you have, with your partners?
And to talk about the scalability of the teams, do you ever envision that, you know, just as, for example, there are systems integrators and consulting firms that work on behalf of SAP or things like that, that in the future, that is something that, that's a model that could be explored? Secondly, on intelligent automation and the McKesson win, could you just give a bit of background on the evaluation and the competitive benchmarking that the customer went through? And Tim, for you, in three years' time, what does success look like in that space?
Yeah. So on the charging now, so today, we have done a number of charged-for consulting assignments. That's not our prime focus today.
Our prime focus certainly my prime focus since I've come on board is to make sure that all of our partners are engaged, that we have local resources engaging day in, day out, and the central resources to bring in for, you know, deeper expertise and deeper and deeper projects. In terms of scalability, I think there will be in the evolution of partners as they come on board now, and we're engaged with partners before they come on board, which is really important because you want to hit the ground running. That's one of the reasons for getting ISF into our partners, as well, pre-going live with CFC. I think there will be a period of, you know, sort of more intense partner success focus with our partners.
Some of them are saying, "Look, you know, you've done it with this one CFC. Thank you. Now we will take those learnings ourselves and implement them." So I think that's absolutely likely. In terms of scalability, the model is absolutely scalable. You know, the more expertise that we have, internally, the better, the more it is, you know, seen as a really viable place. We have lots of people who want to come and spend part of their career in partner success because they understand it's quite commercial. You know, it isn't, you know, it isn't generic and soft and woolly. It is, "Can I make a difference to the partner's growth rate and profitability?" In terms of could we have a different mechanism going forward, I think that's quite possible. It's just not our focus today.
But we have got one or two people who we've specifically hired to look at, for example, the consultancy model. I think that's quite likely, at some point going forward.
Just to move on. I need to obviously respect all of our clients' kind of confidentiality and stuff, but I wasn't personally close, so I can probably talk in more general terms. Certainly looking from the outside, McKesson is very experienced in acquiring automation, and they have acquired automation before and had a standing partner in that space. So one could imagine that the merits of our solution meant that they affected a change there to do it.
As I said before, our solution has a lot of advantages in terms of its space density, its throughput capability, the security with which it holds the inventory, which, as you can imagine, in that field of pharmaceuticals is very useful, that you can't just kind of walk up to the machine and start trying to, you know, rifle through anything. And so it had a number of advantages. As I say, they're not a first-time customer in automation. They're not an inexperienced customer in automation. It's not a random win, if you see what I mean, that we should discount. It's a very smart, very sophisticated business that has dozens of warehouses across North America.
And it was, you know, a good win for the OIA team leveraging on Ocado's long track record in this space, not just in OIA but in general with the automation, and running the facilities.
And success in three years?
What would success look like, I think we'd like to become a large-scale player. So it's, you know, we can see whether it's the Knapps of the world or the Schaefers or the Van derl andes or the AutoStore or the other players in there can generate revenue in storage and retrieval systems. We'd like to be a mainstream player. So, we'd like to be billing hundreds of millions GBP of client wins. Well, billing might not be the right word.
We'd like to be signing contracts worth hundreds of millions of pounds a year in that kind of time frame.
Okay. Thank you.
It's Luke Holbrook from Morgan Stanley. Just on your nine module deployments for this year, if we strip out the three new CFCs that are due to come live, it looks like your existing customers are only deploying maybe one or two modules. So what's driven the extent of that slowdown? The second question was, I think you ended the year at GBP 105 million of capital commitments from your existing partners, but your Capex guidance is GBP 475 million. So are we to expect that the delta between the two is intangibles, and what would be driving that increase? Thank you.
So I'll take the easier half of that question, which was the first part.
As I mentioned before, the way that we work with our clients historically was that in an initial warehouse program, they would start with, let's call it on average I don't have the exact average in front of me, but let's call it kind of circa 50% of the capacity. And we ask them not to draw down other than pre-committed. So they can tell us at the start, "I want 3 out of 6, and three months in, I want number 4," but then they can't come and ask us for number 5 a month later, right? What we ask them to give us a large window where we go in and put that kit in, right? So the sites that we've put live, as we mentioned before, have got 111 drawn modules and 49 undrawn modules.
So they are predominantly drawn, but the clients obviously sometimes some of them are migrating some business into them, but most of them outside of the U.K. have been around new businesses. So they're growing into their module capacity, right? If they'd built a normal warehouse with a normal MHE vendor, they would have started with 100% of the capacity, and they would just be slowly utilizing that. We enabled them to only have to take half of it or something like that, right? So they are growing, but they are growing into what they already took. Until they get there, they don't need more.
So we have one site, for example, internationally that has seen constant small drawdowns that we're doing on it because it's hit that level, and it's drawing down, and it's up at north of 70% of endgame capacity, and it's drawn down several times this year. We're able to draw down in smaller increments than a module at a time. As we go forwards, we will enable because of the work we've done with Re:Imagined and site designs and because of what we can achieve that Stephen was mentioning before on the Capex front, we can be smarter about how we deploy new sites. We can deploy new sites with, let's say, 15% or 20% of the capacity upfront. And then you will more quickly after launch expect to see the underlying growth in their business causing a drawdown.
At the moment, the underlying growth in our client's business is predominantly filling what they've already drawn down. And as I mentioned earlier, there are, somewhere just under five sites where they are coming up to the limits of what they've drawn. And so I'd expect to see more sites drawing in 2024, but most of the work in 2024 is to get the balance of the sites to the point where they are then needing to draw in 2025. That's why we expect to see more drawings in 2025 from existing sites from the 49 modules we mentioned earlier, plus I think I said there was 11 more modules in the sites that will go live in 2024 and 2025 that are undrawn at go-live. So there'll be 60 modules to draw down.
We need to grow and help the clients grow their underlying business, so they need to draw down those modules.
Then you had a technical numbers question.
Okay. So yeah, technology CapEx spend will be the majority of that difference, but it will be flat year-on-year with the 2023 year. The balance will be MHE CapEx, material handling equipment CapEx for go-lives in 2025, 2026, and 2027, early spend in those sites. That's the key difference.
Okay. Thanks. Sorry, just to clarify. Just one on the May 2022 seminar, you said 30-module run rate of deployments. I think you caveated the interims last year. That was more like 20-25. But from what you've just said, is it what you've shown today, is it fair to assume that's more of a low-teens run rate?
I think we'll see low-teens 2024, growth, quite a bit of growth on that you know, quite a bit of growth on that in 2025 and significant growth on that in 2026. So I think that, you know, we don't know whether it's 60, 70, 80 over that three-year time frame, something in that kind of region, but weighted back more back weighted because of the filling of the existing modules in terms of the partner success and because of the rollout of Swift Router beginning 2025. So kind of a growth trajectory there. But over the three-year 2024, 2025, 2026, something in that something in that region of kind of 60 to 80 new modules going live is probably approximately right. Obviously wait to see about you know, there'll be some new sites that could go live in 2026 that we could sell in 2024.
So we don't have full visibility yet. And then there's the whole point about how quickly do the clients fill the sites, but I would think it's in that range.
Good morning. Thank you for a sort of self-critical look at the business. It is appreciated. So thank you. Just two questions. Firstly, every time we come and listen to you, you have a sort of latest innovation for the business. What point do you say, "Look, that's enough. You know, this business is now optimized for the demand outlook we have?" And then tied to that, a question for Stephen. If you think about the chart there of GBP 300 million or just below GBP 300 million of technology spend, how much of that is really into these sorts of new initiatives? So I suppose the development so not rolling out existing technology, but thinking about what's new.
Yeah. Let me just give a couple of examples to your point. Look, we don't want to stop innovating because innovation is, we think, good for us and good for our clients. But let me give you a few bits of information. We previously stated that an Ocado Re:Imagined warehouse would have warehouse productivity of 330 units an hour. I think based on some of the productivities that we're currently seeing, whether it's in the small site in Bristol or in Purfleet or in Andover or in Luton, I think that number is grown at 350 units of productivity. The operating costs at U.K. labor rates of that site, including management labor to the management to manage that labor, is about 2% of sales.
So going from 350 productivity to 700 productivity through the development of really clever technologies would only save our clients about 1% of labor costs, going from 2 to you know, it would save them 50%, but it would go from 2 to 1 in the warehouse. So there's a diminishing need, and it's a big difference going from 150 that we were at in Hatfield to 350 has saved a material amount going from, as well as the advantages of the orders being processed in 10 minutes, not in three hours, and the other things that we can do at Swift Router. So we've taken a massive step from the kind of Hatfield/Dordon designs to what we can deliver with Re:Imagined. There is clearly a slower demand to try and drive 350 to 700. Now, as it happens, I would have said, "We don't need to go any further.
OIA is driving some demand from clients or potential clients for some capabilities that will naturally help drive the 350 to 400, 450, or maybe even 500. But the OSP business itself would not go out there and do very heavy investment to go beyond 350 at this point until we had dozens of warehouses live and hundreds of modules live because that last 1% for our clients is, you know, not worth 100s of millions of our CapEx, to your point. Driving CapEx spend in improving our customers, our clients, partners, routing efficiency is still very meaningful. It's a lower level of CapEx to do it. And if you take the most efficient if you think that's a ORL, we're at about 11%. So a 10% reduction in delivery costs is that same 1% saving that a doubling of productivity in the warehouse could make, right?
So we're continuing to invest in that area. We're continuing to invest in the front-end technologies in particular to make our clients, customers find it easier to shop, to reduce the frictions, and the behavioral science stuff that means that it takes the three to five shops to become loyal. So if we could make it much easier, then that might happen at one to two shops rather than three to five shops. That'd be very meaningful for everybody. So there are areas of focus. We want to finish delivering the Re:Imagined products, make sure that they are reliable, and that the first stages of cost down on them are as we do because there'll be some parts of, you know, you can make a material cost saving with a bit more engineering.
We've had to develop the 400-series bot and then a second derivative of it. We then developed the 500-series bot, the 600-series bot. Those are expensive programs. We don't have a 700-series bot program. We've not started one. So we were doing the 600 program while we were in the first stages of rolling out the 500. We have been developing and have developed a freezer version of our 500-series bot so that we can run and build automated freezers in a scalable and modular way because the third-party freezer automation supplier turned out to cost double or triple what we were expecting, when we were looking to do it using existing third-party kit. So those are the type of areas where, yes, that's why Stephen's talking about we can reduce the spend.
And I think your question really is what's kind of, I think part of your earlier question is, what's part of kind of maintenance CapEx of running these platforms versus what's development? So if you were to take.
well, it's sort of three pieces of that. So you think about the CapEx at the moment. There's maintaining what you've got. There's building new sites for partners. And then there's a third bucket, which is the sort of R&D. Yes. And then another that's including the OpEx part of the R&D.
If you looked at the techs the total tech spend, so the CapEx portion and the cash, as we showed, and the CapEx and the P&L part of it, it's roughly speaking about 30%-35% is what you would need to spend to maintain scalability, resilience, security, upgrades to, you know, Chrome and iOS and, you know, AWS and Google Cloud and, you know, to kind of you wouldn't want to go under about 1,000 people, and we're operating at about 2,500. About 1,500 are involved in the development of new, clever things, and that's the part that you could reduce and go at a different pace when you felt like you've kind of got over the hump of developing a lot of core stuff.
You could have a lot of people doing very clever things, but you've developed the core to such a point where you're doing more clever things, but overall, you've got less people.
Thanks.
Thank you. Good morning, William Woods from Bernstein. The first one is, the, the half-year results. I think you talked about getting 10 new CFCs over two years. It feels like the 5-5 or 4-6 split has now gone to 3-7. Are you still convinced that you'll get the 10 new CFCs, and is that dependent on Kroger coming back? And then the second question is on Kroger. We obviously saw some improvement in Project Endeavor, but you've said very little on the actual relationship there. I suppose two parts. Why hasn't or has Project Endeavor gone further than Monroe and Groveland?
Do you think you'll see any thawing of the freeze this year? Thanks.
I'll let John talk in a minute, but I would just say that from a relationship perspective, we have an excellent relationship, deep and excellent relationship. We have visibility of the warehouses that are likely to open in the next two years because they usually take two years to build. Beyond that, it's kind of not so easy to do because you could have a client who asks for you know, could have a client mid-year who says, "Build me four more," or might not have one.
So we're not yet at a scale of clients and client growth that's predict that we can then say reliably in a way. In the retail business, you can look at the trend, and you can say, "Okay, every week, I acquire x thousand new customers, and why have them convert?" And, you know, you try and move those numbers over time, but it's not very difficult to project out 12 weeks and 20, you know, you get increasing levels of inaccuracy, but it's pretty easy to look forwards. Right now, could we get orders for 10 new warehouses this year that we would deliver in 2026 or in 2027, maybe? Could we get orders for one, maybe? It's difficult to know. I don't know if John, if you want to talk a bit more about the yep, Kroger piece.
It's so, I mean, I think if you look at the relationship today, it is intimate and getting more and more intimate over time in the sense that we've got a team in Cincinnati. We're putting more people on that team. And there's a very good exchange of all the data, all the information, all the ideas, all the action plans. So, for example, you know, last week, we had a very large team over doing three days with the Kroger team of integrated planning, really down to the detail, who does what next in order to improve the profitable growth.
I think in terms of taking the learnings from individual sites and implementing them across the network, you know, yes, there is good evidence that there are quite a few sites making very good progress. Of course, we all want it to be, you know, faster, further. You know, we want to see better numbers. You know, all of those things, and that's our job too. That's our job to do. But I think we are pretty, you know, we're pretty focused. We're both pulling on the same end of the same rope to get it, you know, ever-improving growth and profitability. And I would say, you know, growth actually at the moment, growth is the, you know, it's pretty positive.
Sorry, just to clarify on the 10, I think you said that they were kind of under construction or being built over the next 10 years. I appreciate you can't see what you might get in the future, but are you still comfortable with the 10?
Yeah. I can refer mention that one. So we said sites. We were which is a mixture of CFCs and Zooms. So we have 7 CFCs going live over the next two years, and then we have a further 4 CFCs going live in the year after that as well. So we have good visibility around those numbers. The balancing figure are Zoom sites either in the UK or overseas.
Understood. Thank you.
Yeah. Thanks, Charles Allen, Bloomberg Intelligence. Sorry, just a few sort of detailed ones.
On the, you said GBP 33 million that's going to be paid to Ocado Technology at least next year. For how long does that go on, the contractual arrangement, to pay?
It's index-linked until the end of the lease because that was the originally envisaged closing date, which was 2032.
Right. So that's GBP 33 million a year for another eight years. Correct.
It's GBP 33 million a year index-linked for another eight years, yes.
Right. And is that therefore the same for counting the 14 modules that you have are including in live modules, although obviously, the Hatfield ones are not actually working?
Yes. They're live, as in the live, as in other words, we're saying paid for.
Right.
So we have an arrangement whereby if they replace they can replace them with new modules in new sites.
So if they want to build a new site, then we will swap the commitment from this site to a new site.
Okay. So, a new warehouse well, so you probably could have at least two new warehouses that could be built with the modules being paid for.
If Ocado Retail came today and said to me, "They'd like to build a new six-module site to go live in two years' time, but they don't want to pay for Hatfield and the new mod and the new site," we would agree to that, is what I'm saying right now, if that's what they wanted to do.
So that GBP 33 million will be paid until 2032, unless they ask to replace it with something else that we're happy to then say, "We'll swap one out for the other," because one would be a short-term commitment till 2032, and the other one is a longer-term commitment. And the fees on Hatfield are lower than they are on other sites because it came with lower productivity.
Okay. Okay. And then sorry, just one other one. GBP 27 million to write off the lead Zoom site. I mean, that seems quite a lot. Why is it so much?
We're looking at the whole overall optimization of the network there. And when you close these sites, you have to take charges for the length of the lease until you dispose of the lease. So if you've got a lease, you take charges.
And we've also taken some charges against the Tufnell Park site that has never opened, that's been a contentious site for a long period of time, and one that is not necessarily a core part of the network at this point in time. So we've taken some charges against that as well, which was a Zoom and spoke site.
Thank you.
Good morning, Nick Coulter, from Citi. I'll go one by one to save you scribbling, Tim. Firstly, do better module or modular economics encourage your partners, or do you want to encourage your partners to build smaller CFCs? And then, I guess, link to that, kind of better economic run rate. It does Swift Router, you know, what does that mean for Zoom? Are you kind of getting to a continuum of warehouses where Zoom perhaps isn't needed within the portfolio?
That's the first one. Thank you.
Nick, two great questions that we could dispute half the morning on. So smaller CFCs are better for more short lead-time deliveries as that becomes an advantage because you are departing from a site closer to the customer. But larger warehouses do have economies of scale in rent and security and comms lines and managements, right? So our automation itself is proving to be incredibly modular. So we have got within about 5% to 7%, I think, of the productivity of the kind of six-module sites in the two-module site that we ran in Br., i.e., Bristol versus, say, Purfleet. The productivity in Purfleet is a little bit higher but not much higher and would be easily offset by savings in transportation.
But there is a waste calculation because if you want to carry the full range and there is just a property occupancy costs, I should say, that need to be taken into account as well. I wouldn't build three 2s that were all in half-hour's drive of each other. But similarly, I would avoid building a 10-module site in one city that needs to be trunked 2 hours to another city to offset its volume. I'd rather have a six and a four. There was another piece of software that we have not yet delivered as part of Re:Imagined, which we call Orbit.
Orbit is something that we will deliver but probably not. We've started to deliver little components of it but not all of it yet, and not all of it in 2024, probably in 2025, which will enable the smaller sites to have lower stock holdings, lower waste, and similar inbound economics to the bigger sites. That's not yet available but will help drive the ability to have multiple smaller sites in the same kind of geography. There are some places where it might be more interesting. If you think about seismic zone, for example, it might be better to have, you know, four 3s than one 12, for example, just from a risk perspective, or where you've got a city with, you know, very, very bad traffic issues. It's very much a dependency. But yes, we can build them small.
We can turn them on small. The economics on them are really good despite them being much smaller. The Zoom one is, I think the way I would think about it is it's the sub-one hour is the Zoom space. The kind of two hours-plus or maybe it's even one hour, one and a half hours-plus is the warehouse space. Definitely, it erodes the need for many consumers. A lot of people use Zoom today as a four-hour service, right? I need something four hours. Standard Ocado.com can't do it. I will use Zoom. That ought to be able to be serviced from the standard warehouse.
Therefore, the immediacy warehouse needs to only focus on the real immediacy need of, "I want it in the next one hour," or in areas that, for whatever reason it is, are not close enough to a large warehouse, they still get deliveries, but their minimum delivery lead time might be six or eight hours, then obviously, Zoom could play a part as a fill-in kind of service. And that's why you've got to look at it all as a bit of a network. But it is definitely true that Zoom, as a result, is more for the densest populated areas, and that ultra-short lead-time deliveries. It is worth us remembering that it is only about 18 months ago that the likes of Gorillas said that they were going to be larger than Ocado Retail in the U.K. in 2023.
I believe that the combined sales of Gorillas and Getir was under GBP 150 million, and Hannah's is two and a you know, retail's GBP 2.5 billion. So, you know, they were out by a factor of 30 or something on that on the on that assumption. There is demand for these short lead-time services at a premium price across many markets that we see it. And Zoom, we will develop Zoom to be the best way of, dealing with that if there is sufficient demand. You've got to be able to have demand of GBP 10, GBP 15 million of annual run-rate sales from, you know, in a 5 or so-kilometer radius. And if you're not going to do that for sub-1-hour deliveries, then don't o you know, then, then probably it's not demand not good opportunity for it.
Great. Thank you. Orbit is where I was heading.
Then, then the second question is on OIA. You kind of mentioned greater than 300 UPH, and it sounds like you've got ambitions to go considerably further. But when you think about is there a you know, is there a capital ratio that you can reference? Because that's probably the sticking point. Is there throughput to capital, or how do you think about that? And how do your potential clients think about that, please?
Sure. Look, I think you know we are unbelievably focused on not on technology for technology's sake. The reason that we spend so much R&D is to be able to create cheap-to-replicate unbelievable technology. Because there are some things that we work on in a On-Grid Robotic Pick, for example, to go beyond probably 60% range penetration of velocity, you need something other than the simple suction grippers that you've seen working before.
We can buy end-effectors that do that, but they are tens of thousands GBP, sometimes into the GBP +100,000 if they're actually effective and have kind of sensitive operate more like a hand in terms of understanding that they don't break the glass when they do that. We're trying to build those for single-digit thousands GBP so that we can create solutions we can give to our clients, and help our clients to save tons of money. I think if you look at our kind of current pricing for On-Grid Robotic Pick, it is something like approximately about 1/3 of the human cost in most of our markets, which means it's cost-effective down to a market that would have labor costs at a 1/3 of the U.K.'s today and where you expect that labor to only rise at CPI.
Whereas in most markets that we when we talk to anybody where labor is today a 1/3, it's actually rising much faster than CPI. Therefore, OGRP will have an effect there. We're also working, and one of the reasons it's OG On-Grid Robotic Pick rather than robotic pick was to try and actually make the overall CapEx lighter with it than without it because we were trying to take out the pick stations and the pick tunnels, right? So the aim is to try and reduce the overall CapEx where you put automated frame loading in. The frame loading works more consistently. The ultimate aim is to take out some of the conveyor that feeds it because you need less buffer between the thing going on upstairs and the thing going on downstairs because the robots are more consistent and reliable than humans.
We're constantly working to strip that out so that we can pass on most of the savings to the clients. It's not a question of, "Hey, look, we can do something without a human, but you've got to pay us the equivalent of five humans to do it."
Got it. Thank you. Then last one, just as.
I think that would be number four.
I had 1A , 1 B. Just do a quick one.
So we're taking longer debating it. I'll answer it quickly. Yes or no. Go.
Thanks, Ron. So just a very quick one then, just to benchmark. So you normally disclose in your bond docs, which we may or may not see at some point, but just the percentage of labor in your warehouses. I think you referenced 2% as the kind of the normalized level.
Then also the percentage of labor in your delivery as well, just so we get a sense of your labor cost to serve, please.
I mean, the labor is significantly larger in the delivery than it is. I think I mentioned before that delivery is usually around 11% of sales, and the majority of that is labor. In the warehouses, the labor I said before that at 350, you'd expect it to be around 2%, including management. We're currently hitting 230s, 240s. So you can estimate for yourself what that means, mathematically sitting here in the it's going to sound like it's about 3%, in those warehouses.
Great. Thank you. That was a quick one.
Great. Thank you. Thank you for all the great questions. Unfortunately, we are beyond 11:00 AM at the point this point, so we're going to have to call it there.
But obviously, we will be catching up with all of you in the coming days. Thank you so much for all the engagement. And for those online, we will be getting to your questions. Sorry we haven't had the time today, but as usual, we'll be we'll be getting back to you, very shortly. So thanks a lot.
Thank you.