JP

First of all, thanks to all of you who were able to input into the session. It's really helpful to make sure that we make the most of the 60 minutes that we've got together. Nick, Phil and I got our heads together on Monday, I think it was. Looking at the input, the different bullet points we had in mind, they all were fairly equal, but the number one, by a bit of a margin, was quantifying the impact on inventory and working capital of demand sensing and integrating it into tactical planning. So I think that's going to be our starting point. 

What we thought would be the best way of approaching it was to ask Nick in particular to talk through his journey because he started with this during his time at Shell. We hope that through that we're going to address a lot of the questions that you guys have submitted. We'll open up into discussion and of course, you can ping any questions to Nick, Phil or each other as you wish. If that's okay with everybody, I'll hand over to you, Nick. Please tell us a bit about your background and tell us your story.

Nick

Good morning. I've known JP for probably the better part of the last six or seven years and joined many of the face to face sessions and discussion groups over the last few years. I'm head of Customer Success at Orchestr8 where I've been for just over a year now. Prior to that, I was the Head of Supply Chain Planning Strategy for Shell Lubricants and so was responsible for the effectiveness of the process, the organisation that we wrap around that and some of the system strategy for that as well. 

I'll return back to that in a few moments. Prior to that, I spent a good decade with GlaxoSmithKline at the very beginning of my career, which was way too long ago now. I started out at Robin McBride in North Manchester.

In my capacity as Head of Customer Success with Orchestr8, I get to talk to a lot of businesses. Some of the things that came through in the input survey are very typical of many industries that we speak to. So what we thought would be useful to sort of prime some thinking around how to bring the topic together was to basically give you a quick run through two of the significant initiatives that I ran with Shell Lubricants, which provided a very significant ROI and transformational result. 

Bit of background: I was there from 2006 through to 2019 so a good period to see industry trends, performance challenges and so on. These things will probably be very common across the group here in terms of performance pressures. There were intense pressures year on year with customer service level, working capital inventory pressures, operating cost pressures It felt a little bit like ‘whack a mole’ most of the time in that you'd have a project, you would go and fix one of those, the clock would spin around into the next season of projects so you'd go and fix one of the other things and then you'd go around and you come back and rinse and repeat. On that cycle, fairly often, forecasting was predominantly on SAP APO,  an ERP-based planning platform. We'd done that globally so Shell had rolled out a single platform across the entire globe. Very large business, huge numbers of bottles, drums, casks of lubricant oil go out around the world…5 billion litres a year. It's really quite a size.

Forecasting was just getting harder and harder. Over that ten years, the product tails were getting longer, customer facing activities were getting more and more volatile. Keeping staff trained in how to do statistical forecasting was getting harder and harder. On the supply side, supply plans, the replenishment plans, whether they be the distribution plans or manufacturing plans were in a constant state of flux. There was constant disruption to supply plans, partly due to the fact that the forecast was so error prone. That meant that there was an awful lot of expediting and firefighting and it didn't really matter what I did in the project space…that didn't change. Some of it got worse the more we pushed some of the project buttons like squeezing inventory down, which then put pressure on service level that had a proportional impact on things like firefighting and expediting. So we typically saw the day in the life of the planner, whether that's a planner in the UK, Malaysia or the US, was getting harder. 

As we look further down the supply chain, some of our key suppliers were reporting in increasing numbers that us, as a customer, we were getting worse and worse to deal with because of the impact of the forecast driven supply chain and the bullwhip, which I'm sure you're all familiar with. Our signals to suppliers were being disrupted, were full of expedites, were full of hassle and firefighting on top of that. One of the things that I was also responsible for was the organisational design. Like most corporations and, I think, one of the comments in the input alluded to this and Shell was by no means an exception, there was a constant, constant pressure on headcount. So over that ten years, I probably had three bites at the cherry to reduce the organisational headcount internally within the organisation which was essentially increasing a lot of pressure on planning performance, on supply chain performance throughout that period. I had less people to deal with an increasing amount of disruption and instability within planning. 

Despite numerous projects year on year, we didn't really shift the needle on that. As I mentioned at the top, we tend to find this repeated in many of the companies that we end up speaking to. So I did two things: first of all, in the world that I spent most of my first twenty years of supply chain in, the forecast was king. Very much in the Ollie [Oliver] Wight mode of, get S&OP and get forecasting right, do consensus forecasting, get IBP, and assuming that fixing the forecast was going to be the main ‘go to’ place. 

So I did a project using the first iteration of a technology you may have heard of called demand sensing. This was invented by a chap called Rob Byrne, who founded a company called Terra Technology. I came across this back in 2011. I tested it in the business to say, okay, I don't know how demand sensing works, I don't know if it would work in the oil industry or in our oil industry…we're very much like a batch blending type industry: we have tanks, we have liquids in tanks, we blend them, we pig and pipe them to filling machines, we put them in bottles, we put them in different forms of container and we ship them out on pallets and products. 

Again, that's probably familiar to a lot of you, even if you're not familiar with the oil industry per se. So I didn't know if it would work in that environment with customers spanning retail, distributors, commercial customers, motor industries…so a real broad range of customers. So we tested it. I literally put a huge amount of data, which I could get from the single SAP system globally and I tested it across a number of different countries because different countries operate differently and I could see that it would make a difference. 

In that first iteration here of being still in a very much classical, SAP APO, forecast driven world, I could see that I could improve the forecast accuracy. That gave me some relief on things like safety stock calculations, on some of the changes in forecast that I could make in the short term. I literally couldn't throw any more people at forecasting, so I threw technology at forecasting to make an improvement. I typically saw 10% to 15% of accuracy improvement over the best of my APO demand planning forecasts. With that leveraged as much as I could from 2011 through to 2014-15, in Shell numbers that was over $100 million in benefits that we recorded. 

Because Shell has rather large numbers of inventory, we had a couple of billion dollars of inventory globally at that point. I implemented it globally so, from a project perspective it was a very strong return on that: more than ten times ROI on what it took me to do it so it was worth doing in the context of still planning in an ERP system. Roll forward three or four years to 2014-15, that constant pressure on the corporate metrics of working capital, days of inventory, opex, customer service etc. was just relentless. So whatever successes I'd had in the previous period had just been now assumed as normal. So by 2014-15, I still needed to return in the following year another 10% or 15% improvement on those metrics, particularly the working capital and the opex metrics.

The planning community that I had - I worked with the regional heads of planning for the main regions of China, Southeast Asia, Europe and North America - everybody was scratching their heads as to what else could we do? We had run all the projects we could possibly think of to improve customer service OTIF, we'd done as much as we could do on the forecasting area, but we still had very volatile supply plans, very volatile effects going through the end to end supply chain at that point. In 2015, I came across a few publications which described a different type of MRP than the one that I was using, the only one that I thought there was called demand driven MRP. So if you're familiar with push versus pull-type supply chains, this one caught my eye.

I didn't know much about it to be quite honest so I did some research and ended up talking to three companies that had been on similar journey: Unilever, Allergan Pharmaceuticals and British Telecom. Three very different industries. I got some positive feedback from those direct examples of customers that had done this before so I embarked on a period of testing. I did an extensive set of tests that validated what would be the impact on my inventory, what would be the impact on my service level? And again, I took a huge amount of data out of the SAP system, put it into the demand driven engine and ran it for three different parts of the business. I ran it for North America because it's huge, I ran it for a grease manufacturing plant in Belgium because it's different and a bit strange and I ran it for Hong Kong because it's heavily networked in Southeast Asia.

I ran the simulation over a twelve-month period or, rather, using twelve months of data. It didn't take twelve months to run. I ran the simulation and compared the results to what had actually happened. The difference was really quite dramatic. That helped me build a business case, really. There was a lot of change management, a lot of stakeholder management, but I basically got Shell, which is a very conservative organisation, to actually step away from thinking that everything you needed to do was in SAP. It was helped at that time because SAP announced the sunsetting of APO. so that was a very helpful intervention ‘own goal’ by SAP at that point. 

We did, as many companies are doing now, we shifted planning away from the very old school legacy type approach within the ERP platform and put a SaaS planning platform on the side, or above it or next to it, whichever way you want to view that. So I implemented Orchestr8 as a demand-driven platform to drive a particular set of benefits beyond those forecast improvements that I'd achieved in the demand sensing piece. Those benefits were very real, very tangible, very measurable. We stopped making stuff we didn't need because we stopped driving production plans and distribution plans based on forecasts. We became demand driven so stripped out a whole layer of forecast error propagation along the supply chain. 

All those points that I mentioned, where were feeling pressure before on the production plants, the instability of the supplier plans, and on the suppliers that supply the components and materials, we managed to produce a much more stable, calm plan because we became demand driven and decoupled the nodes in the supply chain that had a direct impact on reducing the working capital. If you look at your inventory - and I would be amazed if this doesn't resonate strongly with you, most of the time we have too much of the wrong stuff and not enough of the right stuff. We're always chasing the stuff we don't have and sitting on excess stock of the stuff we made but ends up not being required. Quite simply, when you stop driving production and all of the other requirements based on the errors in the forecast, you liberate not only capacity to make the stuff that you do need, you also overall reduce the working capital that's in the warehouse whilst maintaining a higher service level. 

It sounds a bit fantastical because the world that I was in prior to this point, I was doing a project, then I'd go and fix the other one and another one, and I'd go back to the first one because I'd disturbed the impact of it. I was always trading inventory versus service level versus opex. I could never quite get them all to line up. But the eureka moment of this was, by changing the planning method, I unlocked something that I'd never really considered before, which was how to stop the errors in the forecast from propagating all the way down the chain and stopping the bullwhip from ripping down the chain as well. 

Many of you will have done the beer game or the coin game, where you simulate tabletop up and down a chain and show how the bullwhip works in a classroom setting. If you haven't ever measured it in your planning system or your ERP environment, it’s worth having a look at just to see what that ripple effect is along your supply chain…to see how potentially minor perturbations at one end in the customer variability which might not be so great but, by the time you get two or three nodes down the chain, you will absolutely see that bullwhip. 

I implemented that globally and effectively turned off APO and switched on Orchestr8 as a cloud planning system. It had a significant impact on service level, inventory and operating costs but also some of the non financial type things like a ‘Day in the Life of the Planner’: that relentless firefighting, always expediting, always chasing shortages…that decreased significantly simply because were planning on the actual real demand and buffering or setting up the supply chain to handle that in a way that meant that things calmed down significantly. The impact on the suppliers that were sending materials and components to us improved as well.

And, just to wrap up, since then what we've managed to do is combine both of those approaches together. I'll pause there. 

JP

Thank you very much for that, Nick. I, go ahead. 

I

Sorry if it's a dim question, please can you explain what you mean by the difference between demand-driven and forecast driven? 

Nick

To achieve that, we set up the forecast in a decoupled way so that we can account for the variability in the demand. So, not to dwell too long on this because it's a whole education session in itself, demand-driven MRP is a pull system, whereas forecast driven MRP is a push system. You push inventory out in forecast MRP to meet the forecast, whereas, in demand-driven MRP, you use the actual demand signal, not the forecast to pull the inventory through the supply chain node by node. 

Does that answer your question?

I

It answers my first question and generates about three others! Can I just confirm what kind of customer lead times you are working with in this environment? 

Nick

It varied. In Shell, we had 24 hours lead time, 48 hours lead time sometimes. When we set up what we call the decoupling in here, we use a concept called buffers, which are different to safety stock. Safety stock is a bit of a blunt tool in the MRP world. We use decoupling buffers, which people think is like safety stock and, broadly speaking, they are but they help us deal with the variability, particularly around those shorter lead time-type products. We respect all things like lead times, MOQs… all the things that you would very much recognise in your MRP world are very much part of the DDMRP world. We just construct how we replenish differently. 

A

A quick question. I guess you have all this internally produced, right? Kind of sending the signal to your own factories or did you also deal with external triggers or both? 

Nick

The actual engine of replenishment is exactly the same. Whether you're a material planner, a finished goods planner, a third party purchase planner, the actual mechanics of how that replenishment is calculated, how you construct the setup for that is the same. It doesn't really matter whether you are internally manufactured, externally manufactured, or a mixture…it's the same process that you use for both. 

A

And just a follow up question, Nick. You named Orchestr8 as being the system, I guess. Was there anything specific you put in place with the external manufacturers? Did you use Ariba or something like that to share this with them or was it directly connected? 

Nick

Typically the way I did it was there were already connections between the Shell’s SAP environment and external vendors for most of them. Some were still on emails and Excel sheets, but some were electronically connected. The we typically could construct the SaaS planning environment to the ERP backbone - because you still want your ERP to drive the sales orders, the purchase orders, the stock, the STOs and the works orders, et cetera - everything remained the same. If you were connected to the ERP environment to get your comms, then you still maintained that. Within Orchestr8, the solution that I put in, for those that weren't linked to the ERP environment there's a vendor portal, so we could have other vendors dial into the portal to get what they need or see the information they need. It helped to bring more people in that probably hadn't previously had the business case or the incentive to dial into the existing SAP portal. 

X

I'm just wondering if you ever come across such a situation when the client gives you very short lead time, but your supplier, as external or internal supplier, will have an extremely long lead time. When you're in such a situation, how would you manage such a gap in the lead time? 

Nick

Yes, it's a very good question, X. We're actually implementing in a business in North America right now that has exactly that situation. They produce consumer-based hardware products, which are sourced globally from Southeast Asia into North America. Lead times can be anything up to six or nine months on that so, as I mentioned before, we take into consideration the lead times, we take into consideration the demand variability over that time. We take into account things like minimum order quantities of certain products. We also take account of things like optimising the shipments as well, particularly on ocean going freight, in the example I'm using now, filling containers is very important. You don't want to waste any space on containers because the logistic cost is so high these days. Basically, whether you're short lead time or long lead time, the way we construct what we call the decoupling and the replenishment honours those lead times. It tells you how much inventory you need to have at the destination point, what is the frequency of supply you need, and how much do you need on the water at any given time. It allows you to see and control those elements so, if you're on a really short lead time in the selling country, that's part of the design, part of the setup, but it works whether it's short or long. Phil, did you want to add? 

Phil

Phil Ribbins here, good morning. I'm the CEO of Orchestr8. Just a slight addition to that is what we find is because in decoupling and setting up buffers in the supply chain based around those parameters Nick just ran through, what you're doing is you're setting up a capacity in the business to manage variability that you can't actually forecast at the granular level. Actually, in some ways, when you get to a very long lead time, because you're separating out what you're positioning in terms of variability control versus the volume of lead time, those differences get very small because your actual variability, by the time you take, say, a six month lead time and compare each six months to another six months worth of demand, the variability between those six month time periods is actually relatively small. 

You can actually then achieve a very high level of service by positioning the right set up in the supply chain across that lead time without having to cope with the variability. Of course, if you're trying to cope with that variability at a granular level on a week by week or day by day basis by SKU, that's still just as variable as on a short lead time product and that becomes almost impossible to manage with a long lead time. In some ways, that management of the volatility really produces a fantastic outcome in terms of long lead time product. 

Nick

The client that we're implementing at the moment in North America is now starting to see the impact on that because they see a more stable plan at the longer lead times, which they find very welcome. In a traditional MRP world based on forecast, that's a very difficult thing to manage. Whereas, doing it in a demand driven way, they're literally only a month into go live, but they're already starting to see the impact. 

JP

Any other questions? 

H

So, Nick, do you have any context or experience with Orchestr8 in any scale up type of environments where you see that growth is 80% one year, the next year it's 150%. Really, when the demand is probably the most difficult thing to estimate because it comes in shockwaves. Can you speak to maybe a situation like that rather than something as developed as Shell, right, where I would say marginally shifting, not those big shocks? 

Nick

Yeah, definitely. Something which I really struggled with in Shell was actually linking what we're doing at the execution end of the business, the actual planning and execution operations with that strategic end. Where we're maybe forecasting and have a business goal to be 50% bigger, 80% bigger, 150% bigger. There's a piece in the middle which we call the conditioning. And this is kind of like your S&OP process of this. It's the part of the process where we can actually say, okay, if things are growing at this rate, what do I need to set up in the middle in the way that I've constructed my buffers, my lead times, my network design, et cetera, to be able to meet that demand? And when you hit the limits of, for example, oh my God, next year is going to be twice as big as this year, you can instantly start to see that and see where are the breakpoints in your current supply chain setup. So, if you're going to exceed a capacity or you're going to find a new bottleneck somewhere in your design, then that's the part of the process where we actually look for that. There are various bits of analytics that help you do that, but it's effectively linking your strategic direction all the way through to your execution. 

And it happens in reverse as well, where you've got declining sales. There's a couple of businesses that I know of have experienced both ends of that in the last few years. The one that I mentioned that we're implementing at the moment, they had a fantastic pandemic because their products are domestic products. We were all locked in homes and thinking, what can we do? So we start decorating our houses. They saw fantastic sales because people weren't going on holiday so they bought new bathrooms, they bought new kitchens, they did things. Same with paint manufacturers: they went absolutely crazy and couldn't keep up with the scale of sales that they were seeing. Wind the clock forward two years, you're now on the downside of that because there's only so many kitchens, bathrooms, bedrooms, and other things that you can decorate and remodel. So they're having to manage it both up and down on that same strategic curve. 

It operates in the same way: you still need to condition and set up your supply chain. We typically do that monthly as part of an S&OP process to say, where am I going to be in the next few months? What do I need to do? What do I need to handle coming up? You configure your buffers, your setup to do that. That might be for promotion, that might be for tenders that are coming up. That might be for new customer wins coming up. You might be opening a new warehouse. You need to model that. You need to prime that with inventory. You might need to stock build. All of those types of decisions we take care of on a monthly conditioning, S&OP type cycle.

N

I guess the forecast is still important then, but your forecast isn't driving your process. As you were talking before, I was kind of thinking, oh, maybe you don't need a forecast anymore, you work out your availability buffers but your forecast is then massively changing buffers, doesn't it, rather than drive it. 

Nick

Yeah, exactly. I made that mistake in 2015 when I was thinking, oh, great, I don't have to do forecasting. It was one of my first misunderstandings of this shift in concept. You absolutely still need a forecast and you make your forecast as good as you can because that's what's helping you to do that strategic setting that I just described. But you're not so desperate to get your forecast to be so accurate as to what day is it going to sell in the week or the month and so on, because you're setting up something that can handle the variability. You're just less obsessed with the Nth level of granularity of the forecast so it actually takes a bit of pressure off your forecasting process but you still need a good forecast. That helps you set up the buffers and see what sort of volume over time am I looking to be able to support.

N

In terms of the change management bit when you go through that process… I'm thinking it might be more from the supply side of the demand side, because I guess supply side people often like to complain about the forecast and blame it. I think they also like having that forecast there because it maybe kind of takes a bit of pressure off them to a degree. In terms of people saying, actually, that's not how we're working now…with the variability and the buffers that you describe, how do address the kind of change management to get to the buy in that side? 

Nick

Yeah, it's big, because in the world that I was in prior to making these changes - and you'll recognise this - everybody gets very good at second guessing. So we have a plan but we second guess that plan. I saw this when I first implemented. Everybody agrees, lead times MOQs, et cetera. The system says, make a batch. I look at what the factory did and it made five batches. Why? Well, we always make five batches! But the system only wanted one. It's that kind of scenario because everybody sort of suffers from the forecast, but also hides behind the forecast. We hide quite a lot of bad practices in supply chain, which we don't think there's any solution to because of the forecast: people second guessing the plan, people doing things slightly off the plan, not de-expediting stuff they should or ending up expediting something. You'll recognise all of those things. So, yes, there's a big change management there but it's a good thing because when the system says, right, you now need a batch, you genuinely do need that batch, not you might need that batch. And there's a big difference between might need something and actually need something. And that's quite a big shift for different parts of the organisation. 

Whether you're in logistics, distribution side, whether you're in production scheduling, for example, whether you’re in supply planning or material planning, you're going to get the same type of signal which has a far higher degree of certainty that you should act on it than the previous signals where maybe if I act, maybe if I don't…I'll kind of wing it a bit. It's no fault of the planner, because in the error strewn, forecast driven model, you're actually wanting your planners to use a bit of planner intuition and try to second guess. We actually encourage that in the previous model whereas, in this one ,we actually try to detune that instinct and say, no, follow the system because it’s going to be more stable, it's going to have a greater degree of confidence around what we're asking a supplier or a manufacturer to do. 

Yes, it is a big leap of faith. Whenever you buy a planning system - and I've been through this so many times - everybody gets burnt at some point with a technology that you implement and it doesn't really deliver. I've got the T shirt and the mouse mat from so many projects that fall into that space. The thing about this one is that I tested it, so it was actually less of a leap of faith because I did so much due diligence. I probably spent 18 months doing the due diligence because that's just the reserved, risk averse nature of Shell as a company or the culture within the company. I had to get so many proof points that it was going to work and it was going to deliver that, when I actually came to do it at a project level, it was less of a leap of faith at an individual level.

Yes, there's a leap of faith and we really advocate having a proper project with change management and helping people on that. We call it the thoughtware, not the software. You've got to help people think about replenishment and think about inventory in a slightly different way. But people do get there. If I got there, I know other people can get there. 

S

I had the exact same question as N. Do I need a planner now? We run our forecast accuracy, which is a two month lag, we run 95%...we forecast 100, we sold 95…that's kind of the way we measure it. So we are extremely accurate. 

Nick

Yes, very good. 

S

5000 SKUs. I'm looking at this from the perspective that I heard about DDMRP in the past, the French business, they keep talking to me all the time. They want to demo all the time. 

Nick

It's very big in France, hot on this topic. 

S

The only thing I have here is that part of the change management. My only concern that I can envision is, okay, DDMRP kicks in, I have the most accurate forecast that I can. We go to quite a deep level of granularity into this - we go channel planning - all of a sudden my demand signal tells me I have produced everything so how is this linked to capacity? 

Nick

Really good question. In exactly the same way that your current processes are linked to capacity. Today you have a demand plan, you have a supply plan that you put against that demand plan and you test that supply plan. Whether that's a production supply plan or a purchase supply plan, it doesn't really matter. As long as you can measure that against the capacity of your supply, then you manage that. We operate in exactly the same way. So the buffers require a replenishment: we can project forward what the future replenishments will be based on your forecast. We can show you what the effect of that is on your capacity. So all of the same planning concepts that you'd recognise including rough cut capacity planning, detailed work centre capacity planning, aggregate capacity planning… If you're looking at total network planning, you operate all the same considerations there.

But one interesting comment though is that 95% bias…it’s terrific. That's quite unusual. Most companies that we speak to don't achieve anything like that. What's interesting is even if you've got very good forecast levels at the finished goods level at the top of our end to end chain, depending on how deep your bill of material goes, or how deep your bill of distribution goes or both, you still tend to see bullwhip running and generating along the supply chain even if your finished good signal is quite accurate. I don't know if you've measured that or if that's something that your supply chain planners feel? 

S

We see that a lot. Most likely the biggest driver of that are MOQs imbalances between components and finished goods. One thing is filling, one thing is missing, one thing is the raw material itself, the component that is the biggest driver of the bullwhip. We need something in four weeks time, but the lead time is six months. This is for one component, the other one is three months and one is ten thousand and the other is one hundred thousand. So it's that imbalance creating the bull whip because it's firefighting all the time. 

Nick

Exactly. And that's something which in that design that I talked about where we configure where to put the strategic decoupling buffers and how to consider the lead times of different products, that's all part of that conditioning design piece that we do so we don't leave any parts out. And we can factor long lead time components versus short lead time components…what to hold, what not to hold…even things like if you can't get hold of a particular component and you might have a substitute component, we can handle that for things like dual sourcing or alternate bill of material type moments where you are in trouble and you are making a different planning action. It's a very common scenario. J

D

Predominantly this sounds very much like it's a purchase order demand planning tool. Can it also support internal capacities? Storage of goods between retail estates, warehouses, planning locations, RDCs, NDCs, et cetera and even down to a more granular level of ABC analysis on inventory in a warehouse, fast moving, slow moving medium, et cetera and help me plan where the stock is supposed to be? 

Nick

Yes, absolutely. 

Phil

Just to extend for that, one of the reasons the answer is yes is because if you're planning inventory effectively as part of your capacity to manage demand, right? So it’s not a safety stock over which forecast layers on top of it's actually your built in capacity to manage the demand level that your business has. We can then model things like capacities both in terms of throughput and in terms of inventory. We can do capacity planning on physical stock locations as well as production lines as well as suppliers, as well as transport lanes. Any of those different elements to the system, they're just at a point in the supply chain which has demand, it has supply and it has something in the middle, which could be a warehouse, it could be a factory, it could be an assembly site, it could be anything. All of those different elements can all be capacity planned. 

D

So that was my next question. Does it do scenario planning modelling? So, for example, if we want to do a range extension and increase the SKUs by x thousand trying to understand the implications for not only the supply chain externally but internally in warehouses and stores and what that does if I was to give you store footprints and cubic capacity of warehouses, that thing, it would help me drive that kind of granular level it would give you?

Phil

Yes, absolutely. You could see the impact of those different setup choices. I think something else I was going to add to something Nick said earlier: the setup of the supply chain allows you as a business to start to link your inventory positions to those factors of lead time and batch size and demand profile that are driving your requirements. You can actually start to manage more specifically, what do I get or what is driving my need for inventory in a particular place? What would happen, for example, if I did reduce my MOQ by 50% and what does that do to the amount of inventory I therefore need because the inventory is directly linked to those parameters. 

D

Just going back to the whole demand conversation, this is a demand plus forecast position, isn't it? I'm assuming by demand it takes historically our ERP sales and then we apply the regimen of what we think we're going to see a 10, 15, 20% increase next period, next quarter, next year. That's at SKU level or can be? 

Phil

Absolutely. Your setup can be different for different product ranges. You may say for this part of my product portfolio, I'm quite happy to use my historical average demand profiles going forwards. For this part of my portfolio, I want to use 100% of my forecast volumes and just focus on forecast. Or you can blend those things to help set and then as you go through your conditioning steps, that's where you would actually optimize your set up to meet those different demand profiles that you're seeing. You're getting a regular cadence of monthly optimisation of your supply chain set up to meet your demand pattern as it changes over time. 

D

Final question in terms of the setup of an infrastructure like this is cloud based and it plugs into data warehouses, data lakes, that stuff?

Phil

Absolutely.

D

Very good. Thank you. 

JP

Thank you, D. M?

M

Yes, so I just wanted to clarify…essentially we have our forecasting system. We do forecasting as normal, right? This technology or this system sits in a place where it replaces your normal MRP. Your planners would take that forecast and would go in and produce, for example, in three months time: we need so many batches to be coming in and to the deliveries that come in. This system would then take that and replace that with a better demand planning essentially and it will sense that. Am I correct in that?

Nick

If you think of our solution in sort of different components, we've got a demand planning and demand sensing end of the solution, which if you didn't have a forecasting solution, then that's a really effective, AI driven, very modern, very new, effective DPDS tool. That's the one I referenced before that I'd implemented in Shell previously, very effective. We've embedded that within the Orchestr8 platform now. But, essentially, you take that forecast and you put it into the supply planning side of the solution which is basically looking at your inventory, looking at your capacities, looking at your existing plan and giving you a very effective replenishment plan to then communicate with whoever you're getting replenished by. 

That might be internal factories, might be third party suppliers, could be component suppliers. It could be - to D's previous question - it could be doing the distribution requirements between your different depots, it could be using your RDCs and your DCs where you need to distribute inventory across your network at the right time, in the right quantities, in the right places based on the actual demand that those depots are seeing. It helps you to size that network and size those requirements.

M

What data feeds are you setting into it? Is it every customer's ordering pattern at the granular level? For example, if you have six different clusters, ten countries in each cluster and ten countries, each country has got ten different warehouses or ten different distributors that you distribute to…at that granular level that data is being put into the system? 

Nick

Yeah, we typically take it at the SKU location customer level. If you're doing the demand planning, demand sensing, we see that at the customer location level. If you're doing the supply planning piece, then we're very interested in where your suppliers are, where your stocking points are, where is your inventory going to be held so that we can then create the replenishment plans and communicate the orders that are going to be required typically back to your ERP system. It could be to your data lake if you're doing it through that. Typically, we transact backwards and forwards between the ERP backbone in your business. 

M

So this is automating essentially your planning side of the world whereas traditionally in forecasting you've got a lot of automation tools that are currently there, AI based whichever base you want to do it, but that's been existing for quite some time. What this is doing is taking that by using the granular information that you're getting from each customer and using it to get a better plan of what you're going to get rather than a manual exercise of a planner to plan it out.

Phil

It's also a different dimension to the plan mode because what it's doing is putting a layer of planning in between the forecast and just the order. If you think of the way that traditional MRP runs, you run your forecast, the forecast goes straight into MRP and you get orders out without any filter…one drives to the other. What we're doing here is we're putting in a filter if you like, and a tactical plan to enable the supply chain to meet that forecast volume. The actual orders are usually driven by the actual demand that comes in as opposed to being driven by the forecast. The simulation tools and the planning tools allow you to have a full capacity plan for the supply chain. You can run simulations out, you can share the volumes for example with suppliers, with your factories, anybody that needs to be part of that decision making. The actual orders are not directly linked to the forecast.

M

But what information are you using for actual? For example, my customer can go in and order for next week's delivery. They're not telling me in three months out what they want in May this year.

Phil

Yeah, exactly. When you get that order, that would debit your inventory that's available at that point in the supply chain and that may or may not then trigger a resupply. Think of it like in the elements where you've got a replenishment part of the supply chain, it's more like you're setting a maximum level of inventory, which represents a capacity as opposed to a safety stock. The stock in your business will be relative to, as I said, the lead time and the batch size. Your inventory in practice will be split between inventory on hand and inventory on order so you've actually got control then of everything across your lead time. 

M

Okay. 

JP

A, I think we've just got time to get your question in. 

A

Yeah, a really quick one if I understood correctly, Nick: this will completely replace the classical MRP, right? It will just be replaced through the MRP. 

Nick

That's right. I think D mentioned this is all about one software…it's not in effect, it's not about one software, it's about a methodology. One reason that I selected Orchestr8 for Shell is because it was an entire replacement of my existing planning environment…I didn't need to operate two environments. So this is a really important point: not everything, is going to be decoupled and buffered. There's some stuff that you might want to run make to order, for example, because you don't want to hold inventory just because of the demand patterns or the particular variables of that product. So in the tool, it's not a one-size-fits-all…you don't apply the same rule to every product. 

You might have reorder cycles or reorder points around the buffers. You might have make-to-order where you’re trying to hold no inventory. You might have things like rate-based products. For the really fast runners, you want to adopt a planning rule that can really support something that's moving very fast through your business where you don't need to hold so much inventory because your variability is low and your volume is high, for example.

It's an interesting thing to make sure that you can apply the right rules to the right products. And it operates finished goods, bills of distribution, distribution centres, RDCs and it operates through bill of material as well. So things like your production, your components, materials, raw materials, et cetera. So yes, it does replace the MRP. 

JP