Nathalie Post  
In this episode, I'm speaking with Ciaran Jetten, Manager Centre of Excellence, advanced analytics at Heineken. He shares the story of how they got started and scaled advanced analytics and artificial intelligence at Heineken and elaborates on their value driven approach and mentality. So, without further ado, let's get into it.

Hi, Ciaran, and welcome to the human centred AI podcast. I'm really excited to be talking to you today. But for the people who don't know who you are and what you do, could you give us a little introduction about yourself?

Ciaran Jetten  
Yes, of course most happy to. So my name is Ciaran Jetten and I will be 49 years old this year. And so as background in business studies, where I studied in Groningen on beautiful memories there. And after working for KLM for three years, I joined it and basically started in the area of data and analytics, which is now 20 years ago, already quite a long time. For a long time. I've been working for several IT companies, so really as a supplier. And then roughly six years ago, well back then everybody called a big data, which now we would typically call AI or advanced analytics as in the case of Heineken. So I also started my learnings in this particular area. And that's been my focus ever since. So, after working for as an independent contractor for a couple of years, I joined Heineken for actually as an independent contractor, which is now almost three and a half years ago. And after working as a contractor for Heineken, for about a year, both Heineken and myself were really happy about all the arrangements on the opportunities. So then I joined Heineken as an internal employee.

Nathalie Post  
Yeah, great. And so right, because also in anything, your role really evolved, right? Can you tell me a bit more about that?

Ciaran Jetten  
Yes, of course. So anyway, I joined Heineken three and a half years ago. Initially, my focus was on a programme aimed at revenue management, where we were building a solution with an external partner. And then after a couple of months, I heard there was a plan to build an entirely new team focused on big data. That's why I reached out to ask, okay, what is the plan? And then I basically heard that there was no plan. And so I thought, okay, so I have many ideas about how to approach this and what the actual topic is, and how to make a farce. So I pitched this story with my manager. And he liked it. He said, Okay, well, you start and go ahead. So actually now, three years ago, I engaged with an external partner who helped with first engineering, support and some data science support. So basically, with a very small group of people, we just got started, and all with the aim to create enthusiasm and start building stories of Okay, so but this is how we can apply data science in in Heineken. And so back then, you know, I was kind of the lead for this initiative. So back then it was really just an initiative project. And then through the success, we heard and evolved and became really a formal department and team. So formally now, I'm the manager of the global advanced analytics team. Yes, indeed, evolved quite a bit over the past couple of years, but in a very exciting way. And I think there's still a lot more to come.

Nathalie Post  
Yeah, no, I mean, that's great to hear, but also so excited. Like you were there really from the start and seeing this evolving and going through that process? Because can you explain a bit more about like, how that went. Like what were kind of key steps along the way and key learnings in that?

Ciaran Jetten  
Yeah, so as I mentioned, important parts of what we do is just start with some first projects. So basically the first half year, we were really flying below the radar, and engaging only with several people who we knew were interested and also to make an effort. So I think that's a really important part. And second thing that we did was that we enlisted some help from McKinsey. And they helped with creating some artefacts, so we had some experiment guide. So really, explaining and showing, okay, this was the flow. So basically ending up with a beautiful slide deck, which typically is what what companies like McKinsey can do. But it helps a lot in sharing and selling the story. And also setting up some some initial governance, which was very lean, really aimed at speed. So this was like an advisory board with some senior people from our management's. And they really helped to create some focus on, okay, what what are the interesting areas to start experimenting in, but also giving a lot of support, to finding the right people to engage with and removing barriers. So in getting started, this was a really important part. And once we had finished a couple of projects, we start sharing stories. And I think sharing the stories is one of the crucial things if you want to get started and not sharing stories of what another company did. But really the stories within our own context. That's one of the first stories that we had was on on our stocks. That's where we collaborated with retailer Tesco. In Hungary, they provided us with a lot of data. And our after eight weeks, we had a model that could predict out of stock events, and then also showing where the pain was. So if there was an out of stock in a particular shop, or if it was in the distribution centre of Tesco, or if the programme was actually in our own distribution centre. And this really helps people to understand, okay, so but if this is what we can do, then they see the opportunity. And then they start to realise, okay, because they know all the issues for our stock events is very well known issue within a company like ours. So finding these stories, really, really helps and helps to further expands, the group of people are interested and also helps to expand budgets, because that's also what we have the stories we showed. Okay, so this is what can happen. And then especially I just received additional budget to further build a team and there's a really expand the number of projects and experiments that we could start.

Nathalie Post  
And so you're talking a lot about experimentation? Can you give a little bit of a quick view of how you approach experimentation? Like, what are your, what are your ways of doing that and creating that mindset?

Ciaran Jetten  
So it all starts with defining the right question. As far as the think phase is really important. And it's not only about thinking of like the business issue that I have, What is the problem that I'm trying to tackle? So the the example of out of stocks, okay, so that's the problem. So we use some elements from lean startup. So the value proposition canvas, where we define a persona. And ideally, we already have this, or a representative of this persona in a first workshop. And then just naming okay, but what are the pains and the gains that this this persona is experiencing? And this was a really important part in shaping the question and really understanding what the process is and where a particular pain point is, and already define okay, what could an end solution look like that would actually support a pain point that this persona is experiencing. So, in support of this, we also put a lot of emphasis on value. So, again, the example of out of stocks as well there are wide spreads research documents on location, but what is the potential value if you can lower stock events, which is a big hazards one of the really big cases as making the relationship between the pain point and the value is an important one and the persona itself. And then what can this persona actually do? So, what action can this person take to in this case, lower the out of stock events? And this was a really important part of the conversation because if you do not have the right persona or the right end user, or if you cannot define what the action is that the person is going to take, or maybe you can define it, but it's just out of the sphere of influence. So, this definition is a really important part. Only if this is clear, then it makes sense to start looking at okay, so, then what is the data that we need and what kind of model do we need? So, this is all in our thinking phase and once we have defined this, then we focus on speed. So we use three or four Sprint's of two weeks to build a very simple products. That's where initially we take the first sprint to make an analysis of the data. So what what is actually happening? Do we have sufficient data? And is the data quality right? Or are there maybe some some strange things that we can eliminate the second sprint second two weeks we focus on building a model. So starting with a very simple baseline model and then seeing okay, so, what simple choices can we make to choose a model that actually makes some valuable predictions and then a third sprint is focused on simple automation as far as getting flow that we can do an out of stock prediction on daily level. So then we have a product and then the second part of the experimentation is testing it. So there we really focus on again involvement of this end user and doing a real life test. Preferably doing AB testing so having access to a control group and then doing this for a specific period of time. So that at the end, we can do a value measurements and see okay, what so now we have a test group where we followed the predictions recommendations to prevent these out of stock events and the control group where we didn't, what has been the difference and is statistically significant. And then again, do a calculation of okay, so, if we would further implement this and scale this, what would be the actual value? So basically, the outcome there is that you have like a very beautiful and complete business case that you can use to make a decision on okay, so are we going to invest in making this a real full fledged product or not?

Nathalie Post  
Yeah, no, exactly. And who do you have involved in that whole process? Like what kind of team members skill sets? Like, can you tell us a bit more about that?

Ciaran Jetten  
Yeah. So obviously, we have the data scientist, who plays a key role as the person to choosing and building the model. For this person in this experimentation phase, the data scientist is involved full time, in part time supported by an engineer to make the data pipeline, which can be sometimes really simple and made of email chains with excels at because the purpose is not to make a perfect solution, but it's a solution that works in a short period of time. And last but not least, we have a translator involved as well. This is also a role that we see as a key role in the success of of implementing AI solutions. And it's an important part also of how we see that we drive the change.

Nathalie Post  
And so these translators, do you like, what makes the translator a translator?

Ciaran Jetten  
Yes so sometimes people also call it an analytics translator. So typically, what we see is that there are two different rules where two different languages are being spoken. So we have our the business world for people talk about OSA on shelf availability, and an OOS out of stocks and customer satisfaction, of course. And, of course, tonnes of other terms that are being used on the business side. And then on the side of the data scientists and the engineer, they're talking about, recall, they're talking about pandas, they're talking about bayesians, they will both have their own languages. And I think this is like a classic, a divide between more technology and business oriented people. So how we see the role of the translator is the person that actually really building the bridge between those two areas, and the ambition that we have, it's not something we thought of ourselves, but it's coming from McKinsey is that for every data scientist, we have 10 translators. And for us, the translator is really a person who sits in the business, I suppose it could be a business controller or logistics manager, or a brewing manager, or sales manager as a whole in these these different areas of our business but who has received an additional training additional education, to understand the world of analytics. So that's, if the data science scientist talks about accuracy, that this person knows what it actually means and how to interpret this as well. And then the translators are the ideal persons who can think of business issues, because they are deeply, they should have deep experience of their own particular business area. But think and think of okay, so how can AI help support in my particular business area? So that's the first part two, so it's basically the think. And then the second stage is to build a solution in collaboration with the data scientists. And then third, which is probably the most important part is implementing and making sure that the solution is actually being used. So it's really on change management.

Nathalie Post  
Yeah, yeah. Because talking about that, can you explain a bit more on how you're embedding well, first of all these analytics translators, but also artificial intelligence in general, within an organisation that is such a global organisation like Heineken. How do you deal with that?

Ciaran Jetten  
Yeah, so we're doing our best, but if we will ever be done.... I don't think so. So but this is really something where you need to take a long time horizon. That's why I said I started three years ago. So embedding first I saw maybe it's classic. So thinking about okay, so awareness, interest, desire action, so still also within Heineken. There are quite a number of people who have very low awareness on what what is what actually is AI. So there we have already come quite a way because we spend a lot of time on on evangelising, as well by doing a lot of presentations, workshops, also actually giving people notebooks to do, just push enter, but just to experience some Python code. So it's really doing about a lot of diverse activities, to create the awareness and the interest but the biggest impacts that I see from really embedding is having people involved in an actual experiments or the actual use of a product. Because then people are not hearing things or seeing but really experiencing. And it creates just a different view on their own own business area. By going through an exercise where some sort of predictive models being used. That, in general, and I think more specific, it helps a lot to, so I call it like the to find the coalition of the willing, because there are always people all over the company who also even today, so they approach us, because they were like, Oh, you know, I just started in Heineken. And I've always been very interested. And I do have some sort of background in this. And, as mentioned, these are the ambassadors. Yeah, as we're really keeping these people close. So we involve them in our community, where we have very regular updates, where people from all across the globe are sharing their experiences. But having these people and really using them as ambassadors, I think it's a really important one in ultimately, embedding AI in in our business.

Nathalie Post  
Yeah, no, that is, I think it is such a valuable answer, or well, also so valuable to have these ambassadors that really will drive it almost from the bottom up in a sense, as well, as you know, you have the top down, but also, it's coming from both sides in that way. So yeah, super interesting to hear your experience in that

Ciaran Jetten  
Maybe also, just to mention, right from the beginning, we've had a very important sponsor on almost the highest levels, which was our CFO last year, and she has been the one that said, okay, so I'm gonna make first bucket of money available, just to invest and get started. That's why having this senior sponsor has always made a big difference. Because in the conversations that I've been having with many people in the business, I could just always drop her name and say, Laurence, she is sponsoring this initiative. And that ultimately helps a lot, of course, and, and so what has actually happened since the beginning of this month is that we have a new CEO. So from van Boxmeer, he has left and now we have Dolf van de Brink. And so you know, he's from a different generation. He's about my age. And he has a very different view on technology in general, and also data. That's why and it's already clear that he also very much supports this direction. So I think it will also help us in further, maybe even accelerating all our activities.

Nathalie Post  
Yeah. Yeah, no, that is great. And also, generally what we're hearing is really that both sides, the top is crucial, but going from the bottom to the top as well. So having that interaction, I think it's great that, yeah, you have that at Heineken.

Ciaran Jetten  
And now thinking of the top. So I mentioned already the part of the sponsor. So it's been driven by the upcoming change of our CEO. But also since just a couple of months, we now have a representative for technology in our executive team, as well, until now, that wasn't the case within Heineken. So we now have an executive team member who was responsible for digital and technology as well having a representative on technology, but also specifically about data and analytics is going to make a huge change.

Nathalie Post  
And so maybe to just based on what you've learned over these past years, and like scaling, analytics within Heineken, what would your advice be for organisations that are just getting started with advanced analytics and artificial intelligence to approach these things, but also in their journey to scale up.

Ciaran Jetten  
So I think I already mentioned a couple of things. So I think the first one really, to get started was build these stories and how to build stories. So we have used a one liner, which is, don't get ready, get started. And it is very, very true, you know, do not build an extensive plan or get started. Because I already mentioned that just going through the experience has the biggest impact. That's why, don't get ready get started. And then  because you get started, you're investing in building these stories, which you can then share across the organisation. So this is a really important part in experimenting and getting started. Once you really want to scale and create impacts, an important lesson that we learned is that it It should be completely aligned with business programme. So just solution can be very good and very valuable. If it's not supported by a business programme, the real change management bars, so the risk of failing is large. So where we have seen the biggest success in really building a product and scaling it across multiple countries is where it was related to programmes. As for Heineken, again, this was being individual data driven marketing. As we're really on the marketing side, we have a similar programme, data driven sales. And the third one around planning. That's what that's where it really comes together and where you can make the biggest impact.

Nathalie Post  
And I'm assuming thing, you also really have where you were talking about the personas and who you're building it for, if the end users are also more internal, you have the more identify them related to their current needs at that time, I'm assuming. And so maybe, well, we've talked a lot about, like, where you've been over the past three years and what you've done. How do you go forward from where you're now? Like, what are your main things that you're focusing on within your department.

Ciaran Jetten  
Yeah interesting question. Yes, what I just mentioned is that if you really, really want to create impact, as well build these products and scale them out. As well, we started scaling out our first products a little bit over a year ago. I still have the feeling were at the beginning of really scaling out. But now we are accelerating. So what will happen is that we will be working on more products and implementing them in a lot of countries, as well for expansion of our product portfolio is one of the things and then depending on circumstances in the markets and our guest position, we will be able to further expand the team. One of the areas where I'm now also putting in efforts is basically around ultimately around retention of employees. And why because we so we have been quite successful in attracting talented people. But also we've been investing in employees to become a translator as I mentioned earlier. That's what I really want to look ahead to see okay. So how can we further help these people in developing themselves and creating career path because typically for specialists in general, but for certain in this area. Within Heineken, we do not have a clear cut or not even clear cut, we do not have a career path at all. That's why what we're working on is building an education portfolio based on profiles, personas of people that we have in the company. And then also making, we're going to make examples or forecasts of what our potential career paths, for example, may be someone that started as data analysts. I want to be a translator, or maybe this person would like to move to business control or, or become a data scientist. Or if you're a data scientist, so yeah, of course, you can become a senior data scientist and maybe lead data scientists. But then there should be options to branch off into becoming maybe a sales manager as well, that really focusing on retention of the people that we are investing in, I think is a crucial area in the coming years. And I mentioned earlier, just building further awareness to do the change management, that to embed AI, and the use of the solutions. So that's going to be continuous. But it's going to be a big chunk of, of the work that we do.

Nathalie Post  
Hmm, yeah. Well, I mean, I'm super curious to talk to you in a couple years and see. I mean, when it comes to data scientists and analytics translator, I think there is, you know, the market is quite competitive. So I think it is really great, like you are carving that career path. So to have a way forward also, maybe as a closing question, but what we could advise to people who, you know, maybe are in a slightly different field right now, but would want to branch out and go into a data science direction or analytics translator direction, what would your advice be?

Ciaran Jetten  
I would say do it! Yeah, so of course, it depends a little bit on the background that you have. I do know, there are quite some different interpretations and definitions of data scientist. So that's also the reason why we've built on our own definition. And I also know that there are different definitions of the translator, as we know, a couple of companies that have defined translator as also being capable of building simple machine learning solutions. You know, both very possible. Yeah. But I think that the majority of people, so then I'm counting like the majority of people being not that technical. investing time in understanding and developing the skills that you would need for the translator, is very valuable. And actually, I think, maybe already in like, five years, or six, seven years, we will not be talking about this translator role anymore, because everyone should have these skills. It's more that we're now in a transition, that we named this as a specific role. I cannot imagine a future where people will not have these skills. That's what I truly believe that if you do not understand data and speak the data language, you know, you'll make yourself redundant.

Nathalie Post  
That data literacy, I think it also will become more and more part of the curricula of people in studies in general education. I mean, maybe in a couple of years, we'll have quantum translators or whatever we might get.

Ciaran Jetten  
So maybe just to summarise my advice. If you do not want to become a dinosaur, invest in these translator skills.

Nathalie Post  
I think those are amazing closing words. So thank you so much Ciaran, for taking this time.

Ciaran Jetten  
It was my pleasure.

‍