(0:03) Welcome to the supply chain planning podcast, where we speak to supply chain professionals (0:07) about their experience in planning. I'm Ben, co-founder of Horizon, and Wim is our co-host. (0:13) Hi everyone, I'm Wim, I'm managing director of LongArch. 

(0:16) For today here, our guest is Hadi Eglidi, who is the head of demand management at VAT Group. (0:22) Welcome, Hadi. (0:23) Thank you. 

Thanks. Happy to be here. Thank you for the invitation. 

(0:26) Yeah, thanks to you. So as a beginning question, we'd like to know a bit how you got into supply (0:32) chain, because you have a bit of an unusual background, I would say, for people who work (0:37) in supply chain. You have experience as a lecturer and a data science background. 

So (0:42) how did you get into it? Was it planned? (0:44) It just happened, if I want to say it in a nutshell. So I get this question actually (0:50) many times, very often, that how did you get to supply chain planning? But if you look at my (0:57) history, you see that, okay, it was not also completely random. So as you said, I have a (1:03) background in science. 

I have been a lecturer at university. I was a group leader. I did a lot of (1:10) research, publishing high impact journals, journal papers and such, patents, technology (1:16) development before in the R&D. 

So I had a background more on the technology and engineering (1:23) and physics. But at the same time, I was part of two startup activities also. And I had an (1:31) MBA where I got quite some good theoretical background in supply chain planning, but I didn't (1:39) put it into use until I started studies in data science, advanced degree, something similar to (1:46) a master's at ETH, in data science, machine learning, deep learning and such. 

And then (1:52) when I came to the thesis, I wanted to do my thesis along the way of supply chain planning (1:57) and demand planning, or forecasting, actually business analytics. And then I got in contact (2:04) with the senior management of the company, got quite some data, got some tasks to do. At the end, (2:10) I developed a pretty complex dashboard where I analyzed basically a good part of the supply (2:17) chain, put some machine learning, some deep learning on top, and then tried to forecast future (2:23) and then get some insight, for example, how the dynamics of the supply chain is, how the inventories (2:28) are in the supply chain, and with the goal of really forecasting what the demand of our company (2:35) would be. 

And that's how I got to the topic. And I continued. And then when the opportunity (2:41) appeared, I just took the chance and moved to, like wholeheartedly moved to the planning (2:48) world, basically. 

Okay, so you really came from the techie side, because the words that you're (2:55) using and throwing in the conversation are for a lot of people like, yeah, I heard them once, but I (2:59) would have no clue what to do with it, at least for most of the supply chain people. But that's (3:03) the world you came from, and you actually moved into supply chain. How did that go? Was that a (3:09) difficult transition? Was it easy? How did you make that move? Yeah, that was different. 

I mean, (3:14) the environment is very different. I was before in the technology development part, I was the head of (3:20) content center, smart systems. I was very much also working in the same company on smart systems (3:25) and such. 

And I was very much surrounded by people like me, very technical backgrounds, all the (3:31) developers and such. But when I came to the supply chain part, I saw a lot of business people that (3:37) were speaking a different language. So that was the first thing that hits you, that you see that (3:43) people think differently, more strategically, maybe. 

And then, yeah, I had to find my place there. It (3:49) didn't take me much time. But yeah, basically, I could identify myself as like the more techy guy (3:57) that also learned the business part and then tried to a little bit to get a kind of a unique twist (4:05) that could bring some of these technicalities to supply chain planning or to demand planning (4:10) specifically. 

For the audience, can you briefly introduce the company you're working for now in (4:15) the field of activity? Absolutely. So we are VAT Vacuum Valves. As the name suggests, we are (4:22) majorly producing vacuum valves for anywhere that you have vacuums. 

We are a manufacturing company (4:28) very much, kind of mechanical engineering company. I could also be more specific. And then we supply (4:35) vacuum valves to whatever industry out there that uses vacuum, which are many industries, (4:41) and majorly semiconductor industry, where all the chips, all the electronics are actually (4:49) manufactured in vacuum. 

And these vacuums need gases in and out, regulations and all that. (4:55) Everything is in vacuum. And we are the market leader in providing the valves that they need. 

(5:02) So we have something maybe close to 80% of the market in semiconductor industry, which is, (5:06) by the way, one of the probably the largest supply chain that you could find. (5:11) And we, at the same time, also do something we call adjacencies that go a bit beyond valves, (5:17) for example, motion components, modules that are a combination of different valves and such. (5:24) And was it hard to, because demand planning often risks to be just a numbers game and data (5:30) scientists looking at curve fitting or generating forecast. 

But that only makes sense if you also (5:37) add the business understanding and what is really behind it. How did you acquire that knowledge as (5:42) well? I mean, we have our own specifics. I mean, if I don't know if you want me to go already into (5:49) the details of the demand, the specifics of the demand that we have, but pretty complex supply (5:55) chain. 

Imagine that we are very far from the end customer. We supply the valves to the original (6:02) equipment manufacturers. They give the equipments to the fabs basically. 

The fabs create chips for (6:09) electronics goods market companies, electronic goods companies like Apple, Samsung, and such. (6:15) They sell the final product. Therefore, we are in the supply chain. 

We are a couple of layers away (6:20) from the end customer that brings a lot of bullwhip effect, if you will, that brings also a lot of (6:28) volatility in our demand in a way. And then at the same time, since the market is kind of like (6:34) the applications are fragmented, we have a large portfolio, a lot of products, and the supply chain (6:40) is pretty complex. And the suppliers that we have are also many. 

The audience will typically be people (6:47) active in supply chain, and everybody in supply chain has heard about the chips shortage we had (6:52) in the past, was it one and a half to two years? Did that create a ripple in your demand as well, (6:59) or was that too far up in the supply chain to have an impact on you guys? (7:04) No, no, absolutely. So imagine anything that happens in the supply chain, we see it with a (7:09) higher amplitude than anybody else probably, because we are kind of several layers down in (7:15) the supply chain. And then the bullwhip means that whenever there's a bit of a hiccup on the (7:21) electronics goods market, that gets enhanced, every layer that you go in. 

And then by the time (7:27) that it reaches us, it might have an amplitude of 10 times even. So in our case, it might happen (7:33) that we have a growth of easily two digit growth one year and two digit decline in the downturn. (7:40) And all these shortages and such, we saw it also very much. 

Actually, what happened in our case is (7:46) that it drove our business, because then there was more demand, there was shortage. Whenever there (7:51) is shortage, it means that the utilization rates of the fabs are high at a high level, (7:57) meaning that they want to increase capacity, meaning that there will be investments, (8:01) and whatever they invest, it will go to us, because we are basically providers of these (8:07) critical subsystems that they use in their systems. However, it means that at some point, (8:13) all these constraints are removed. 

When the constraints are removed, the companies start (8:17) building up inventories, because they were scared of what happened in the past. They build up (8:25) inventories. And then as soon as the demand slightly goes down, they stop building, they (8:31) start using their own inventories. 

And that is the driver of the market going down, (8:37) meaning that the companies that are providing the components, they feel it more, because the (8:43) customers that are making the equipments or the fabs and such, they start using from the stock (8:49) that they have already, and they stop buying. And I think most people active in the supply (8:55) chain one day have played the beer game. It's almost compulsory education for people active (9:00) in the supply chain. 

The lesson you learn there is to counteract bullwhip, you need to be (9:05) transparent and communicate with the other layers in the supply chain. Is that something which is (9:10) happening actively on your end as well, or you're basically a sitting duck because there's no way (9:15) of counteracting it? Another specific of our industry is that it's a kind of a secretive (9:21) industry. It's driven by some big giant companies, for example, in chip manufacturing, (9:27) there are only very few players like TSMC, Samsung, now Intel. 

There are not many. (9:35) Also when it comes to equipment manufacturers, there's ASMF as the big company, the most critical (9:41) company in the supply chain probably. And then there are others like Applied Material, (9:46) these equipment manufacturers that are important, but it is not that they easily share information, (9:52) easily share their business with others. 

Of course, you could always extract from their (9:58) quarterly reports what they publicly announce, but it is not always the case that you could easily (10:06) understand the dynamic of the supply chain and what will come. And it is also not the case that (10:11) almost anybody understands. It's a very difficult supply chain to forecast in a way. 

And then even (10:19) when we talk to the companies in other layers of the supply chain, they oftentimes also complain (10:26) that they don't have visibility. It doesn't mean that there is absolutely no visibility, (10:31) but it just means that it's difficult. There are always risks to any decision that you make. 

(10:36) However, there are some, I mean, from a more data science perspective, there are some indirect ways (10:42) that you could really get some very good intelligence about the markets. (10:47) And these are the things that we are, by the way, also very much trying to do (10:51) internally in my team, that we look at different indicators. As I said, we have (10:56) gathered something close to maybe three and a half thousand different indicators of the market (11:02) from different sources and such. 

What type of indicators are these? You don't have to mention (11:07) specifics and maybe... Yeah, no, no. I mean, these are not very also secretive and not all of them. (11:12) Some of them are really the ones that are not easy to get, but there are, for example, the (11:19) equipment revenue that is going to be generated in the market. 

How was in the past? How much is (11:25) going to be in the future? You could have it in different categories, like different technologies. (11:31) For example, when it's, I don't know, the smartphones, the personal computers, data (11:36) centers, the 5G, all these things, you could also have data about those ones. These are related to (11:42) different layers in the supply chain. 

Then you could come down to another layer that is (11:46) equipments. You could have the revenues of the equipment. You could have some good overall guess (11:52) of the inventories in the supply chain. 

For example, how much are the inventories of the (11:57) baths? How much are the inventories of the equipment? How much are the inventories of... (12:01) Okay. The inventories is a difficult topic. Maybe that you don't have with that much granularity, (12:07) but then there are some other, for example, different cheap technologies where you could (12:13) have more granularity on the technology, and then you could know these topics like them. 

(12:19) Okay. But from a data science perspective, what you just described, getting that information (12:23) and then creating those models, is it something that you fully do in-house? Is it something that (12:29) your referrals also rely on external providers? Because similar companies or other companies (12:34) might also have those data science projects going on. So do you keep it internally? (12:40) Absolutely. 

Getting data is the major part of it. Then a storytelling with that data is another (12:47) major part of it. But that comes after getting the data, high quality data, reliable data. 

(12:53) So this data we get from many different sources. In my team, we have very large interface with (12:59) basically a big part of the company. We always talk to the people, to the sales organization, (13:05) which is in direct contact with the customers. 

We at the same time get some data from different (13:11) sources. Sometimes we pay for this data that comes, for example, from some consulting companies, (13:17) from some market intelligence companies. And then there are some other data, as I said, (13:24) from different sources that you also go after and try to get it. 

But as I said, it needs its (13:30) own effort to get the data. But a major part after that is what to do with that data, (13:38) how to make a sense of it for your own business. And that part, because also you have another role (13:43) within VAT before your demand management role. 

So talking about AI, I think you also manage (13:48) different projects in there. So how do you differentiate between keeping a data science (13:53) project in-house, developing in-house capabilities or going externally? And how do you integrate (13:59) data science within supply chain as a team? So let me focus. I mean, before moving to (14:05) supply chain, I was part of the technical side of the company more R&D and in the area of smart (14:14) systems where AI is also used. 

But when it comes to supply chain, it's a bit more difficult because (14:22) there you're not really the developer. Your main task by far is the taking care of the business. (14:29) And you're not a software developer. 

You're not even a code developer and such. (14:35) Therefore, you cannot really do everything from A to Z. And the north star for you should be (14:43) having an advanced planning system, bringing a software, a software that does all these things (14:51) in a professional way. But getting to that maturity, to that level, to have that advanced (14:56) planning system is not easy. 

You have to go a couple of steps. And on top of that, you have (15:02) some applications where APS cannot fulfill. You need all this KPI tracking, benchmarking, (15:11) like seeing where it could improve and such. 

And then for that one, you need to have a certain level (15:16) of, first of all, understanding and capability of doing it hands-on. And therefore, I very much (15:23) advocated always and tried to do to build a team at the interface of data science and business. (15:31) Meaning that there are people that are business experts in the team. 

There are people that are (15:36) data science people of the team. And then there should be a mutual kind of understanding, (15:42) a collaboration, and then solving the problems one after the other. (15:46) That's a very interesting topic that you named these two roles, the data sciences and more the (15:51) business oriented. 

If you could choose, is that the way you would do it? To have those two different, (15:57) distinct roles, the geeks and the business people? Or would you want a profile, a kind of person that (16:06) actually is capable of doing both at the same time? What is your goal? Also, do you want those (16:11) geeks to be more business savvy? And do you want the business people to get more technical? (16:16) Or are you fine continuing with the two roles in the team? (16:19) Absolutely. That's a very good point. Actually, what you need to do, you cannot change people, (16:24) right? You get people with a background in the data science and you get people with a background (16:28) in the business side of the task. 

And then you train them both. Meaning that the data scientists (16:36) that come in should have an intensive time of really understanding the business, should be (16:42) open-minded to understand the business, and should have this kind of mindset and personality to go (16:48) and talk to people, what people do in the business, right? And not much in the technical side. (16:54) And then the people that are the business people, they should also be open to learn that machine (17:01) learning, the technical side, so that they could define the requirements that they have (17:06) to the data scientists. 

So they could develop the tools for them. (17:10) And then if you look at, I don't know, how big is VAT now in terms of, for instance, turnover? (17:16) Is it realistic for any company to start building this capability? Because if I hear you explaining (17:22) it, it really is a strength of VAT having this capability. Is this feasible for any company to (17:28) start infusing this data science into their supply chain department or team, however they're (17:35) structured? Or do you need to have a certain size? How do you look at that? (17:39) I would actually challenge it the other way. 

I would say that sometimes if you're small, (17:43) you might even need more of such people. That's interesting. (17:48) Because when you're small, you cannot afford to go and jump on buying an advanced planning system (17:55) because you're just not mature enough to really go and outsource all your tasks to a software to do. 

(18:02) You need a certain level of maturity. At the same time, you're growing. A lot of needs pop up on the (18:08) way. 

And then you cannot go and buy a software for every problem that you have or outsource it, (18:15) ask somebody to do it for you. Therefore, you need to have these dual profile people (18:20) that can take care of the business at the same time, could do the hands-on data science part, (18:26) so that you could provide this agility. These new things that are especially in the last couple of (18:33) years are very much needed in supply chain. 

I'd like to challenge you a bit further on that. (18:39) I think from conversations that I've had, it's been only those biggest companies that have this (18:46) data science within supply chain. And from my perspective, it's been because resources, (18:50) of course, to get all of that data and to create those models specific to the business takes quite (18:55) some time, even if the person has some expertise. 

Whereas maybe for the smaller business, there might (19:00) be a solution out there, perhaps not as advanced, but something simple that could fit their needs (19:05) at that moment before they get a lot bigger. But you would say, no, actually, smaller companies (19:10) also would need that. No, definitely. 

So what I'm saying is not that you develop everything (19:17) in-house. Of course, you need companies that serve also that gap that exists. And that gap is not (19:23) very much service, to be honest. 

Then when you go to the space of planning, you see all these APS (19:29) companies, end-to-end advanced planning softwares, but you do not see much of these companies that (19:35) could come in and serve the companies that don't have, first of all, much resources. They cannot (19:43) take the risk of really switching to an advanced planning system right away. They need a certain (19:49) level of maturity before that. 

But it doesn't mean that they don't have the needs. They have the needs. (19:56) And therefore, if there are companies that somehow can position themselves to serve such types of (20:03) needs, for sure, they are welcome. 

But if there are not so many of such companies, you better (20:09) not stop yourself and develop these skills by yourself within the company. And you're now coming (20:16) in from the demand management side, where I think it is a number game, trying to forecast as (20:22) accurately as possible using 3,500 external KPIs or insights. Do you see the same possible impact for (20:31) people with a data science background in other elements of supply chain? Or do you think that (20:37) no, amount management is really the place where that can have the biggest impact? Or can we draw (20:44) that further into other elements of the supply chain? No, it is because, for example, machine (20:49) learning is not only about forecasting. 

Machine learning is about clustering, classification, (20:56) lots of these things like abnormality detections, outlier detections and such. And all of these (21:03) have applications across the whole supply chain planning. For example, you want to see which (21:09) suppliers are selling the goods to you with a very high price. 

You don't have that visibility, (21:15) because you have a lot of suppliers, you buy from so many different companies, you have dual (21:20) suppliers and such. Here, a simple kind of classification or clustering technique (21:27) could help you to identify shortlist the ones where you could go and negotiate. (21:33) And so switching gears a bit, I mean, like Wim said, this is more about numbers and machine (21:37) learning, but on the, let's say the people side or collaboration side within VAT, because I can (21:44) imagine if you have those customers, and if you're in this complex supply chain, you do try to get (21:50) information from them. 

And also from people internally, I can imagine salespeople who have (21:54) information. How do you deal with that? Maybe first with sales, like how many people are you (21:59) interacting with to get information from? I can say that we are in touch or collaborate with more (22:06) than 70 people all across the sales organization in different countries. We also are in touch with (22:13) and collaborate with the product management teams across business units, and then sometimes are in (22:20) touch also with the supply chain managers that do the supply chain management, that you are the users (22:27) of that forecast, at the same time contribute to that forecast. 

Yeah. So if you speak about those (22:34) salespeople for a bit, we're now talking about the numbers and machine learning on it. How do you, (22:39) let's say, merge that kind of information from salespeople, what info they have on customers (22:45) or of the market, with whatever you have identified from a data science perspective? (22:50) What we do is that we do kind of a bit of sequential planning, meaning that we always (22:56) start with the sales organizations, we start with the customer forecasts, any AI calculations that (23:02) we want to bring into it, and then we make a baseline based on that, make some scenarios and (23:07) such, and then it gets reviewed by the sales organizations. 

Then we get all that information (23:14) or all that forecast back, and then together with the business units more internally, we come to an (23:21) agreement and come to the final numbers. But basically, we get input from all the parties, (23:28) and then we oftentimes have very many, actually, short meetings with every stakeholder with a very (23:36) clear agenda. Is it on a monthly basis? Is that a monthly cycle or a weekly cycle? We do monthly cycle. 

(23:41) We do monthly, but within a month we have different stages that we go through, and we really have a (23:48) touch point to almost every stakeholder. Trying to get a new initiative started or trying to make (23:54) a change to a process, how do you approach that, or how would you recommend people approach it? (23:58) We've spoken to some people, and some of them say, look, try to do it first on yourself, get some (24:03) initial good results, then propose it higher up, or should you immediately try to get validation (24:09) and other people on board? No, the way that we do is this. We are the experts, so you cannot ask, (24:16) not all the opinions are equally important. 

We are the expert in this topic. We define the process, (24:25) we define the improvements, we show it with numbers that such a thing could improve, (24:30) and then we define the roadmaps and the changes that have to happen. (24:34) With that, normally we go top down. 

We go to the senior management, get the sponsorship, (24:41) get the agreement, we present it, we maybe iterate it, and then with that one, we go and make the (24:48) change. Then these KPIs are also not something mysterious. For example, accuracy is one KPI. 

(24:54) Forecast value added is one KPI that needs to be improved. Efficiency, meaning the demand hour (25:01) spent on the forecasting is one KPI. These are the things that we track, and then at the end, (25:07) numbers talk. 

Okay, cool. Something that kept in the back of my mind since earlier in the talk (25:13) where you were also explaining, we don't want the data scientists and a business person, we want to (25:20) merge them or blend them. Also, I think it's a bit linked also to what you said just recently (25:26) about we are the expert. 

I loved how confident you said we are the expert and we are doing this, (25:31) but somehow you probably need to convince people as well that you are the expert. (25:34) Is there any education or evangelization needed there as well to convince these other functions (25:42) to listen to you, or how do you handle that? Yeah, absolutely. As I said, at the end of the (25:47) day, numbers talk. 

What we do is you definitely need for driving change, you need that confidence, (25:55) you need that level of understanding from your side, plus a bit of risk-taking. (26:02) When we drive it by ourselves, it means that we also take the risk of failures. (26:08) Actually, one essential thing here is having quick wins to show that things are going to the right (26:15) direction, and then showing multiple of these quick wins, and in the midterm, show the numbers, (26:22) the KPIs. 

Overall, then you get the momentum, and the more you succeed- (26:31) It opens up the door and then allows the further conversation. To be able to keep that up, (26:35) because you also mentioned continuous improvement, I think your team, because you're referring to a (26:41) team, you have multiple people reporting to you, they need to keep on evolving as well. How do (26:46) you handle that? In this setting of a very technical aspect combined with this very (26:52) complex collaboration aspect as well, how do you keep your people growing? How do you handle that? (26:57) I try to mentor people. 

I have a very long history of research. I always read books. (27:05) I always go and try to figure out what is the best practice, and I try to a bit mentor the team, (27:12) and then define the tasks, also ask them to do, if even do sometimes literature review, (27:17) which might sound a bit weird in the setting that we have, but we do such things. 

(27:21) Is that working? Do they like it? (27:23) It does. It does. As I told you, I bring this scientific way of thinking with me, (27:31) which is not very much practiced maybe in the supply chain planning. 

Then we always start with (27:37) research, basically, with learning and research and questioning the status quo. (27:43) Then I always tell to my team that, okay, I give you time to just go explore. (27:49) Don't get your hands dirty right away, and go explore what is the best practice here. 

(27:54) Then they oftentimes come back with very interesting suggestions that I learn from them. (28:01) Any tips or hints you have for our audience on the good resources that they can tap into, (28:06) or are you willing to share some of the data or the learnings you've collected? (28:12) What would you recommend people to also go down this route? (28:15) I suggest that the planning teams, especially supply chain planning teams, (28:21) to develop these dual profiles. If you look at just the evolution of the whole planning from, (28:29) say, 70s, a long time ago, people were doing all the planning on papers by writing and all that. 

(28:36) Then the MRP, material resource planning systems, came in, computerized. After that, (28:41) there was these ERPs, actually, that were introduced. These are all along the same line, (28:48) but new, basically, revolutions in the planning world. 

Then came these advanced planning systems (28:55) and end-to-end planning. That was the next revolution after the ERPs. (29:00) After that, a couple of things came in, like this machine learning forecasting. 

That was something (29:05) that came maybe around 2015, 2016. People started talking about it. My suggestion here is this. 

(29:13) Try to get your hands dirty. I specifically suggest that, apart from coding, use the (29:21) cloud services. Cloud services, actually, are very powerful tool sets that enable teams to (29:30) put together a small software event for different business applications that they have. 

(29:36) I see it as one of the trends in the market that I imagine in the next five years. You see that a lot (29:42) of teams start really using these tools that are available right now. They are out there to (29:49) democratize, basically, application development. 

Of course, here I should add that the business (29:56) planning teams are not, again, not software developers. 

They need some partners, they need some companies that should come in and enable them to do such things and take care of the infrastructure and help them acquire the technology.  

I imagine there will be companies in this type of space. 

Then there might be companies that develop solutions so that teams don’t need these solutions.  

30:45: You know your history, any books to recommend? 

31:03: In general what I suggest, to read this very old book. The Goal by Eliyahu Goldratt.  

32:02 the Toyota Way by Jeffrey Liker.  

32:40 books related to Six Sigma. 32:50 also suggest in direction of market intelligence.