MLOps Zoomcamp: Free MLOps course. Register here!


Business Skills for Data Professionals

Season 12, episode 2 of the DataTalks.Club podcast with Loris Marini

Did you like this episode? Check other episodes of the podcast, and register for new events.


Alexey: This week we'll talk about business skills for data professionals. We have a special guest today, Loris. Loris is the CEO and founder of Discovering Data, where he's on a mission to build a bridge between business leaders and data leaders. Loris hosts the Discovering Data Podcast, a show for business leaders and data professionals who want to turn data into outcomes. Welcome Loris! (1:42)

Loris: Yeah, absolutely. Thank you. Thank you for having me here. It's crazy to be on the other side. I watched the show recently as a listener and every time you're invited, there's this weird Inception thing, where you don't know what's actually happening. It's super keen to be here, so thank you. (2:07)

Loris’ background

Alexey: Yeah, it's always an interesting experience to be on the other side – not to host but to be a guest. Before we go into our main topic of building skills for data professionals, let's start with your background. Can you tell us about your career journey so far? (2:26)

Loris: Yeah, sure. In a way, it’s very typical for data people. I started out with a background in engineering and then I got passionate about physics. I thought for a moment I would one day become a researcher – a professor in quantum physics – then I got to understand how academia works and I realized that wasn't quite what I wanted. (2:45)

Loris: The first job that I landed after my PhD in quantum physics was data science, which aligned really well with a lot of the work that I've done in engineering as well. My background is Information Engineering, so it's a lot about entropy, mutual information, statistics – a lot of statistics – and data science, as we know, is founded, really, on statistics. I managed to convince someone that I could have done the job and they hired me. That was the first data hire in that company – a startup in marketing. So – very scary. (2:45)

Alexey: That’s amazing. Somebody without experience as a first hire in data science, that's a success, right? [chuckles] (3:53)

Loris: Exactly. Typical 2018, right? That was the thing at the time. If you didn't have a data scientist, you were behind – that was the business perception. So the chief executive said, “Okay, we need to hire a data scientist,” and started asking around to everybody, “Do you know a good data scientist?” And a friend of mine said, “Well, I know someone that is pretty smart.” I had that classic conversation, sitting at a table across from the other side was the Chief Technical and Chief Executive and they were like, “Okay, so we have a bold strategy and lots and lots of complex business questions to answer. Now you’re clearly smart. You’ve done a PhD in physics, so you work it out. This is your desk. He's your MacBook Pro. I'll see you in the stand-up tomorrow.” (4:01)

Alexey: That's the entire interview process? Or did they actually ask something? (4:47)

Loris: That was my day one. The interview was actually very interesting. We can talk about the interview, but day one was that. On day two, the CTO looked at me from across the room during stand-up and asked me, “Okay, so when is the project going to get done?” [chuckles] And I'm like, “Okay… things are fast here.” So that was a bit of a shock coming from academia. But that's where I learned the ropes and ended up doing a lot of data management and data engineering, just to get the data that I needed [chuckles] and just to avoid using Excel. (4:51)

Loris: At the time, there was a very compelling business need – they wanted to create some models and integrate them into the app. So it was exciting. I couldn't believe it. I was like, “Wow, we get to see the whole workflow from the source data all the way to the application of the user and get some feedback and iterate.” It took us nine months to deploy the first model into production at the time, which for me was forever. Then I learned how hard it was for data scientists to put anything into production at that time. So I felt, “Well, maybe I was actually lucky. Maybe it was actually an achievement to manage to deploy a model in real life.” (4:51)

Alexey: What was the app actually doing and what was the model? (6:14)

Loris: Oh, the app was just recommending. Autopilot is a SaaS business. Now they are rebranded, so you wouldn't find it on LinkedIn. But what they were doing was marketing automation. You could create journeys, and tag people, and make the whole workflow for someone in marketing a lot easier and more intuitive. The platform was incredibly flexible. We had this drag-and-drop functionality with blocks. There were actions and triggers sort of similar to ConvertKit, for those that have an experience with that software, but way more powerful. You could do literally everything, and it would natively integrate with everything else. [chuckles] (6:18)

Loris: They had a lot of engineers. I think more than 30% of the engineering team was focused on integration. The product was really, really cool. My job was to make the dashboard and the reporting way more actionable. Because the problem was information overwhelm – you will log in, and you will find logs, real-time information on what was happening across all your campaigns and across all your contexts and leads. But it was very hard to focus the attention on “Okay, here's what you can do right now to improve the campaigns.” (6:18)

Loris: It was analytics – it was reporting and sort of basic descriptive work, and there was also some diagnostic work on top. The idea then was to leverage the human intellect to cover the last part, which was prediction and prescription. So a bit of anomaly detection stuff – very simple from an algorithmic standpoint, but very convoluted from an architectural standpoint. [chuckles] (6:18)

Transitioning from physics to data

Alexey: Interesting. Your background was engineering and physics, yet the work (the job, the app that you worked on) was in marketing. This is like two different worlds. How did you actually convince them to hire you? Was it enough to say, “Hey, I know physics. I'm smart.”? (8:11)

Loris: [cross-talk] Well, they were a tech startup, right? So very technical people. It was very easy to convince them, to be honest. It was one of the easiest interview processes. We started talking and I described some of the projects that I worked on. In academia, in the context of university, I’ve actually done some really interesting work that never saw the light of the wider community it was stuck in – published a paper, it ended up on the IEEE, and that was it. But if you unpack that, that was a work on reinforcement learning. (8:30)

Loris: I was looking at non-polynomial hard problems for those in the audience that love mathematics, which I think are most of them. I used reinforcement learning to create basically a game where a bunch of actors would take actions, and their actions would compete and cause other actors to feel the effects of these actions until the network eventually converges to a non-optimal, but very close to optimal solution in like record time – in like 30 or 40 iterations, compared to being an intractable problem, if you want to approach it exhaustively. So that was cool. And then I just happened to meet this person. I wanted to build a laser lab in quantum physics. I switched from whatever I was doing when I saw lasers – a shiny thing – I couldn’t resist. [chuckles] Like, “Let's do it.” [laughs] (8:30)

Alexey: So previously, in academia, your work was published in papers on IEEE and after we published, the work was done, but then you joined a company, which was a completely different domain – marketing. And to deploy something, it took nine months, which was very unusual for you. Right? (10:05)

Loris: Yeah. I was expecting less. The basics of the model were ready to be tested on a larger sample than what I had available and were ready to do some basic A/B testing after a month from when I started. Because I had that good friend of mine in engineering, I said, [chuckles] “Hey, I need some data. Beer’s on me, launch is on me – help me out.” And so he did. He did help me out and gave me a dataset and I built the initial models off of that. The actual integration to live, that was a beautiful exercise of learning how DevOps works, and how the world of software development works, which is very different from statistics and data science, as we know. (10:23)

Loris: So that was exposure end-to-end – from the user needs to the engineer that needs to deploy a server in AWS to make that thing happen. Then eventually, that engineer was too busy, so I took over that part. I became very quickly more of a data engineer than a data scientist. Luckily, I knew the basics of coding, so I managed to survive. Eventually, we published that stuff into production, which was great. And then I started developing this passion around, aligning people on concepts and trying to really understand what they're trying to achieve and map those requirements into data requirements. I found that part extremely refreshing and energizing, so I focused on that. That was a lot of fun. [cross-talk] (10:23)

Aligning people on concepts

Alexey: What do you actually mean by “aligning people on concepts”? This entire thing – maybe we can unpack it a little bit. (12:11)

Loris: Yeah, sure. It's also one of the points in our outline for the talk today. “Data that makes sense – context, semantics, and meaning.” I guess I found the first hints that this was a problem when we're having a conversation after a hackathon. We did a hackathon one day – it was fun, everybody was free to spend one day solving a business problem they were in love with. I found myself contributing to a customer success problem. We did that and at the end of the day, we were drinking beers [chuckles] and this friend of mine said something that completely shocked me. I remember, essentially, freezing and staring at her in the eyes. I couldn't believe what I heard. I was like, “Is this what you mean? Or is that what you mean?” (12:19)

Loris: I realized that in that moment, we had two completely different understandings of what we did for eight hours together. She built a picture in her head that was different from mine and they were both very valid pictures – just not the same picture. [chuckles] And I thought, “Oh, okay. Interesting. How many ways can we have different views of the world, even if we use the same words – the concepts that we associate with those words can be different if our backgrounds and our priorities are different.” She was in customer success, while I was more in data science and data engineering. [cross-talk] (12:19)

Alexey: Customer success is like customer support? What is customer success? (13:57)

Loris: No. The customer success department was more focused on some of the 5% of customers by revenue and try to get them to grow even more. (14:04)

Alexey: Account management, right? (14:15)

Loris: Yeah, but more focused on adoption and proficiency of the product. Because even the big ones were actually not using the product – just scratching the surface of what the functionality was that we had available to them. (14:17)

Loris: My work there was to analyze the usage of the product. I was scanning the metadata (the logs) that came from the Elasticsearch background that we had and tried to describe the usage. Before I could even describe usage, I had to describe what “good usage” meant. So we ended up doing the session and whiteboarding “What is success for the product? What are the key metrics? Is it the number of elements they use? Is it the complexity of the graph that they build? Is it the number of tags or advanced features? Is the ability to export that information with other things (integration and ecosystem building)?” (14:17)

Loris: That potentially could have been a lead indicator for stickiness. You could then argue that if people use the product and it's embedded (they use a lot of integrations) then they're less likely to churn, because it's part of the ecosystem. It’s going to cost them time and energy to kill it. So that will tie in with the lifetime value of the customer and all the business metrics that the business was following. So context, the semantics and meaning – yeah. (14:17)

Lead indicators and stickiness

Alexey: [chuckles] You said something that is quite interesting and I wanted to ask you about that. You said that “It's a lead indicator of stickiness.” So what is a “lead indicator”? And what is “stickiness”? (15:46)

Loris: Yeah, good point. [chuckles] Stickiness is not the stuff you find on scotch tape – it's more, in the context of a SaaS, it’s the probability of churning, which essentially means canceling the account and picking a competitor. (15:58)

Alexey: So just like the opposite of churning? Right? (16:16)

Loris: The opposite of churning, yeah. If you have high stickiness, people tend to keep using the product. [chuckles] And so their lifetime value increases. (16:19)

Alexey: And “lead indicator” is some actions or some numbers (a figure) that tells you that they're about to leave or they are more likely to continue using the product, right? (16:30)

Loris: Yeah, that's right. I don't know how many people in the audience have ever seen a causal graph, but you can think about causality and a causal graph like a connection of nodes with arrows. The arrows typically go only one way. Could be that way [points in one direction], or this way [points in another direction], but one way only. And the causal graph tells you, “Okay, if this happens, then this happens and then that happens”. It's about establishing a direction on that arrow. (16:42)

Loris: A lead indicator is an event or some conditions that are very likely to lead to a transition to the next stage. In that case it was churn, but it could be anything else. For example, in my relationship with my wife, I know that my ability to keep a strong separation between my professional work and my home work is a lead indicator of how well we go along together [laughs] and how strong and healthy our marriage is. I hope that answers that. (16:42)

Context, semantics, and meaning

Alexey: [chuckle] Yeah, it does. I interrupted you, sorry. You were going to say something about context, semantics, and meaning, I guess. (17:50)

Loris: Yeah. I learned over time that this applies a lot more with cross-functional teams and larger organizations than it does in startups. The fundamental mechanics of it are just cognitive biases and just the cognitive structure – the way that we process information as humans – which is something that always fascinated me. I never had the opportunity to dive deeper into the topic. (18:00)

Loris: We see a lot of the data information knowledge wisdom pyramid, which is one way of understanding the different levels in data. But I recently learned to extend that pyramid, in two interviews in particular – one with Jessica Talisman, on my podcast (the podcast title is True Data Mesh, semantics, ontologies and then a true data mesh) and the other one with Ron Itelman. Ron has done a lot of work at the intersection of machines and humans, so he loves that machine-human interface. And it's just that multidisciplinary sort of stuff that really gets me out of bed in the morning – studying cognition, psychology, and how we make sense of the world, essentially. And it's so messy, because everybody has a different way of processing information and building pictures of what reality is. (18:00)

Loris: So how do you align people on the same understanding – on the same thing? Very practically, an example is – the definition of customer. It's the most used in the data management space. You ask someone in sales and have one conceptual understanding of what a customer is, then you ask marketing and they have a different one. The problem is that these different understandings then reflect in the data and how people use the data to do what they do. (18:00)

Loris: As data scientists and data analysts, if we want to support the business, to achieve the targets that are set up as part of the strategy, we need to understand that different people might have different understandings of what we see as one metric – customer, how hard can it possibly be? There’s one example with Scott Taylor, where he tells me the story of the hospital with 170 definitions of what a patient is. It's like, “How can you possibly have 170 definitions of what a patient is?” (18:00)

Alexey: How did they even count this? (20:41)

Loris: At some points, things got broken, someone suffered – potentially physical suffering, not just psychological suffering – and they must have gotten to a stage where they engaged with an external firm to do some audits, and they looked into the databases and accounts. They counted 170 definitions. It doesn't surprise me. So how do we use data to actually have an impact on the organization when we know that our cognition – the very system we use to make sense of the world – is so unpredictable, and so random, and so different? [chuckles] That fascinates me. (20:43)

Alexey: I guess another good example would be the definitional of churn, right? You can define churn or “stickiness” in so many ways. Everything you do, when you need to analyze data on churn, depends on this definition. If you use a different definition, then your analysis will be different. (21:23)

Communication and being memorable

Loris: Yeah. In this one – I think this could actually be a fantastic introduction to the next point that we have in our outline. There’s this one point that I wrote around communicating – building better models by building relationships and earning trust, which connects really, really well with communication and storytelling. It's something that (I'm really curious to hear from the audience about this) I didn't even conceive of at the beginning of my career. For me, marketing was sort of the enemy. [cross-talk] (21:46)

Alexey: Coming from physics, right? And reinforced learning. (22:33)

Loris: Yeah! From physics, man. This is like science and hard facts and numbers. Everything has a standard deviation, everything has a mean – everything has a model that can describe it. And if you don't have a model, I can sample it and I can estimate the model. This theory and the books were written many decades ago. To me, that was solid, verifiable, reproducible thinking. With marketing, I always saw it as magic. I also associated persuasion as something very bad. You don't want to be the one that persuades people, because it sounds like you’re manipulating them. It sounds like you’re getting in their heads, right? [Alexey agrees] (22:37)

Loris: You want to stick to the facts. You want to show them the numbers. And I changed my mind. I changed my mind big time on that topic. In fact, I think as data professionals, we don't do that nearly enough. If you think about it, we're competing for the same amount of funding. There's only so much money that the organization has to invest in projects and we're competing against each other – the sales folks, the marketing folks, they know how to tell a story. They know how to be memorable. They leverage the way that our brain works, the way that we process information – not necessarily to lead the organization in the wrong way – just to have that edge. Just to be remembered, really. (22:37)

Loris: In data, we just blather a lot about databases and models and we forget that we're talking to people that don't know what we're talking about. They really don't. They really don't know what you're talking about [chuckles] with some exceptions. I mean, think about data maturity scales – zero to five – zero being all paper, where we barely use a laptop and 0.5 being the majority of organizations. Maybe I'm being too harsh. I think Doug Laney put it around 2 being like 45% of the organizations that they use for statistics, back when he was a gardener – between two and three in the data maturity space. (22:37)

Loris: I think two is when you do have digital systems, but you don't really have a strategy. You don't really use data. You talk about data, but the data is in excel (in a spreadsheet) there has no heading, no metadata in the title is “_v1” or “_final_final”. That's kind of level 2. So it's still pretty bad. [laughs] Level 3 is when you start having a bit of a function – you have a data team, maybe an analyst or two. And level 4 is when you start doing more predictive, and 5 is the prescriptive, but it's really embedded within the strategy of the organization. Most organizations don't really do that sort of stuff and they don't have that data literacy that’s high enough so that the business can understand what the heck we’re talking about. (22:37)

Making data digestible for business and building trust

Alexey: How can we make business understand what we're talking about? What can we learn from marketing to actually do that? (25:43)

Loris: Yes. For me, this has been the focus for the last year, to be honest, and it's gonna keep continuing being the focus in the next two or three years. I think the key is to unlearn some of the concepts and the beliefs that we built up over time, especially for those that have resonated with the story that I shared and the way that I felt about marketing, and learn to see it in a different light – as a tool. (25:53)

Loris: Like any tool, we know that if we want to do classification, we can hop on SciKit Learn and we have a whole plethora of systems that allow us to do classification. Just in the same way, if we want to have any impact in the organization, we need to use those tools. A tool is a tool, right? It's not bad, it's not evil, it’s not good by itself – it’s just a tool. It's how we use it that gives that tool the objective. Is it good? Is it bad? Are you doing it ethically? Are you doing it the wrong way? But we can’t just say, “Hey, because there are examples of people that are manipulating other people with marketing slogans, then I want to be a purist. I'm gonna just step back, walk away, won't even get my hands dirty.” (25:53)

Loris: I think that was the mindset that I had to overcome and learn how to be like them, but while retaining my intentions. My intentions as a data professional are to have any impact, to be honest. I hate the idea of working on things and just having zero adoption. That is my biggest nightmare when I do any data work. It just doesn't feel good for me. (25:53)

Alexey: How does somebody coming from a physics background – from a mathematical background – work on marketing? How do you unlearn these things? How do you start building trust with business people? (27:36)

Loris: Yeah, it's a fantastic question. I think it starts with active listening. But before you can listen, you have to learn to be comfortable with not knowing what the hell people are talking about. And it's incredibly frustrating. I have fresh memories, because I just started a new position two months ago. I'm what's called now an Industrial Engineer, but essentially, I'm an analyst. I'm trying to help the business reduce costs, increase margins, and reduce waste, which is what data people do. That’s what any business wants to do – one of those things, if not all of them at the same time. (27:55)

Loris: There's so much domain knowledge when you join a new company, that it's overwhelming. We also live in a moment of volatility, so I'm sure that a lot of the listeners have fresh memories as well, of what it’s like to start a new job in a new organization. And it's awful. We do it because we have to do it and we can't wait to go past that first steepest part of the learning curve, so that we become comfortable – we know where we are in the organization, we know who our stakeholders are. Stakeholder mapping is one of the first things you do in any new position. That gives us reference points, we build a kind of map of what we're doing, why we're doing it, who the people that we need to please are, who we need to be careful of, etc. It's basic survival, if you think about it. (27:55)

Loris: As a data professional, if you want to really have an impact, you have to get close to the business – “the business” meaning anyone in the organization that is trying to achieve something tangible and they're struggling to do it, or they have difficulties and they need help. So first, you need to know who they are, what the problems are, whether the problems are worth your time. So there's a whole prioritization of projects you can work on, because you’re only one person. As an individual contributor, at least – if you're leading a team it’s different, but the same principles apply. You still have to prioritize. That requires business literacy. That requires mumbo jumbo [chuckles] going to meetings where people talk about stuff that, to be honest, we’re not ready for. Most data professionals don't come from a business school – they don't have an MBA, they've studied technical stuff. The engineers… [cross-talk] (27:55)

Alexey: I don’t know anyone who actually has an MBA, at least those from the data world. [chuckles] (30:44)

Loris: Yeah. Because we don't need it, right? To get hired, you need to demonstrate that you have a prolific GitHub account and that you are active on Kaggle. You don't need to demonstrate business knowledge. But to have an impact – after you're hired – you need to know how a business operates. You need to know to speak that language. (30:49)

Alexey: I think I wasn't completely correct when saying I don't know anyone. It occurred to me that I do. But usually, the people who studied MBA, studied business – they don't start as junior data scientists. They start as head of product or head of data, right? (31:09)

Loris: Well [chuckles] I hope not. A head of data and data lead are typically really technical positions. I mean, it depends. (31:33)

The importance of understanding the language of business

Alexey: I mean, if you start with the head of product, this is a good place to… well, maybe a product manager and then you become head of product. Then you get into data because you've worked with analysts. So eventually, you might become head of data. In many organizations, actually, the head of data or director of data reports to the product function – to the CPO or whoever – which is fine, I guess. But most of us, we studied engineering, we studied computer science, economics and not business development. (31:39)

Loris: Yeah, absolutely. When we see that the business talks about targets on specific values that they need to increase – they talk about penetration, they talk about differentiation and humanization of the product. They talk about education that they are actively doing to turn customers (leads) that are not even aware that they have a problem and walk them through the five stages – from unaware to super fan. They talk about the cost of acquisition for marketing. They talk about conversion rates for a campaign. Every domain within the business has a language that is different from the other. (32:20)

Loris: And if you change industries, the language keeps changing. Plus, in addition to all of that, there is the lingo that is actually spoken by the people on the job, which may be filled with acronyms and inside jokes and the whole thing. So you kind of have to become one of them if you want to win their trust and make them understand that you are there to try and help, that you, by no means, have all the answers. Because these people may have been there for like 5, 7, 10 – in the case of the company that I’m in – 20, 25 years. I have four key stakeholders and the average time that they have been employed in the same organization is 20 years. So I walk in as the data person, and I'm like, “I have zero domain knowledge. I need to learn from you and demonstrate that I'm genuinely curious. I actually want to learn what it's like to be you at work. What keeps you busy? What worries you? What are your aspirations? What do you want to achieve? Personally, as a team, and for the business.” Hopefully those three are aligned, [chuckles] but in some cases, they're not. (32:20)

Loris: And only when you have that kind of welcome, then you see that people open up and they come to you with ideas. So you don't have to chase for case studies, you don't have to do the research – the research is done by establishing those relationships first. I think those are the foundations of successful data professionals and don't see a lot of that. (32:20)

Stakeholder mapping

Alexey: How do you actually do this? So you join a new company. Then you said that one of the first things is stakeholder mapping – for each stakeholder, you try to understand what they do and why they do this. You also want to prioritize and understand who is more important – like you said, who you need to please and who you need to stay away from. So how do you do this? How does it work? (34:48)

Loris: Okay. You have to know one thing about me and that's – I have a terrible memory. [chuckles] For a long time, one of the biggest challenges I had when it comes to knowing people – it's remembering their names. I'm incredibly bad with names to a point that is almost ridiculous. How can you build relationships? [Alexey concurs] Yes! Sometimes it’s like I have to avoid people, because they recognize me – they call me by name, and I, for the love of God, I don’t remember their name. It feels so bad, so I kind of tried to avoid it. [Alexey concurs] (35:20)

Loris: I've been stuck in that for years. Trust me. So one day, I learned about Notion. Now there are many systems, but Notion was one of the first, and when I learned it was the only one. I was like, “Oh, maybe I can build a system where every time I meet someone, I can Google them (everybody has LinkedIn) get a copy of their photo, attach it to the system and build a database with people.” I started out, actually. (35:20)

Alexey: Like a CRM, right? But for colleagues. (36:29)

Loris: For colleagues, yes. (36:33)

Alexey: That sounds very creepy. [laughs] (36:34)

Loris: I know, I know. I know it's creepy and it sounds really, really bad. But it helped me enormously. [chuckles] So if you ask tactically, practically, “How do I solve the biggest barrier to building relationships?” It’s remembering people's names, what they do, what they care about. I obviously don't take notes when I meet someone over coffee, like, “Okay, let me take notes about you.” No. we're humans – normal relationships. (36:36)

Loris: But as soon as that interaction ends, I know that I have to do it. I have to collect that information, that knowledge, and resurface it whenever I need it. Otherwise, it's lost. And that helps a lot. Because people – we're humans, right? We want to feel acknowledged, recognized, and respected. We want to feel that sense of belonging and connection. It's not manipulation, it’s simply just… normal people are born with a solid memory. My memory is crap. So I need a database. [laughs] (36:36)

Alexey: A second brain, right? [Loris agrees] This is your stakeholder mapping – or this is like a stakeholder relationship management system? (37:36)

Loris: No, this is just awareness of who is who and what their names are. (37:46)

Attending business meetings as a data professional

Alexey: But it's like the first step, right? (37:49)

Loris: The first step, yeah. This is just overcoming my terrible memory. Then, step two is to attend business meetings. There are many in any organization. In any organization – from startups to enterprises, people talk business every day, all day – many times a day. Some of those meetings are more engaging, some are boring, but you'll never know which ones are valuable until you sample them. I know it's crazy, right? I know it's absolutely crazy. Something you will never do. You join and there’s this “data” in your title. What do you do? You just go around and get random invitations to Teams and Google Meets to sit there and just look? It sounds like there is no meaning whatsoever. (37:51)

Loris: The first 10 meetings you go to, it's just a different language. It’s like when you go and visit a different country where they speak a different language like, “I don't know what you're talking about. There's no way I can understand the conversation.” And that's where a lot of people quit. They're like, “Ah, SciKit Learn is so much better. It gives me good vibes, I can build models. I don’t need that. This is bad for me.” Right? And I think that's where we need to go against our instincts. Because the instinct tells you “Play safe. Stay within the comfort zone and just get attached to those pleasant sensations that arise when you know that you're on top of your work.” When you know what you're doing. (37:51)

Loris: Nobody wants to feel like they're lost. But if you stick to it long enough, then those words start making sense. Now you start seeing patterns and connections. Repeat, rinse, repeat. [chuckles] You do it over and over and over. And after – sooner than you think, actually – within less than three months, I guarantee you that you will start seeing things from a completely different perspective. Now those people recognize you – you’ve built a habit of attending those meetings, and maybe you didn't ask a single question for a couple months. Now the first question comes to mind. You'll ask it maybe privately because nobody wants to ask a stupid question in front of like 40 people that may one day be your key stakeholders. (37:51)

Loris: At the beginning, you play safe, and then you build some confidence. Once you know what you know, and you kind of have a feeling for what you don't know, then you maybe dare to ask one question and challenge what someone said. I mean, it's just a wonderful thing, because most people are actually open to get fresh perspectives on problems – and you have it. (37:51)

Loris: That's the thing that we don't see: the value of someone that attends a meeting from a completely different domain or department is that you don't have priors. Your prior is a uniform distribution, to speak the language of the audience. You don’t have biases, you haven't done the shaping yet you. Everything is equally important and potentially interesting, or potentially not – just in the same way. And that is a superpower. If you can sell that, people recognize it, and they open up, and they talk to you. Now you have friends – you have new connections. [chuckles] (37:51)

Organizing your stakeholder map

Alexey: Do you have a system for that? You have this database with people to remember their names – do you also have a system for taking notes after these business meetings? So then you can remember, you can map concepts, you can understand who cares about which topic, and what’s important for people? (41:11)

Loris: Yeah, absolutely – something that you can do on day one. So I come from… I’ve been working on Discovering Data for two years now and the whole system is built in Notion, so I knew the software really well. But I found that for the first four weeks, I couldn't possibly think about it. That's just all pure overwhelm. There's so much information coming to you that I think the only thing you need to do is just chill – do some yoga, long showers, sleep – because there's a lot going on. (41:33)

Loris: But after week four, you start seeing some patterns, and you’re like, “Okay, now I kind of know what the business drivers are. We want to do X. We want to do Y. The net sales value has to be above that number. The margins on our products have to be above that figure. Cool. Now, what are the key activities that I'm going to focus on, to try and sustain and support that business strategy?” (41:33)

Loris: And your line managers know that, because if you have a good line manager – even an average one – if they’ve been in the business longer than a year, they surely know the key activities that are more likely to support that strategy. So you start from those priors – you don't have to invent anything. They will tell you, “Okay. Two low hanging fruits are one and two. We're going to start from those, nice and simple.” So I built a system literally based on those two first key activities. I try to map my actions and map them to the activities. I have notes. (41:33)

Loris: Notion has things like templates, so every time I have a meeting, I have a button – that button creates a very simple, very bare bones template with who I’m meeting with, the date is automatically captured, the title of the meeting, and tags that tell you which key activity that those notes may one day support. And each key activity is a page – you open it and you have all the notes that refer to the key activities. You can do it on a notebook (pen and paper) as well as. I just use Notion because it's easier. (41:33)

Alexey: I’m just wondering, do you use your private Notion account for that or do you use Notion at work? Is it a company account? The reason I'm asking this is because at my work, we don't use Notion. There is a different system called Conference. Some people hate it, some people love it, but this is what we use. I’m then wondering, if I want to implement this should I just use my Notion account? Or maybe I should use whatever the company has? (43:42)

Loris: It's a wonderful question. I started out creating an account specifically for work and then I realized that Notion is literally my second brain. My ideas, my knowledge, the things that I develop are all there, so it made sense to continue doing that. In saying that, it could be tricky depending on… I don't want to offer legal advice, but do read your contract and intellectual property clauses of their contract, because you might be in breach. So the way that I resolve this is by creating different access roles and just segregating. I have databases that are within that same workspace but that are specific, so that there is no accidental sharing with other collaborators from my business and stuff like that. Just something to be careful about. (44:17)

Alexey: Okay. (45:07)


Loris: Yeah. We talked about prioritizing. Prioritizing is a big one. The question that's interesting to me is, once you reach a steady state and you start being productive, you're part of the business, you attend meetings. Now you are independent. You can decide and propose even a use case or a project to the business. How do you prioritize those use cases? That's where I think you really make the transition from onboarding to leaving your function or being on top of the work that you do. I'm curious to hear from the audience, if anyone has a go-to framework to prioritize. What do you focus on? How do you make that call? Is it a dollar value impact? Is it the impact on people and how they work? Is long-term thinking stuff, like data management, cleaning data? Is it transitioning to a data warehouse? (45:13)

Alexey: The way I do it (maybe not the most scientific way) but it’s just by talking to people and understanding what they care about. What are their goals? What are the goals of their department? And they say, “Okay, this, this, and this is important to me.” And if I talk to multiple people and multiple people say that this item is important to them, then this means that this thing, this piece of work will make more people happy. It will contribute to whatever they are doing. And this is pretty simple, right? You don't think about money, which you probably should in the end, or some other impact, but instead it’s “How many stakeholders can this thing make happy?” I think this is a valid way of prioritizing too. (46:26)

Loris: Yeah. It makes me think of network and social media, when you find the nexus of a graph – it's a node that is highly connected. In this case, the nodes are data projects and the edges are how many people that data product essentially can touch. The difference is that those products don't exist yet. They're just in our head – in our imagination – so the job is to really understand where the business is going and find those opportunities. (47:10)

Loris: It's like marketing. When I title a podcast these days, I come up with like five or six different options for a title and I pick one – the other five or six are gone forever. That's where agility really comes in. It's not because we just learn about a particular algorithm and we find an opportunity to use it. A lot of people do that, to be honest. I've seen it over and over. Curriculum-driven development is a thing. People want to get better at specific things. They know that the average tenure in the industry is low. I’ve met people that were like, “Look, I'm just being realistic here. I'm not going to be in the job longer than a year and a half. So I'm just taking every opportunity I can to work on the stuff that I think is hot or is going to be hot – there's a big market for it – and then jump ship and negotiate a bigger salary.” That is one way of approaching the problem, right? I don't think that way. (47:10)

Loris: The community that I'm trying to build at Discovering Data (my podcast) is for people like that. Of course, I have a mortgage to pay. Of course, I want to negotiate a bigger salary. But what gets me out of bed in the morning is to see that I do work and that work has an impact on real people and I can put a dollar value figure on that. Look, I don't know if it's because it's nice to feel that someone comes in and says, “Hey, Loris. Good work!” I don't know, maybe I'm addicted to that feeling. Maybe I'm addicted to the brainstorming and the learnings. (47:10)

Alexey: It's very human to get addicted to people saying good things about you. [chuckles] (49:16)

Loris: Yeah. It's just nice. It makes you feel that you're consequential. That you're not just doing stuff for the sake of crunching numbers – you are actually having an impact on someone or some teams or maybe a whole domain in the organization. That makes me feel so much better. So for people like that. (49:23)

How to support the business strategy

Alexey: I just noticed that there is a comment from Alejandro that I wanted to acknowledge and read. He wrote “I have an MBA,” we talked about people with a business background starting in the data and he started from an individual contributor position again. “This shift is not easy, I'm still figuring out how to profit from that generalist business knowledge.” This is quite interesting. I wanted to ask you – I think most of the listeners are not coming from an MBA, but are coming from more of a quantitative background. How do we get this business knowledge? How do we start supporting business strategy? (49:46)

Alexey: One of the things we wanted to talk about is, “How can I sit with people who are making important decisions, strategic decisions? And how can I speak the same language?” So we already talked about stakeholder mapping, understanding what they care about, prioritizing things, and speaking the same language with them. Then what's next? How can I actually contribute my data knowledge to this model? How can I support these business decisions? How can I support the business strategy? (49:46)

Loris: I'm gonna give you an answer that is kind of intuitive. The answer is – that part is actually the easiest. [Alexey chuckles “Okay”] But! This is a huge “but” – it requires flexibility in the way we manage our identity. As data professionals, we identify ourselves with the depth and breadth of our technical knowledge. That is great, but it's also a barrier. The biggest transformation for me is to rethink who you are and decouple the identity – your identity as a professional – from the tools and the technologies that you use. The reality is that the 80/20 rule applies to anything (the Pareto principle). It applies to data projects as well. So there's the 20% that you can do that gets 80% of the value. (51:01)

Loris: Most of the time, for new case studies, these are things that are experimental – in Excel. And yet, we see on LinkedIn literal wars between people that sustain that Excel is bad and people that sustain that Excel is good. And I've been a victim of that. I took very sharp stances on that position and I said, “From a data management perspective, Excel is the enemy. It’s the reason why data is crap.” Yes, there's truth in that, but it's not one or the other. [cross-talk] (51:01)

Alexey: That's definitely true. [laughs] (52:40)

Loris: That's true, right? That’s true, but that doesn't mean that we need to go banning like, “I'm never touching Excel again. Give me a Python client or I'm outta here!” Because that blocks us from having those conversations. (52:42)

Alexey: You can do a lot of nice things quickly with Excel without even starting your Jupyter Notebook. You just open the CSV file with Excel and play with the data. And then [cross-talk] (52:54)

Loris: Pivot table. That's a superpower. To be honest, it's even better than Jupyter. How fast it is and how easily you can share it. We don't think about that, right? But the other side is, we're not dealing with monkeys, we're dealing with people that have brains. They want to understand what's happening and they even want to feel good about it. That collaboration – they just want to establish, or experiment, to see if it's worthwhile considering you as an ally, as a partner, to what they're trying to achieve. That's an experiment when you start. And you have to prove that it's worth it. (53:08)

Loris: What we don't think about is that we need to also worry about how these people feel. If I come to you with my Jupyter notebook and I showcase all my exploratory data analysis and that “package that does all those cool graphics!” Yeah, maybe there is a “Wow!” moment. But then the next thought, most of the time, is “Oh, jeez. He's so smart. That's a big brain and I'm behind. I don't even know how to launch a bloody Jupyter Notebook. I don't even know what it is.” And now all that work is inside your head, as a data professional. Instead of initiating a conversation, it becomes a monologue. That's the problem. It kills the conversation. And we want those conversations. The more conversations we have, the better we understand the domain and the higher the impact of the models that eventually will touch Python is going to be. We'll get there. (53:08)

Loris: We'll do predictive and prescriptive, but first we need to do description and we need to do diagnostic. We need to be able to answer the question “Why?” That is the most burning question for most businesses. People don't know why things happen. Yeah, Excel is not the only tool. You definitely can get to the description stage, and maybe with some use of your brain and your domain knowledge, you might even be able to do diagnostic without even touching machine learning. But are you going to be okay with that? That's why I said that it's going to be an identity clash. We need to let go of the fact that “You're cool if you use machine learning. You're boring if you use Excel.” That is nonsense. That is completely missing the point. The point for us is to support the business. If the business needs Excel at that moment, we need to be okay with that. We need to step up – grow up, essentially – and go like, “Okay. The tool doesn't identify me. I use whatever it takes to serve the business.” Are you ready to do that? That's the question. Not everybody is. (53:08)

Loris: [technical difficulties] (53:08)

Learning to speak online

Alexey: Yeah, now it works. Thanks. We have a few questions and one interesting one from Raphael is “Your sound is very good. How does one learn to speak online?” [chuckles] Is it only trying and or did you do something else for that? (56:13)

Loris: The sound as in the quality of the audio or the speaking? (56:32)

Alexey: I think it’s both. A microphone is important, right? [Loris agrees] But that's not the whole thing. For example [cross-talk] (56:36)

Loris: Well, I host a podcast [chuckles] (56:45)

Alexey: [laughs] Yeah. But how did you actually learn to speak online? Was it just practice? You thought, “Okay, let me host a podcast.” and then you just learn to speak online or you did something else for that? (56:47)

Loris: I did something that was atrocious. It was incredibly painful – listening to yourself after recording. It was one of the hardest things. (57:00)

Alexey: Ooh. No, no. [laughs] That’s terrible. (57:12)

Loris: I wanted to run away. Yeah. It's so weird. And people say “Yeah, you get used to it.” But yeah… I'm still working on it. I'm getting used to it. But when you edit your own podcasts, you realize how much you blather without really… You learn to speak in a way that is different, like varying the tone, taking the time to pause. A lot of us are actually afraid of silence. [cross-talk] (57:16)

Alexey: You’re very good at this. I’ve noticed that in this episode. The pauses were really nice. It's like you practiced. You did practice, right? (57:51)

Loris: No, I didn't. Well, I host a podcast. So after 55 interviews with people, you do find patterns in editing. I'm like, “Why…? I should breathe.” Sometimes that's what it comes from – from realizing how suffocating it is to just keep talking without breathing. From the listener perspective, when you take that pause… (58:01)

Alexey: They can breathe too? (58:25)

Loris: You allow them to breathe, yeah. It's so much easier to follow the podcast. I don't know if that works. (58:26)

Resource recommendations from Loris

Alexey: We should be wrapping up. So maybe the last question I will ask you is “Do you have any book recommendations or resource recommendations on the topic of today's interview? What kind of book can we read to learn more about business skills? Or maybe you can recommend your podcast for that and you don't need any books? (58:35)

Loris: Yeah, it's a good question. To be honest, I have been looking for a book like that for a while. I’m kind of dreaming of writing one myself, because I couldn't find a book on the topic that is connected – that speaks the language of data professionals. There's plenty of books on businesses, but you read them and you're like, “This is not for me.” So no, I'm sorry. I'm not aware of any books. I do have a podcast that people that are interested in the topic of getting more impact out of the work they do – they can definitely check it out. It's called Discovering Data ( (58:54)

Loris: We just launched a Discord server yesterday. Unlike DataTalks.Club, our community has myself and my podcast assistant at the moment on it. The server is open. [Alexey “Two people?”] Yeah, just two. So to whoever comes in, don't expect a polished system. [chuckles] But we've been running now for two years and we have so many people that reached out and said “It'd be nice to have a space to hang out and share thoughts – post episodes even before the episodes.” My vision for next year is to use that space to grow, to fill this gap that we all kind of know it's there. It's kind of like the Matrix, right? You’ve gotta take that red pill to follow the rabbit and see what there is. Because some people – most of us, really – don't even know that there is a gap. (58:54)

Alexey: To join the Discord server, what do I need to do? Do I need to go to the Discovering Data website and scroll down? (1:00:26)

Loris: in the future, yes. At the moment, we are a business of one. It's way simpler than that. You go on wherever you listen to your podcasts, you follow Discovering Data, the last episode – the one with Stephen Shedletztky, the guy who worked 10 years with Simon Sinek. He's an absolute authority on the space of psychological safety, leadership, that kind of stuff. We had a fantastic conversation. He's an incredible speaker. I absolutely recommend him, if you're looking for speakers for your podcast, by the way. In the show notes of that podcast, you'll find the link to the Discord server. So definitely join. Happy to see you there. Maybe we can bridge the gap together. It’s so much more fun. (1:00:33)

Alexey: Thanks for the chat. It was amazing. Nice to talk to you. And thanks, everyone, for joining. I did not expect that so many people would join so early. So thanks, everyone. That was fun. (1:01:23)

Loris: Thank you. Anytime. Hit me up on Discord or LinkedIn. Happy to connect with anyone in the audience. (1:01:36)

Alexey: Yes. We'll include all the contact details, and you perhaps will also send us a link to your Discord server, which we’ll also include in the description. (1:01:42)

Loris: I can definitely do that. (1:01:50)

Alexey: Okay, nice talking to you and have a great weekend. (1:01:53)

Loris: You too. Ciao. (1:01:57)

Subscribe to our weekly newsletter and join our Slack.
We'll keep you informed about our events, articles, courses, and everything else happening in the Club.

DataTalks.Club. Hosted on GitHub Pages. We use cookies.