What should a supply chain data transformation roadmap look like?

We previously discussed some?small steps that supply chains can take to get their data transformation journey started?but a major challenge is knowing which way you’re heading and how to navigate the terrain over the medium and longer term.

This session particularly focused on how to balance a long term strategic view that aims to be inclusive and exhaustive with the more short term and tactical needs to address current pressures as they arise with Bryan Oak from Kompozable as our guest subject matter expert.

The discussion also touched upon:

• gaining?clarity on the outcomes?you want to achieve and the general direction of travel for the business,?without getting bogged down in analysis paralysis;
• knowing where you are today by?understanding your current data architecture, sources and quality;
• the pros, cons and costs of data lakes and the risks of them becoming data swamps;
• taking a?change management approach?to avoid bloated solutions;
• maintaining ‘metric discipline’ because even if they are correctly calculated, they can still drive the wrong conversations and decisions;
• how the?‘democratisation of data’ is a double edged sword?and often leads to a proliferation of applications and interpretations that actually create more confusion;
• how to know which interventions and over-rides (human or AI-driven) are value-adding and the risks posed by ‘arrogant models’ and staff who are 'well intentioned';
• that any data transformation roadmap should?give equal weight to training, education and understanding?of master data, parameters and governance;
• whether the current economic climate is?reducing commitment to transformation initiatives.

Key recommendations included:

1. Understand the strategic view, purpose and desired outcomes over short, medium, long term horizons – have your “North Star” in mind, and the intermediate stops along the way, but don’t neglect the immediate priorities;
2. Prioritise!! Segment and chunk up the challenges, use cases and choice of tools and technologies – it’s too much to tackle in one go;
3. Dumping all your raw data into a “data lake” is probably expensive and risky. When it comes to solutions, technologies and tools, “one size will not fit all”;
4. Identify the data needed to support/inform decisions and metrics – choose a scope, don’t be try to be too exhaustive/all encompassing;
5. Have a shared understanding of the underlying data model for your organisation and how it supports your operating model – current and future;
6. Build solid foundations:

• Start with the data you have – what is it, where is it, what state is it in?;
• This means at a logical level you need to understand your data architecture and which applications house which pieces of data;?
• Standardise data sets – definitions, metrics, etc.;

7. Experiment, learn and fix root causes:

• When processes or models are not giving expected results or people are changing or challenging the output, perform root cause analysis and correct the source of the problem;
• Evolve and mature your data management capabilities as you go, to reinforce and sustain the improvements you make along the way;
• Model behaviour, educate and develop ownership from the top;
• Measure improvement in Data Management as well as Supply Chain improvement/outcomes.

Further reading:

https://www.forbes.com/sites/stevebanker/2023/04/12/micron-is-an-exemplar-of-what-a-supply-chain-transformation-should-look-like/

Transcript (edited & anonymised)

…

JP

We did a session previously on this, which was more about what are some of the easy steps that you can take, to get started on the data transformation journey. We didn't manage to cover the last part, which was really mapping out that whole data transformation journey. I suggested to Bryan that maybe he could outline some key thoughts and key ideas to catalyze your questions and comments. So, without further ado, I'm going to hand over to you, Bryan.

Bryan

Okay, well, thanks a lot.?I know I've met a few people on the call before, either online or at the London event. The focus of this is about building a roadmap for supply chain improvement using data. Looking at people's input into this, I think there's always this quandary between?how do you balance a long term strategic view that's trying to be all inclusive and exhaustive against something which is more short term and tactical and addresses the current pressures and issues that you're facing today.?

I don't want to lead or shape the thinking too much because it'd be very interesting to get your experiences and your inputs about this but from my perspective, what I see is that there's a little bit of a dichotomy here in terms of this long term versus short term. There are lots of challenges and opportunities to address across the end to end supply chain, whether that is demand forecasting, inventory optimization, supply chain visibility, supplier performance and whatever customer profitability, product profitability, whatever those activities are, there are a lot of.?

What's important to you and the priorities for your organization is going to be different and there's very definitely a need to position all of this within the context of the urgency, value and difficulty of addressing that particular challenge.

I think the other bit is that there are an awful lot of solutions available to us. I'm a firm believer that?there aren’t right or wrong solutions?because there are so many different ways to skin a cat, if that translates. There are?good and bad choices?that you make along the way.?

The important thing is to have made good choices. Probably one of the main ones is that in this world of supply chain and data, not all solutions are actually applications, data or technology. A lot of them are organizational behaviors, practices and culture and the rest of it.?

If I boil it down, I do think there are some fundamental things though. At the roundtable discussions that we had at the London event, we absolutely had a couple of sessions where it was positioned that actually?knowing what outcomes you want to get and what problems you are trying to solve was absolutely number one. If you don't know the outcomes you're trying to get, then that's the first place that you need to start.?

I think the other bit is around that long term bit,?taking a long term view and trying to make sure that what we're doing is congruent with the business strategy, the operating model that you have today and any changes to that operating model as you move forward. That can be a challenge as well because actually quite often the business strategy isn't necessarily the clearest thing in the world, right? That's where it becomes important in terms of that long term view to be looking along the path as far as you can see and understand the general trajectory that you're going on. If you understand what outcomes you want and you understand the direction of travel, then that means you can then identify what are the decisions we have to make, what are the insights that we require and what's the data that's needed to support that.?

I think the next bit after having that direction of travel in mind is you don't have a choice, you can't not start from where you currently are. That means that?you need to understand the data model that underpins your operating model and your business today. It does mean that you need to have a good view of your data architecture, what data have you got, where is it, what state is it in, etc.??

The other bit, which is not very helpful, is that in taking that long term strategic view you?are trying to be exhaustive. If you get stuck in that mindset too much then you won't know everything that you want to do from the outset. If you spend too long trying to pontificate about that end state, you're going to get stuck in that analysis paralysis phase.?

So it’s very much about building for change, making sure that what you're putting in place can evolve and importantly that you learn that you operate in a way that builds in the underlying fabric of how you then keep that going.?

How do you mitigate the risks then? How do you prevent yourself from going down cul de sacs? How do you protect the return on investment and the costs??

How do you design in a more modular way, if you like, a more adaptable way to deliver the performance that you need? What does a supply chain data roadmap look like? It'll have some common characteristics but it will be different for each organization. With that in mind, do we want to open the floor up for some comment section??

JP

Thank you. I will open it up but I would actually like to ask A to comment actually if that's okay, because you highlighted the challenge that you have with your data lake and it's a common challenge, which is why I wanted to pick on you. There's a perfect way of doing it, which is strategically and in a logical order but you also have to deal with the internal stakeholders who want maybe everything at once. I just wanted to ask you to expand a bit on that and reflect on Bryan's thoughts around that dilemma.?

A

On one hand, we had this idea of supply chain transformation and digital transformation going on for a while. That meant mainly different programs looking at different types of platforms, usually to serve the needs of a particular function more or less in parallel and more from the IT side.?The whole concept of a data lake started to started to pop up in different places. We start to see the first examples of, okay, probably we need X or Y report. Let's bring that data into a data lake with that particular purpose. As we continue through the journey, we started to see that these initiatives were more and more frequent. We said that if we want to have probably a more structured view of how we do things from a supply chain perspective, let's think more about it. I think what I said then, JP, was that while we still drive the construction of that data lake and all that cloud based data elements on specific outcomes of specific solutions that are needed or specific use cases, we also reflect on, okay, if I've done this use case for logistics and then there are these other use cases, when is that addressed?

It comes down to a point in which I say actually I need to change my view. Now rather than each use case, I have a look at all the stuff that I have there, which are the big pieces that are missing for me to have, let's say the whole set of logistics elements there, to continue to exploit the logistics information, for example. When exactly do we need to change the lens and change the perspective to see from a use case from a completeness perspective. I think that's what we are digesting right now.

From a supply chain view to an IT space, it has helped a lot in order to pose those questions. Do we have answer? Definitely no. I think we have a couple of areas where we more and more feel that we are closer to being complete. In that sense, the next question then is once we know that something is out there, how is that we promote the use of what we have? Again, that's where we are and the question that is going on in our heads. Does that make sense??

JP

It does, yeah. Thank you A. Now, I'll invite anybody to jump in either just please speak up or raise your hand, particularly if you've had positive experiences of balancing short term versus long term or scaling up the use cases beyond their initial scope as A was just outlining.?

Bryan

Let me jump in there. I think the first thing is that the really vast majority, and you can correct me if you're approaching it differently, but the vast majority of the data that we use to inform decisions, make decisions, analyze, try and diagnose problems, try and predict, try and whatever is structured data. It's structured data, it comes out of systems and it generally is coming from some structured database. It's got some way of relating it together. If you then jump immediately to well, what's the right data storage type platform that you put that data in??Data lakes are very good when you've got semi-structured and unstructured data. They’re not the best tool really if what you're doing primarily is looking at structured data.?

So I think there are other approaches and this is where I say it's not the right or wrong answer. I think?there is a danger that if you take all your supply chain data from all your different systems and dump it into a data lake that you don't get a data lake…it's a data swamp. I think the other bit is the potential to add value, to generate value if you do it without any concept of the use cases that you're trying to deliver and that's not the case from what A just said, it can become slow and expensive and risky.?

You can’t ignore any data that you store and particularly when you store data in more than one place, it's cost.?If you make that decision to take all your data and put it into one place, then it's going to cost you. In order to make good use of it, you're going to at some point be moving that data or transforming it and loading it into somewhere else where it stores data. The cost of storage of data starts to quickly get out of control if you're not careful.?

I believe in an iterative use case approach to things where maybe what you do is broaden out the horizons but I think that use case driven approach is a better middle way of doing things. But you also have to make sure that you're avoiding creating very tactical solutions that drive you into cul de sac or select tools that are not appropriate for the range of use cases that you're going to be tackling.

C

I'll jump in on a point there. I look after the supply chain product and data teams. The data teams are data science, data quality and reporting. We're going through a supply chain transformation at the moment. We had a number of disparate sources, eleven different databases where were taking our data from. We’ve moved them on to one source now, but?when we first did our call for metrics, we had 30 availability metrics alone come back. So we found the tail was wagging the dog and so took far more of a change management approach as opposed to a data transformation or data platform approach where we asked users to justify how each metric was going to be used and at what decision point. The ones they could justify made it through, the ones they couldn't have gone.?

That, both from a tech point of view,?helped us keep the number of metrics that went into the system and so the size of the database down. It also really drove people to say, this is actually useful, I use this every morning to drive X decision, or this is just a nice to know. Some of those metrics are in our self serve tools, but they'll be used once in a blue moon. That was really good just switching it from being a data program to a change program as an approach was really powerful.

Bryan

When you said that you had multiple availability metrics or whatever it was, did you then effectively get everyone to standardize one way of calculating those metrics? Some industry standard or something?

C

It wasn't so much that…there were completely valid metrics. We had a start of day, midday and end of day availability metric. You'd end up in conversations where people are saying, oh, my end of day is awful. Why does that matter? We should be looking at your middle of day metric. Both?metrics were completely correctly calculated, but just drove the wrong conversation, the wrong decision. We've still got supplementary ones, but we've got our poor availability metric that we use now that everyone knows this is the thing to focus on.?

JP

M, you'd like to come in??

M

Yes, on that point because it's something that I'm currently in the middle of. It's not a full end to end data part, but we kicked off a thing around stock and availability reporting. It's something that C said that may be an approach we dig a bit further in our business that, believe it or not, even finished good stock?when we're trying to get it all into one system for one set of reporting, we're getting four or five different versions of the truth for the same answer. What an unrestricted stock figure was, you wouldn't believe how many variants of that were in across all the systems.?People in different functions, wanting to understand stock positions and accessing things themselves, would get different answers. It's been creating lots of emotion, shall I say. If commercial people are looking through business intelligence suites and looking for stock information, for instance, they believe all in the garden is rosy and then we have service problems, but what they're looking at is not what the real version is.?

We found that, like C said, lots of different bits in the background and different people doing different things. We're now in that process of discussing with, obviously IT systems, how we filter that same type of measure down to have one version of the truth. That, to me, is where a base point for us now needs to be. You have one figure and that's what everybody, no matter how they pull it from wherever, that is, the one source of truth at the time it is extracted and that's where that sort of consolidated. One integrated data source that everyone uses for reporting becomes useful.?

If we can get to a place where there's one number one set of the truth, that's where the endpoint needs to be. We're now in that journey of how the heck do we get there? It's good to see that the C’s of the world have gone through and seen that. That's been quite useful to hear and quite refreshing.?

JP

I’d like to ask P from XXX to come in because in the pre survey input, P, you mentioned that you have some current initiatives around integrating your internal data, I think particularly with the SAP and other tools that you have. Would you mind telling us a bit more about that, please??

P

I also resonated with what M said about one source of truth or one true source of data, that everyone agrees upon one issue that we're having at XXX. Because we've been using SAP as a master system, really, for the last ten years, since we started up, and now we're starting to tack on additional applications to it, respectively. We're starting to purchase additional applications where we’re setting up EDIs with SAP data flowing both ways.

Some of those applications have been initiated unilaterally by certain functions with purposes that have not really been agreed upon by everyone, with inputs that are not really clear to everyone. Bryan earlier used this term data swamp, right? You start to have this situation where while you're integrating things, technically, you're not really integrating things from an organizational or from a quality perspective. You start to get more data in the system. You start to get more data points, more data volume, but the quality of that data is not clearly understood or clearly controlled. The sources of that data are not clearly understood or controlled. From my perspective, my role is supply chain readiness so what I focus on is preparing the supply chain for changes, making sure that when changes happen, that we're ready. From my perspective, it can often lead to misleading situations where people think, based on data that they're seeing, that, no, we're good here, everything's fine.?

Actually we're not, because either the data is not clearly understood or we're looking at the wrong piece of data. Or there's even another set of data that no one is aware of. When I talk about integration, that's really one of the key risks or also opportunities that I see at our stage of the enterprise, which is still very young and very much in the growth stage.?

JP

Thanks very much for that, P.?

L

Maybe some reflections? What P outlined in terms of people potentially say, oh, we are good, but actually we are not good. This is also, in my view, driven by the interpretation of a customer demand, because on one hand, the demand is flat. This doesn't mean that we sell it because we have storage agreements with the customers. We may need to produce it for a particular period, but it doesn't mean that we sell it. All of a sudden the surprise is big towards the end of the month, why we have not generated the revenue following the customer demand. I'm looking after supply chain transformation activities with our main customers and they are driving the agenda.

Coming back to the data, we have one leading ERP system where this is our truth of source or truth of data, right? This is where we have then connected SAP APO as our planning and scheduling tool. It's also connected to SAP IBP (integrated business planning).?To get people from their way of looking at data and also working with data which is purely in Excel outside the system, convincing them and highlighting the benefits of an integrated way of working?means do it right from the very beginning in your leading system in order to avoid changes and data adjustment, which may take you hours to really have the data integrated. This is also then the connectivity which we look after with our customers, which is in two ways from an order processing point of view.

The tactical order management, adjusting volumes and dates towards also exchanging information. What's my capacity versus demand? Am I right on time in terms of my supply towards your demand? Or am I late? This is the beauty going forward, or where we are actually currently in as part of the transformation to have data integrated and to change the way how we look into data, how we communicate toward our customer, as well as how we will get more agile and flexible going forward.?Still, there are stakeholders within the organization who still come back and say, okay, I'm using my Excel, I don't trust the data. Here it's repetitive trying to convince people changing the way, how they process data, how they look into data. And that's really time consuming. Just as a reflection and comment.?

JP

Thank you, L.?

JP

D, it looks like you had your hand up briefly?

D

Yeah, I did. It's interesting that we all have the same problems, but what I see here is that I work in a pharma company.?The main thing that digitalization is doing is democratization of data. Making data very broadly available to many people. In fact, it makes the data much more powerful. If you have an inventory number or service level number or whatever numbers you're using to make decisions, it used to be that specialists looked at that. Now, I have thousands of people looking at that, not necessarily understanding it very well. Potentially, if it's not the right number, then it has far more consequences. So, in fact,?digitalization is making data much more dangerous, but also much more powerful, of course. I think that the key work that we really are going to have to do.?

For me, the challenge of digitalization is not to invent new tools, right? Because people are using new tools every day. The challenge is to really have the right data definitions and the right KPIs to make sure that all that data that gets used by hundreds of people is used for the right purpose, in the right way, so that there's no misunderstanding about what the data means or about what it should be used for. I think that's where really we need to do a lot of work. We have a lot of work to do, certainly in my company, because in fact our standards are not clear enough to really be able to digitalize.?

To have an EDB and all your tools pulling from an EDB from a data lake requires super strong global processes and global definitions that everybody uses the same. And I don't have that. In fact, we're using a data lake to try to create that so we have data coming in. I don't have integrated ERPs. My company has two ERPs. We've got some SAP standards and effectively JD Edwards standards, and we have to pull that together and try to harmonize it. So it's a bit painful. For me, bottom line, it's really important to understand how the data is used by the various tools in the organization, by so many people. Now to make sure that we have super clear and super simple definitions, because the average user in the company does not have the same level of understanding of what the data means as specialists in the teams right in the supply chain team.?

A planner might know very well what critical path lead time means. The average person on the shop floor on a commercial block has no idea. We have to standardize, simplify, and align on a global basis to be effective. It's going to be a lot of work.?

JP

It's related but different C when you mentioned how to measure the effectiveness of interventions on machine generated forecast. I guess that's another variable to throw into the mix with all of this stuff. Would you mind just telling us a bit more about how you're finding the challenges with the machine generated elements within your data as well??

C

Yeah, so one of the things that we have been interested in for a long time is?where the users' interventions add value and where they are value destroyers. So we've got demand forecasting tools. The same with your inventory optimization tools. You've got a computer that either picks a forecast or picks what it thinks is the optimal inventory and then uses, for various reasons, good and bad interventions over the top. What we want to do increasingly in a way is to be able to say in these instances when you intervene, it typically adds value, so it improves the forecast or it moves you to a more commercially optimal inventory position. In these instances when you intervene, it has no effect or it has a negative effect.?

That might mean we should always intervene on this category because its performance is lower. Or you seem to be intervening on a Tuesday often…why? Because the forecast is always pretty good there. Or why have you moved inventory in this product? Because it's putting more stock, but it's not actually driving more sales and you're driving more waste out of it. So, yeah, that's a big area of focus at the moment to be able to coach and drive productivity with users. We start doing some stuff with analysis to try and pinpoint those areas. I think it's a growing area, and if anyone else has got any other reflections on that, I'd be interested to hear.?

JP

P’s virtual hand has shot up, and so has T’s, so it looks like it might do. So P first.?

P

Yeah, I rather had a question on that because I'm not too familiar with machine learning. We've done a few RFIs and RFPs on the topic with various vendors, but?I would have thought that a good machine learning tool would actually do that by itself and start to learn that in this case I keep getting overridden and it's actually an override that is leading to worse results. I'm going to start either ignoring the override or warning that this override is not a good override. That not the case with solutions on the market today??

C

From my experience, not really. There are things that we're looking at, we've got a few initial projects looking at exactly that, using machine learning to identify what’s a good intervention and what bad one is. In terms of market solutions limited in that space.?I'd say one of the interesting things I've noticed even with tools that use more simple time series forecasting or ones that use more sophisticated machine learning is what we call quite arrogant models. They don't understand their own uncertainty. As you get further out, because they think their forecast is bang on, they then order less because they're going to waste. As you get closer and closer, you see your supply orders three days out, two days out, suddenly ramp up and that's a surefire way to irritate suppliers.?

I think we're probably in the foothills of what ML can do from a forecasting point of view, but just scratching the surface from the interventions.?

T

Yeah, very interesting point. We have been discussing that very topic around intervention of forecasting and one of our banners is actually doing that. In France they are taking our SAP forecast, calculating a forecast error on it, but also then calculating their own forecast separately in an access model. I would say it's not machine learning, but it's a different way of forecasting. They're then manually importing that forecast in the forecast and replenishment tool. They say that they're probably increasing or improving on the buyers by about maybe 5% to 10%. That's fantastic. Store level forecasts are relatively low. We're talking five sales per week is a really fast mover so you're starting to change 5% or 10% on to a really low demand signal.?We're not keeping in mind the demand signal on this lower level that they're changing it in stores…so what is the impact of pack sizes??

If a replenishment happens and a pack size is ten pieces anyway, then changing your forecast from one to two, I'm not sure whether that has any added value for us. Definitely for those articles that have got a really high MOQ and a really high pack size to get that replenished, is there that much point in putting in so much effort to overwrite the forecast whereas other factors play a higher influence?

C

Yeah, it's really interesting you say that from a non perishable perspective because?even within perishables we're having a similar discussion of forecast matters for sushi, but does it really matter for dairy?even? Some of our butters have got 15 nights life. You can cover a multitude of mistakes of your forecast over 15 nights. We're trying to focus efforts on these areas that have an impact. Some of our ambient products may not matter at all. Inventory to cover uncertainty and that's it. I think in some instances you get rid of the forecast entirely and probably have an optimal operation.?

Bryan

I think one of the things is that when you talk about using data within our planning process, the forecasting process for the inventory management process, the rest of it, then we have to recognize?it all starts with the master data that you've got and the parameters that you set up within those systems.?

Certainly the downstream bit of it, the demand forecast, is based on sales and output but then?everything in between, everything between is utilizing data that's master data. And I think that comes back to that desire of people to tweak the outputs because they don't agree with what the machine is telling them. Question then is, what are you doing about the root causes??Go back and try and understand not just when it is adding value or not adding value but, more importantly, why is that error occurring??Where do we need to fix the input data, where do we need to improve that? Or is it we need to educate people??

I think you've all come up with examples where actually, one of the challenges, and particularly?when you get to democratization of data, D, as you said, it's about what level of education do people need to understand how the data affects the supply chain and how to utilize it in the best way, as well as the standardization of the measures.?

I think even for something as basic as MRP, the number of organizations that still can't run a proper MRP process because they haven't paid attention to the parameters that they set in the master data and how that calculates works so that they understand what the input of safety, stock levels, minimum order quantities…whatever parameters your ERP system gives you.?

That it leads me to one other statement that we can be addressing particular problems and we can look at the data. We can use different technologies and solutions to help us get the results quicker and the rest of it. Fundamentally though, there's three things. There's the quality of the data, quality of understanding of the people, and our ability to manage data.?

As we're going through all these processes of doing analysis, improving process business, we've also got to be evolving our data management capabilities all the way through. So that's master data, that's parameters, it's data governance, it's ownership, it's responsibility, it's leadership, it's all those.?

Capability and organizational bits can't be left behind whilst you are doing better analysis, improving the business processes, investing in technologies, et cetera, all those types of things as well. So there's a massive organizational component to this which is made more complicated in larger organizations that are certainly larger multinational organizations as well, just because there are more stakeholders.?

JP

Any final comments, questions, reflections before we start to wrap up?

T

Just one provocation. We're going through quite challenging times in terms of supply chain transformation.?How far ahead or how big is it on people's roadmap now with the current budgetary constraints, with the current economic climate, maybe downturn in demand or maybe increase in stock levels, are people noticing that it's becoming more and more important or are they thinking, well, it's quite expensive to invest in a very expensive tool so let's not do that??

M

Despite the current challenges, as of in a couple of weeks, it's almost nailed on. We should be signing off a full move to SAP S4 across the entire business. We've been so fragmented and we've been talking about it for five, six years. We've even brought a Chief Transformation officer into our business. The CEO's right hand man, and he's developing a team, putting people into process owner roles and waiting for board sign off. It's a massive end to end business transformation, not just supply chain and data. I don't know if it's just we've been backed into a corner or we've got to a point where cost is irrelevant. If we carry on, in a year's time, two years' time, it will be worse.

JP

Thanks, M. I mean, it's interesting because certainly since the pandemic, the general sense, at least my general sense, is that it highlighted the need for supply chain transformation. There's this rock and a hard place of the economic situation which makes it harder to justify the investment. Has anyone seen the opposite where it's been pushed back down the agenda??

Bryan

There are a lot of businesses, a lot of SAP customers in particular, with the pressure of the end of life support for ECC, APO etc in 2027. Knowing that?for a lot of big organizations, it's a multi year journey that adds an extra additional challenge to that long term versus short term debate discussion, because that's going to take a while to deliver it. What do you do about your tactical pressures? Because now you've compounded it because now you’re taking all the mind share, all the budget, all the IT resource that you need to do the tactical things as well. I think they've still got to try and navigate that long trajectory with the short term activity.?

I think the other bit, from a supply chain perspective, there's always that golden triangle there, which is, what's the lead time, what's the cost, what's the service level that we can give people??

You said about investing in expensive tools and technology, I think this is where the experimentation bit of it comes in. There are some very good tools and technologies that can help which don't cost millions. In some cases, if you're a relatively low level of maturity, they might be good enough to see you through for the next period.?

JP

I think that's a good time to park it there. We've almost come full circle coming back to this long term versus short term point that we started with. Let me say thank you to all of you for giving us the last 57 minutes, particularly Bryan, for leading us off and giving us some structure in the discussion. If you have any thoughts, any other particular points that you find useful to discuss in a similar way, do drop me a line, let me know.?