Secret Ops Podcast | Uncover the World of Operations with Ariana Cofone
On this Episode
Marc Jansen, Data Scientist, talks about how he got into the world of operations and discusses the challenges, particularly in data science, and the importance of balancing multiple objectives.
We also dive into the power of technology and data in solving problems in smarter ways within the operations industry.
Highlights
[00:00:31] Marc's background in operations
[00:08:11] Finding the right career match
[00:12:38] The role of data science in operations
[00:15:37] Solving complex operational problems with Data Science
[00:25:11] The importance of data champions
-
Ariana (00:00:06) - Welcome to Secret Ops, the podcast uncovering the world of business operations, one episode at a time. I'm your host, Ariana Cofone. And for my first guest I have the wonderful Marc Janssen, a senior data scientist at Amazon, as well as a former colleague of mine and a very good friend. Marc, welcome to Secret Ops. I’m so spoiled to have you as my first guest.
Marc (00:00:30) - Thank you for having me.
Ariana (00:00:31) - I wanna start out just by context, setting your experience in operations. Cause operations can mean a lot of different things to a lot of different people, which is why I think there's confusion around it and you have a particularly interesting background. So can you walk us through how you started in your career to where you ended up today?
Marc (00:00:49) - Yeah, so I, I, I guess there's many paths to anyone's story, right? But let, let's focus on the operations path, if you will. So I'm from the Netherlands, did my undergrad, did not know where I wanted to go with, with all that. I did a mix of different things from physics to economics to museum studies of all things. But I, just enjoyed lots of different topics to learn about and study. I ended up falling into this world of operations management, which is kind of a mix of different areas of study from technical bits, to maths, to the sort of business context and also the problem solving aspect that lies in this, in this kind of area.
Ariana (00:01:33) - Did any of it bridge into the technological application of how do we automate some of this?
Marc (00:01:38) -I spent some time in Singapore during my studies. I got a chance to do a project in university in Singapore, and part of that was to do an experiment about my work. An experiment would not involve actual companies cuz it wouldn't be impossible to do so. Instead we simulated, the environment with a computer game and let students play it and sort of run their own little company inside of this game. It's not a fancy game, it's not the kind of game , Ariana, it's very boring. And I have to build the UI and I am, you know, I'm not good at that. But that was my first time realizing how cool it is to build stuff with code or with software and to then let people experience that world that you've created for them. Granted my world was really boring, but it was not the successful part.
Marc (00:02:28) - But it was nice to see things come out and then to try prove my theory. It gave me this way of creating a feedback loop that I'd never seen before. Spend some time building and experience a piece of interactive stuff that other people can work with, even though they might not know what's under the hood. That's gonna give you information, data, feedback, and that's gonna help you prove your case, whatever that might be. In my case, it was south theory, but, but obviously today, if you fast forward to where I'm at, that's kind of the way that a lot of data scientists work, right? I think build tools to process data, to make predictions, to optimize and the the real world will interact with that once it's deployed and that's the feedback you're gonna get. And that's where I understood. Or when I create these things that fit in the picture of operations and in operations, I think of how real things work, whether it's people and moving about their day or goods moving through the supply chain, right? That's, that's the broadest sense I have of what is operations. Operations is value creation, but the real meat of it, not the piece of papers that prove that there's value. It's through doing of things.
Ariana (00:03:45) - It's interesting that you say that operations, it's when you actually create value that is felt and seen in a day-to-day way. Whether that's from a financial standpoint, a human standpoint, a technology standpoint, but it's really the effect on the system. Not just creating a process to create a process. We don't want structure just for structure's sake, right? We want it to, to be impactful in some way. Wow. I I've learned such a new piece of your history. You had that light bulb moment when you're looking at taking these theories of bringing them into real life. And then within your work experience, you've applied it within a supply chain context, technology, education context. Now just doing pure data science, was it important for you to try different mediums to apply operations in, to learn? Or was it just sort of following, I don't know, a passion from one thing to another that led you through all these different operational niches?
Marc (00:04:43) - That's a good question. I don't think I've ever really reflected on whether operations is the passion or if it's a theme that I rely on to as a kind of my super strength. It doesn't, I don't think I necessarily have to be in operations and see all these different ways it comes to life, but it's kind of what I come back to. I naturally get drawn to it. So to give you an example, my first role, after finishing the PhD, I actually ended up joining as a data scientist for a food company. So a food company in London, a recipe company to be specific so folks might know various recipe kits you can get subscribed to online. This is one competitor in the UK and I joined as a, a general data scientist. I didn't have a designated area to work on. This company happened to have many things, many problems like so many other companies to work on with data science.
Marc (00:05:34) - So I started off with various things, I had some things around predicting future behavior with our customers, I worked on a recommender engine to, to serve up the best recipes, pun intended, on the website. Lots of customer facing things that are, I would call them a, a bit more quote-unquote traditional data science because those are the majority of problems people work on. Until I actually ended up trying to help on one of the pieces that, that were to be with our fulfillment center (FC). We happened to run around and a lot of the machinery wasn't ours as a company you buy that or lease it. But the operating of the machinery was our, our task as a company. A lot of that was obviously done to the highest standard, but not necessarily within data scientists. So at some point there's this idea to try and override some of the software defaults of this equipment.
Marc (00:06:28) - And the equipment is just a conveyor belt, right? It's basically moving boxes along paths, they need to stop somewhere and someone, in this case a, a picker needs to put ingredients, parts of the recipes and the recipe cards in the box to prepare them and do so without damaging at all. And that turns out to be extremely interesting problem from a routing perspective. Imagine all the vans you see outside or the scooters moving goods around, but in a more stylized fashion inside of a warehouse and just making sure these boxes stop at the right place. It's a big video game. It's a big video game and you need to make sure it works.
Ariana (00:07:05) - This is where our experience is crossed here cause I came in food manufacturing and supply chain world too. Yeah, I would say we were in 20th century and you were straight up in the 21st century. So when I heard about being able to look at operations through this lens and got any exposure to it, I thought oh my god, we could have done a thousand things better had I had this aperture to look at problems through. We talk about operations, my definition is operations is the glue between people, process and technology. And oftentimes operations is approached through people and process, but once you get that technology piece, you can just take it to a whole other level. And when I found out your experience in the same sort of industry but in a very different application, that's where I saw the power of how technology and data can really change the game and make you solve problems in such a smarter way, which is then interesting cuz then from there your next leap was into technology education, which is where we met. What made you springboard into that world?
Marc (00:08:11) - I do believe that you find things that energize you. So I'd found that I like working in this factory environment that's confirmed. Yeah, I found that from a theory perspective, I like the idea of operations, but then all through this process, I also really like teaching. This was just one thing that I kind of enjoyed even before my studies. So I was looking for a role that would combine my background in data science, potentially exploring new business areas that I hadn't seen before. Cause I'd only worked at this one startup, so I'd only seen what business looks like from the point of view of a scale up and you know, having to figure out everything on the fly. So I was kind of interested what two bigger companies work, what do they look like, and the education part. And so this came onto my radar through a former colleague whose, whose partner introduced me.
Marc (00:09:00) - I was looking at various different options, but this one just felt such a good match of both what energizes me and what I can bring to the table as you try and build, build a career. And then finding something that does a bit of both, it makes you excited, energize you, but also gives you a way to contribute, add something to the table early on. Felt like a good, good place to go. So yeah, technology education's a great way to learn and to really make sure and convince yourself that you are really, really on point when it comes to technology or, or the topic that you're teaching about. And I think we had a great opportunity to do this over the course of a couple of years to really dive deep on a number of topics and connect them out to whatever else is going on.
Marc (00:09:46) - I think that's the interesting part. Education doesn't live in a void these especially as you step outside of academic education, which I was used to. This is education in context, very specifically inside of a company, inside of an industry. And so, you know, try and talk about data but not about blockchain or not about the dark web. It's hard actually. Yeah. When you put everything into context, especially if you're speaking at such, well not in our case, at, at sort of senior leadership level, those topics are connected. They are important. And so all of that sounded really exciting and that proved to be true. So.
Ariana (00:10:27) - Yeah, it was an interesting experience within education because the lens of teaching data science, any emerging technology that essentially we had to have knowledge on was entirely through the lens of business and entirely for audiences that didn't understand these concepts or maybe had known 5% of what it means, but they don't know what the full scope of it is. So you have to come into a business and say, all right, here's what it is, here's how it's affecting your industry in your company. Here's how you can think through these things. That muscle at first was really hard for me. Trying to be able to articulate something perfectly in a way that anyone could understand was really hard. But it became an important north star with everything I did after, meaning now in operations, anytime I approach anything, a tool that I want us to use, a way in working, I know how to distill the vision into the reality for the people that are going to be using these things. I don't think there's anything more valuable than that, at least in my experience. Do you feel similarly?
Marc (00:11:31) - Yeah, I know what you're saying. It wasn't only about education in the sense that we prepared a presentation. It's about yeah, digesting ideas, all the work, all the, the internal feedback and spending time consolidating ideas and digesting it and working through it again to the point where, you know, it's gonna land first time around. Yeah, I'm with you. I've yet to sort of be able to take all I've learned and apply it in this new context I'm in, but it will come. I'm curious when that will happen. But I do feel like you do so many hours and then you build up that muscle. But it, it does atrophy a little bit.
Ariana (00:12:10) - It does, it does. I feel like with anything, if you don't use it, you lose it. And then it's little technical skills. It's in teaching skills, it's in all of those things. So after technology education, you've gone now to Amazon, you're working in a senior data scientists role. So if we can zoom out to data scientists in general, how does data science influence operations? What is the relationship there? How do, how do they work together?
Marc (00:12:38) - So I like your triad of people technology and what did you say? Process, right? Yeah. So obviously I've been kind of primarily in the intersection between process and technology when it comes to the physical operation side. But I do also recognize from a bigger point of view, how lots of companies are starting to look at how data science influences the workforce and people and their performance. There are data scientists that now embed themselves within technology teams. I was very close to that in my first role where initially I was a sort of in a separated data science team and then eventually we changed the operating model to make data scientists embedded in engineering teams or engineers embedded in data science teams, depending on what the so project was. But you kind of, that's how you see data science also influencing engineering and technology.
Marc (00:13:31) - How do you build a piece of technology knowing that you might ingest a piece of data today, but maybe tomorrow you need to ingest a lot more. Or how do you build a piece of technology that isn't deterministic, that's a, that's a little bit of a, a technical word, but normal code is fully logical. If you give it a certain input, it will always give the same output. In the world of data science, that's not always the case because you're making predictions based on data that model the models you're building aren't always correct. So you have to deal with this inherent randomness and if you build software around it, that software also needs to deal with it. And so it's a kind of shift of paradigm for a lot of folks who've been in development for a long time to consider that and not call it an edge case because it's by design.
Marc (00:14:21) - That's a big, big shift in thinking. So you've got influence on HR, you've got influence on technology, data science also then also of course has a lot of influence on process. You know, things happen, from a few different lenses. The way we of often look at it, you do what worked best yesterday and try and tweak it over time. Overtime ou get better and you make real gains as you go along. But in the world, data science, now two things happen. Suddenly you can look at it through the optimization lens. So you can look at your problem trying to distill it down to its, you know, important factors and then start influencing those factors. And you can use predictive technology to see not only how did things work up to yesterday, but given what we know about the past, what might happen to our process tomorrow or beyond, right? Yeah. And so the combination of optimization and prediction are changing the game for process. Cause then the process isn't so static.
Ariana (00:15:23) - So in operations and through your particular lens of data science, how do you approach solving a problem? What are you looking at first and how do you even begin to approach some of these very tangly problems that you're navigating?
Marc (00:15:37) - Well, I think before I jump in there, I want to separate two types of problems out because operations, or at least you know, the day-to-day of operations has a lot of firefighting as part of it.
Ariana (00:15:49) - Definitely.
Marc (00:15:51) - So you could call those problems, but let, let's put those aside because I, they're not, they're often not the best to attack with data science just because data science takes time, right? So I, yes, there are problems and some of them you could maybe solve with data science and you can do a lot on the spot. But the, the majority of problems that I would work on, that I work on are longer form problems that are open questions, unanswered questions in a business, how do things work, why do things work a particular way? And these are often tangly because they involve a complex system.
Ariana (00:16:24) - Can you give us an example of a question that comes at you or a problem that comes at you as a data scientist?
Marc (00:16:30) - Yeah. Most data scientists will go to the customer side and a lot of the problems people look at early on in their career will be something to do with marketing attribution. How do we get customers to the door? Why do they come through the door? That's very complex because people have different intent. Do they wanna shop? Do they just wanna compare? Imagine yourself walking into a shop, it's not always, you're not always going directly for the rack or the shelf where something is, you have myriad reasons to enter a store. Maybe you're just browsing. And so that, that's one of the first thorny problems I looked at that took a long while to try and unpack. On the upside, think about, in my case, a fulfillment center has just got tons of stuff going on. You've got machinery working, you've got people working in that space, in that environment, you've got different orders coming through with different aspects that are complex and then there's just incidents and random things that change the context all the time, which is what the firefighting comes from, right? Oh no, someone dropped a banana on the floor. Do you wanna keep that banana there? No, because it's tricky. People might slip.
Ariana (00:17:37) - I mean, but the comedic effect my friend, that would be a reason to keep it.
Marc (00:17:42) - You need a write a model to predict when is an next banana gonna fall on the floor. But some of that kind of comes together. So I don't know if that kind of answers your question, but the world I'm in are often problems that involve many moving parts, whether they are technological moving parts or they're people in the process and they altogether make for a very dynamic space where we just don't have perfect levers. You don't have perfect insight. You need to do something that works well in the balance. And so a lot of what I've worked on doesn't just try and, you know, maximize one thing and tries to kind of balance across multiple objectives.
Ariana (00:18:25) - Is there anything you're specifically looking for outside of data from human element or a process element? How do those components become a part of solving problems?
Marc (00:18:36) - Yeah, that's a really good question. This is what got me from being a standalone data scientist to working in a full operations team and recognizing where my work crosses over. As I said at the start, the thing that gives me a kick is building an environment where someone else can interact with it. And so the first question to ask is the “so what” question? So imagine a world where you did have a full insider full prediction and you're a hundred percent accurate. What could you do? Do you have any actions you can take? That action space has, has to be something meaningful. The, you know, the action space cannot be oh, that we can then determine whether we invest another X million. That's not the action space. We're, we're talking a little bit closer to the day-to-day. Can we make a meaningful change in scheduling? Can we give people more insight into what's going on on the floor? Can we do something different with the layout of an environment that people work in? Those are of actionable things that don't break the bank. There are real meaningful actions that we can take after.
Ariana (00:19:42) - That is such a good point. But you don't wanna predefine what the solution could be, right? But you wanna say, is this gonna have a large level impact? Is the time investment here actually gonna be worth it cuz we can action it? Is this more of a just an insight and a passion problem that we wanna solve? Or if we solve this, this makes the, the day-to-day lives of our team members better? That is definitely part of how you assess whether or not to move forward with the problem.
Marc (00:20:08) - I think that's specifically in operation. There's a lot of data science that is meant to explain and give insight, but in operations specifically, you wanna make sure that you then have, you can connect it to a, a lever, a button, a dial, something that, that gives you better control over the environment that you're wanting to control.
Ariana (00:20:28) - This gets to my next question, which is, what do you think is the hardest part of operations?
Marc (00:20:34) - Oh, well, first thing that comes to mind is timeliness doesn't go away, right? So you're in this spot, you've made something better, but chances are the world just shifts underneath your feet or whatever you just changed in one part of the system affects other stuff again, and it, it kind of falls apart or whatever you've built doesn't stick around long enough, it doesn't get the right adoption. That’s a real thing and ties back to the education piece. And those two dynamics may play into each other, so that could be disheartening. But for real ops nerds, I'm sure know many of them are nerds among your audience are like that. But that's the fun part of it, you know, the fact is something new will come up and we'll have to reconsider. So it's the hardest part. And also at the same time, the most fun part.
Ariana (00:21:26) - So I was a creative writing minor in college, and I'll never forget a teacher said, sometimes you have to kill your darlings, meaning you love these certain things that you've written or, you know, you love these certain things that you create, but if it's not working, sometimes you just have to let that go. And that is a very hard part of operations once you've built something to have that ownership, but also have the maturity to say, eh, let's, let's move on from this. Okay, if someone was wanting to get into operations and specifically within data science, science, where would they begin? Where would you recommend them starting?
Marc (00:22:02) - Find a company that doesn't just do digital work. There's a lot of companies today that, that just do digital work, right? And so if you find yourself in an industry that's primarily digital, chances are you're gonna find yourself not seeing the breadth of operations opportunities that are out there. A lot of the opportunity are in physical space. That means physical product, physical environments, you know, getting people into different faces. Think about that as you look for roles and then make sure that there is a data science team that you can join or that there's appetite to do something in that space and connect the two dots, it's very likely you'll be the first. There's not that many people who are so focused on operations and data science. Outside of the commercial world, I do see a lot more of this in sort of defense and at the more governmental level because they solve those problems more often. They move a lot of equipment, they need a lot of stuff, a lot of supply chain theory and supply chain management comes from that. But yeah, so look, look for those companies that have a physical aspect to them, maybe they influence the supply chain, like car sharing, apps you use for grocery delivery that you might interact with, or other things that involve getting something to your doorstep or getting you to somewhere are all very much intertwined physical world. So those are good places to go.
Ariana (00:23:34) - That's a great tip, I'm gonna take that tip and run with it. For, for those people that are listening who are either decision makers at businesses, what tools would you recommend investing in?
Marc (00:23:46) - If it's specific to data science, the progression is quite clear to a lot of people. You need to be able to collect and organize your data first and foremost. Today, that doesn't necessarily mean you need to build your whole data stack yourself. There are a lot of good solutions out there that take care of a lot of this. That gets you in into place where you can actually look at your data and ask questions but it's not a tomorrow kind of thing. It could be a couple years worth of investment if you're not there or said differently. If you have a system in place, but it's not meeting the challenge because it's an older system, a legacy system, then you know, again, that's a long investment. The next set of tools is all just about enabling your data champions. So the first bit is getting the data in place, and then you have to recognize what types of data champions you have.
Marc (00:24:34) - Some of them are pure data scientists, some of them are analysts and some of them just interact with data, but they need to have their own sort of level of access and ownership. Once you get to that point, you're, you're gonna start asking questions around what do we wanna do when we wanna do very serious deployment? And so then it's a choice of cloud providers. Right now, they offer all different tools that you might want. You can really build, you can really scale out and the sky's the limit, but not every company needs to get there from the get go. Not every person who works with data needs to get there. So it's a, it's a journey of different tools.
Ariana (00:25:09) - How are you defining data champion?
Marc (00:25:11) - It’s a good question. What I've seen before is a data champion is someone at their level that can help other people get access to information and also champions the idea of data. Data champions would be more integrated with teams and be on the spot to try and show them examples of how they might use it. And then there's data champions at the very other end of the company at the leadership level that hopefully in your best case, that's the senior leadership team or the board of directors themselves. So data champions come in different shapes and forms and interests, but it's important to try and ask the question, do we have any? Because if you don't have any, that means either your entire company is all on board, which is very unlikely, or very few people are, and you're not gonna move forward.
Ariana (00:25:56) - All right, we're gonna wrap up with some rapid questions. You ready?
Marc (00:26:01) - Yes.
Ariana (00:26:02) - Okay. What morning rituals do you start your day with?
Marc (00:26:09) - I write notes on paper, so I check my notebook. I've had these notebooks for the last so many years. I keep them all. So I look at my last few pages with some coffee. Iced coffee seasons ending, but I'll have an iced coffee today. But yeah, my morning ritual, take it slow, get your brain fired up. I'm not the quickest to fire on my brain in the morning, so I need some time to sit and think and then get rolling.
Ariana (00:26:38) - Now, on the flip side, how do you wind down at the end of the day? What rituals or what things do you always do?
Marc (00:26:44) - I try and go for a run. So I'm not a morning workout person, it makes me nauseous all day. But for me, running, if I have a chance to do it or if I feel energized it's a meditative thing for so many minutes is kind of like, I'm still churning through all the thoughts and then eventually that dies down and it either just goes away or it's organized. And so that's what the run does for me. I end my run and then I can end my day. So that's a good end of day activity.
Ariana (00:27:17) - What are you currently reading? What's, what's on the docket?
Marc (00:27:20) - On the docket? I am reading Edith Hamilton's mythology. It’s a paperback we had, I think it's from the eighties or nineties, brought it on a trip to Europe. And I thought what better way of having this really old beaten up book to read about Greek mythology. I, you know, you kind of know a lot of these stories if you happen to grow up in Europe and go through this education. I've been fortunate enough to have, but I've never seen them explained and put next to each other. So that's what I'm reading. It's really fun. And so now I'm on this kind of, in this rabbit hole and I wanna learn more.
Ariana (00:27:57) - What's your most important lesson you've learned in your journey so far?
Marc (00:28:05) - Find something that energizes you. Find something that you bring to the table and marry the two up the best you can. And then at least for work, it's great. It turns out it's actually a really good formula for a lot of other things too. Relationships, you know, as what you wanna study, what you want to create, do your hobby. So everything right? It's, actually really good. Those two things. It's neither one of the other by itself. It's kind of both. And then you feel, you feel like you're doing the right thing.
Ariana (00:28:36) - Last one, what do you wanna be when you grow up?
Marc (00:28:39) - ? God, you know, I think I still wanna be an astronaut but that's what I always wanted. I still want, I have this interest and passion for all things space. But I have a fear of heights. I don't particularly like flying. I really don’t like rollercoasters. I get nauseous easy. So, you know, all aspects of this job are not fit from my physical being. So, you know, I'll be an armchair astronaut. That's, I'll give you that.
Ariana (00:29:11) - Last, but not least, where can people find you?
Marc (00:29:14) - They can find me on LinkedIn. Send me a message. I think you should be able to
Ariana (00:29:17) - Marc, truly. Thank you so much for being the first guest of Secret Ops. To the listeners out there, thank you so much for listening to Secret Ops. Please follow us wherever you find your podcasts and check us out secret-tops.com. We'll see you next time.