Marieke Smits 0:00:02.0:

I would like to introduce our next speaker, Robert Frentzen, Financial Planning and Reporting Manager at Heineken, and he will be talking about how they have become boring predictable to improve their forecasting and planning capabilities.

Robert Frentzen 0:00:19.4:

Yes, so thank you very much for the opportunity to speak to you all. Robert Frentzen, indeed, in my role I am responsible for structural improvement projects in reporting and in planning. I'm very delighted to share our story going forward. It's really about how we improve our capabilities. I was very impressed by the presentation of KPN, 99 per cent forecast security. I think we've got a long way to go. I would like to introduce first a little bit, Heineken. Of course, everybody knows Heineken, but I'm going to go a little bit into what is the context for our journey, really, from the set-up of our company, and why do we want to improve our planning capabilities, and what are we doing, and how are we doing it? Everyone knows probably that Heineken is originally a family company, a family-owned company, it is still a family-owned company, but also listed on the stock exchange with a huge and wide product catalogue with beers, ciders, soft drinks, and what have you.

Robert Frentzen 0:01:33.4:

From that comes also that we're the number one brewer in Europe, we're the second brewer globally. We are active in more than 190 countries. We have 160 breweries, and 240, 250 million hectoliters, 350 brands, and almost 90,000 employees, so quite sizeable. From that comes a lot of complexity as well, which you would expect in planning. We'll go a little bit into our strategy and our purpose; is really we brew true togetherness to inspire a better world. We really see beer and our beverages as a social lubricant to really bring people together, and we do that through our legendary drinks, brands, and experiences. That translates into a strategy that has just been refreshed quite recently. EverGreen is the name of our strategy, and it focuses on three key priorities. The first priority is to accelerate the growth, to really use our advantaged footprint and our brands to accelerate growth in the future. We want to make a step up in our productivity.

Robert Frentzen 0:02:59.4:

That's also where our projects are hopefully contributing to, but for the biggest part, we're driven from the focus on future-fit, so we're really working on future fitting our organization for the future. You see here the Digital Backbone, I'll come to that. This is a very big project in our company, where we as a project also need to deal with, basically. This part is also really about our sustainability and improving our winning organization. If you look at this, you see that we want to transform from really fragmented local cost structures and more leverage the scale. I think KPN was a very nice example where the scale was already very much leveraged. We are still very dispersed as to that, and we want to leverage more our global scale and skill above OpCos. We have very complex and non-standardized legacy systems which we need to transform in DBB, which is our Digital Backbone program, towards more harmonized processes and data.

Robert Frentzen 0:04:11.1:

A little bit about this DBB, because it's so important in the context that we operate in, is that we are going to transform our IT landscape from different ERP sessions all around the world to one, cloud-based platform where [unclear words 0:04:27.7] is in the middle as the digital core, and you have all kinds of platforms sitting around that communicate to that. We are really taking the step now, and the investment as a company, to standardize our processes, to analyze our data, and really to enable the future of AI, of other data platforms and - well, we've seen already before how important that is to drive improvements for the future. I think this is really where we, as a big company, need to catch up. Let me explain it differently. We are organized currently in different regions and OpCos. Going back to why are we trying to improve, what we noticed is in our forecasting we are missing the mark. We observed our forecast equity is low, nowhere near the 99 per cent. We have very limited insights into the forecasting process and we're really slow to adopt and course correct.

Robert Frentzen 0:05:34.7:

We see also that our process is very inefficient and disconnected. It was time-consuming, it was error-prone, a lot of manual work, a lot of Excel, a misaligned project and misaligned processes. What you see here is a series of initiatives that we've taken to start improving this, and I'm really going to focus in this presentation on connected forecasting, which is the project we're doing with Anaplan. I was also asked to share a little bit on our AI developments, which we call HEIPredict. We like in Heineken to Heineken-ize the terminology very much, so that's why we call it HEIPredict. This is a little bit the world we started with a year ago. We had many different operating companies that are really working independently, and some of them already recognized obviously that their forecast accuracy is not there, and that they are struggling, so they started developing their own planning platforms.

Robert Frentzen 0:06:46.7:

Some of them, most of them, did it with Anaplan, and some of them, very successful and very happy. Rob Kleinjan was mentioned before with this beautiful video. He is, I think, one of the biggest advocates of Anaplan in the company and shown really the value also to the rest of the company. Also, some of those were struggling, because everyone was inventing the wheel on their own, and that's something that we observed is not sustainable in the long run. What we're doing is to really build a centralized template to roll that out to all of our companies. This is a little bit what I wanted to explain earlier. We have four regions, we have 81 operating companies, which are organized by country and market. There's local entrepreneurship and empowerment there, and they're all P&L responsible themselves, so that makes it pretty difficult from a central organization to harmonize things.

Robert Frentzen 0:07:58.6:

There's little standardization, and we have this big, huge elephant in the room that's DBB that's coming to all of our OpCos, and on one end, it will help us to improve our data and standardize things, but the change effort there is so big that it's going to be very difficult to also change your planning processes at the same time, so we actually have the strategy to do this next to and in parallel to the DBB. So, what are we doing? We started with we needed to prove actually that the whole concept of a standardized way of planning works in the Heineken concept. In Heineken, if you propose something, they want to see the proof first, so that's where we started our journey with last year, and we basically got the challenge, with Spain and Portugal, geographically very close, but actually, in the way that they operate and how their market looks, they are very different. Different product portfolios, and so on.

Robert Frentzen 0:09:02.9:

They basically challenged us, 'Do a proof of concept with this, guys.' We put them in one room, and see if we put them together, if we can come up with a common planning process and with a common way of putting the data together. That's what we did, and in the end, we were able to do that. I'll show that in the next slide, but our mantra, which is really important here, was to be relevant, simple, and scalable. We want to make sure that we keep it relevant for the individual OpCos, but also really push them to keep the planning as simple as possible, not to go overboard with the details. We've learnt from, for instance in Italy, where we implemented Anaplan, they went overboard with all the details in the planning, and then they got completely stuck. We also wanted to be scalable. I'll come back to the scalable part later.

Robert Frentzen 0:10:03.3:

In the second phase, where we're in the middle of, is that for these OpCos to actually build a foolproof or failure, we call it a full pilot, where we actually build the solution, for Spain and Portugal, to prove that this is technical feasible, that we can consolidate also these OpCos into a regional view. Our vision is that if you bring all of the OpCos together on the same template and the same data structure, that we can also aggregate into a regional level and towards a global level, so that you can do a top-down and bottom-up planning from OpCo, region, and global perspective. In there as well, we've shown with these OpCos that we could do it and start measuring the value, and the regional view was also successful. What were the key learnings from it? For the proof of concept, I think, and we heard it before also from the video from Kleinjan, we are much more standard than we think. If you peel down, peel off the onion and you look at the planning process, in the end, it's just a P times Q usually.

Robert Frentzen 0:11:17.5:

It's not that difficult. You need to make sure that your data dimensions are correct, but for the rest, it's actually much more standard than we think. What we learnt is that if you make smart planning hierarchies, you can both guard the standardization, but we can also allow for the flexibility for the local market circumstances in the OpCos. We learned that you need to really think in the end-to-end processes, and doing this exercise actually really helped the evolved OpCos to rethink how to do their planning. They really started to think yes, we've been always doing it like this, but if we can enable it through the [?tools 0:12:02.1] then we can make our process a lot smarter. That's really exciting, what you see with this project, that the rethinking actually by the teams themselves happens almost by itself, because they see the opportunities.

Robert Frentzen 0:12:18.6:

In the proof of value, we're still in the middle of it, but what we've learnt so far is that indeed this reimagination is really important, but we already know to make the change stick, there really needs to be a very dedicated effort. When you're under time pressure, it's very tempting to go back to that old Excel if it doesn't work straightaway, so that's where we really need to evaluate and make sure that people are well prepared for their next planning stages. Also, when people move on in the organization, that the knowledge is properly transferred. This is a little bit in-depth, and this is as in-depth as I will go, but a little bit explaining what I think was a key enabler to make sure that we can harmonize across these OpCos. This is an example of our customer plan hierarchy that we made, and we actually said until this fourth level that you see over there, until that level, everyone has the same values, basically, in the hierarchy.

Robert Frentzen 0:13:27.2:

On level two and three, they can choose. They can choose from fields, where we know that those fields will be available in the future, but from the DBB. We made a very smart thing here, that actually in our data management system, the OpCos once need to register which field they use at which level, so that it can be also picked up later in reporting and so on. I think this allows so that one OpCo that's organized by region, can put the region there, an OpCo that's organized by sales channel can do it in a different way. That really gives the flexibility while maintaining the standard. What's next? At the moment, trying to measure the value of this proof of value, where measuring the efficiency, we're trying to see if we see time improvement in terms of time spent on forecasting or planning. We're measuring the forecast accuracy as hard items, and through surveys, we try to find out if any of these qualitative measures are also really improving.

Robert Frentzen 0:14:45.9:

We're just live with this OpCo, so the measuring is starting, forecast accuracy we already see. I think also thanks to level of attention that we slowly are improving, and yes, the rest is still to come. This is actually in the end where we strongly believe that the standard template can bring much lower deployment costs, because actually what we're doing in Heineken now is we've built an app just before the apps were available, our own app, and we're going to bring that app to the different OpCos where they can basically put their data in and we can deploy really quickly. That's the idea. Therefore, also centralizing builder licenses, which are now in different OpCos, we can also bring the license costs down, and we can much more learn, share, and reapply across our OpCos, because we can share experiences and together optimize the whole process. That versus a non-standard template where we've seen time to implement is longer, the cost is higher, and less opportunities to share and improve.

Robert Frentzen 0:16:04.9:

This is a little bit our medium-term plan. We are having, for the time being, a Europe-first approach. We want to expand. We're actually at the moment expanding to France, and with Italy, Austria and the Netherlands on the roadmap for the company years. We hope to accelerate after, but first, the proof of value needs to come through, basically, for us to really commit to a bigger roadmap but this is, for now, it. We are actually doing this based on the potential value that we see in the OpCos, but also very much we notice, and I think it's the greatest asset that we have, is that many OpCos want this. They all hear the great stories from OpCos and how it's working. There's really a big pull coming, so I think that's really helping us, and that's, I think, an asset that - keep listening well to the OpCos and make sure that the standard is something that they can really work for, is relevant to them, is really essential for us to keep the momentum in this project.

Robert Frentzen 0:17:19.6:

Then a little bit about our journey for AI machine learning. This picture explains a little bit what our thinking is around AI and machine learning. On one end, you have the human forecasting, people know what's going to happen, whether you're going to do a price change, or whether you're going to do some intervention action or whatever. You have the machine forecasting, which can, based on the past, predict momentum, where you're going to end up. Today you see that we fully rely on human forecasting, and there's no forecasting element, but now we are slowly moving towards where we use HEIPredict, which is our forecasting model, and I'll explain in a bit what's behind that. We use that to second-guess our man-made forecast. We use it as a second opinion, so people can compare, start getting more confidence in the model, and challenge it.

Robert Frentzen 0:18:33.9:

Where we think we'll move towards, and I think then also where we'll be looking at an integration into Anaplan, is combining the two. You look at your machine forecast, it tells you when momentum is going to go there, but I know that I'm going to do a price increase, I'm going to do some change, and you add that on top of the forecast to get a better forecast. I think that's where we feel that financial forecasting is leading. A little bit, a quick picture on how our model - it's an internally developed model, and we're looking forward to at a certain point integrating that into Anaplan, but we take our data from our central CIL, it's Company Information Logistics. That's basically where we gather all of our financial information for consolidation and so on. We do pure time series forecasting, but what our team has built is a model that takes all of the different forecasting techniques and basically, on a combination of OpCo and KPI, always picks the best predictor. That's updated every month, and therefore, every model is improving more and more.

Robert Frentzen 0:19:58.3:

At the moment, that's just as a report available, and we compare it against our forecasts, but that's really something of the last few months, where we started learning to use its output and challenge our business based on it. That's a little bit where we are at Heineken in our journey to improve forecasting and how we use Anaplan.