00:00:04 Data scientist role in supply chains.
00:01:33 Comparing data mining and data science trends.
00:03:16 Marketing data science tools to universities.
00:04:14 Coding skills vs. creating business value.
00:06:37 Lokad’s shift to ‘supply chain scientists’.
00:08:01 Duties of a Supply Chain Scientist.
00:09:50 IT and Supply Chain Scientist’s shared duties.
00:11:58 Supply Chain Scientists and data extraction role.
00:14:19 Pitfalls of relying on statistical toolkits.
00:16:29 Data’s growing role in life.

Summary

Kieran Chandler and Joannes Vermorel are discussing the critical role of data scientists in supply chain management. Their conversation is highlighting the necessity for these professionals to extract value from business data, while warning against the trend of overemphasizing programming and statistical skills. Vermorel is pointing out the importance of practical knowledge and business acumen, cautioning against overconfidence in technical abilities. He is introducing the role of ‘supply chain scientists,’ who are tasked with extracting and interpreting data to address business problems, in contrast to IT roles focused on maintaining systems. Their dialogue underscores the challenges in university training, highlighting the scarcity of real-world supply chain data and the over-reliance on programming languages and statistical frameworks.

Extended Summary

The conversation between Kieran Chandler and Joannes Vermorel focuses on the expanding role and significance of data scientists in the commercial realm, specifically in the supply chain industry. Chandler underscores the mounting demand for data scientists, characterizing it as a phenomenon “sweeping across the industry like wildfire.” He observes that this demand, which wasn’t pronounced five years ago, evolves so rapidly now that it surpasses the capacity of universities to yield enough graduates to fill these positions.

Vermorel lends his viewpoint to this shift, observing businesses starting to appreciate the inherent worth in their data and the subsequent need for individuals with the skills to mine this value. Nonetheless, he also draws attention to the cyclical pattern of this trend, likening it to the “data mining” frenzy of the 90s. He speculates that the current fixation with data scientists is redolent of the past interest in “data miners”, leading him to dub data scientists as “data miner version 2.”

Vermorel discusses the rise and fall of data mining firms in the 90s, suggesting a cautionary narrative. He recalls the emergence of hundreds of companies during the data mining period, providing tools for data mining. Yet, most of these companies eventually vanished, casting doubt on the current surge of data science tool suppliers. Vermorel perceives a correlation between these two periods, hinting at a possible repeated pattern of ascendance and decline with the present data science trend.

Interestingly, Vermorel accentuates that his company, Lokad, avoids using the term “data scientist”, opting instead for “supply chain scientist”. This preference mirrors his conviction in the significance of comprehending the business context and value beyond merely mathematical and coding skills. He cautions that proficiency in technical facets, while crucial, doesn’t automatically translate into generating business value within supply chains.

Furthermore, Vermorel talks about the promotional strategy of present data science tool suppliers. He underscores their aggressive marketing towards universities, especially through open-source toolkits that dovetail with the general mindset of academia. Yet, he also provides a cautionary note: triumph in marketing a product within universities doesn’t ensure that the tool will produce beneficial results in tangible business settings.

Vermorel emphasizes the necessity for data scientists to not just analyze data but also bring about actual business change through their discoveries. The challenge is that, frequently, data scientists can examine and present findings, but grapple when it comes to implementing these alterations, as they may disrupt the status quo. It’s more than just conflicts on operational decisions; it’s the larger question of whether the data scientist is genuinely empowered to act and deliver value to the business.

Moreover, the discussion touches on the role of a ‘supply chain scientist,’ a term used at Lokad. The job of a supply chain scientist, according to Vermorel, involves generating actionable decisions related to the supply chain, like deciding the quantity to order. These decisions should be actionable, practical, and profitable. Unlike a data scientist, a supply chain scientist takes ownership of the business value of his or her proposals. This requires understanding enterprise systems and the interplay of data with the software and the people operating it. This ensures a comprehensive comprehension of the problem to be solved.

The task of a supply chain scientist entails understanding the extracted data, building an optimization model, and balancing complexity and precision. Vermorel acknowledges the complexity of the real world, especially in supply chains, which renders perfect mathematical modeling impractical. Instead, supply chain scientists must resort to approximations and heuristics to effectively solve problems. They need to view the whole picture and remain committed to it.

Chandler then introduces the role of IT departments, querying whether they should take responsibility for the software and the people, given they usually implement and maintain the software systems.

This question hints at a tension between the operational, technical, and strategic roles within an organization.

The conversation primarily probes the distinct responsibilities between Information Technology (IT) and supply chain scientists, as well as the challenges that data scientists confront in the current landscape.

Vermorel argues that IT and supply chain scientists hold separate responsibilities within an organization. He equates the role of IT to a maintenance one, safeguarding the constant and smooth operation of systems and processes. IT’s responsibility is to keep things running every second, managing the technicalities involved in preserving system uptime and security.

Contrarily, the role of a supply chain scientist, according to Vermorel, isn’t about maintenance. Instead, their duty revolves around data extraction and interpretation. They must ensure that the utilized data offers a correct understanding of business situations and that the derived solutions would yield profitable results. They don’t need to handle the technicalities, as their primary goal is to resolve business problems by accurately interpreting data.

Chandler changes the conversation to the apparent shortage of data extraction and preparation skills among data scientists, despite these being vital components of their profession. Vermorel agrees, noting that university courses and boot camps largely concentrate on programming languages like Python and R, often neglecting the more practical aspects of the job.

Vermorel clarifies that universities are better equipped to teach certain aspects due to accessibility and confidentiality concerns. Supply chain data from large companies is not readily accessible for training due to privacy issues, while open-source software and statistical frameworks are more available. As a result, students often graduate with a deep understanding of programming and statistical toolkits but lack practical knowledge on handling real-world supply chain data.

Vermorel warns that this overemphasis on programming and statistics could lead to overconfidence among new data scientists. They may mistakenly believe that these skills alone suffice to solve supply chain problems. However, supply chain management isn’t just about programming or statistical analysis; it’s about understanding and making business sense of the data. Vermorel cautions against disregarding the wisdom of low-tech supply chain practitioners who tend to rely on common sense and simple tools, like Excel sheets, to make business decisions.

Full Transcript

Kieran Chandler: Today, we’re going to be talking about a new job role that is sweeping across the industry like wildfire. The role of a data scientist is becoming increasingly more relevant at a time where businesses are placing more importance on data and drawing relevant conclusions from it. Five years ago, no supply chain director had any need for data scientists. However, today that’s all changed with the amount of job opportunities for data scientists seemingly growing faster than universities can seem to produce them. So, Joannes, what’s changed? Why is there suddenly this need for more data scientists?

Joannes Vermorel: Clearly, businesses recognize that their data has a lot of value. As soon as they recognize that, they need a lot of people to extract the value from the data and that’s what data scientists do. However, the interesting thing is that it’s not entirely new. For those who were around in the 90s, or maybe by the end of the 90s, at the time it went under a different name - data miners. People were mining things from the data. So basically, the data scientist seems to be the data miner version 2, or something similar.

Kieran Chandler: If these data miners aren’t still around today, I’m guessing the results didn’t fare so well. Perhaps you could tell us a little bit more about that and is there something we can learn from why it went wrong?

Joannes Vermorel: It’s very interesting because you can see that in supply chain circles, data scientists have become very fashionable. It seems to me that there’s a macro trend where things cycle in and out. Two decades ago, it was about data mining and nowadays it’s about data science. It’s the same pattern going on just under a different name. Two decades ago, we saw the emergence of hundreds of companies providing tools for data mining and most of them disappeared. Nowadays, we see an emergence of hundreds of companies delivering data science tools. We also see data science consultants. So yes, there is something true at the core but there’s also a cyclical fashion effect about it.

Kieran Chandler: It’s not every day you hear the words data science and fashion in the same sentence. What we’re sort of saying here is the data miners used to go by a different name. So, should we not start selling Lokad technology to universities so that the next generation of data scientists, whatever they’re called, are fully trained in the tool and they fully understand how to use it?

Joannes Vermorel: That’s certainly an angle. By the way, all the companies pushing for data science tools are aggressively marketing themselves toward universities. An easy way to do this is to promote open source tool kits, because they fit the general mindset of universities. However, it’s primarily a marketing tool. It’s good in a sense, but it doesn’t have to mean efficiency. It doesn’t mean that if you successfully promote yourself within universities, you’ll necessarily get results in your business. It’s not because you become a great mathematician or a great coder that it will immediately translate into actually creating business value within your supply chains. This is a danger I believe, and that’s one of the core reasons why at Lokad we prefer the term “supply chain scientist.”

Kieran Chandler: Business-first makes sense to a lot of supply chain practitioners because they’re exposed to business a lot in their daily work. Maybe the only exception is in very large companies where data scientists can get swamped by the sheer volume of data or the complexity of their problems. But, is there any catch beyond focusing on

a particular business problem?

Owls and reducing lead times and things like that. So, is there any catch beyond focusing on the correct business problem?

Joannes Vermorel: Yes, there is a big catch actually. A data scientist’s role is not only about analyzing the business; it’s about making a difference and being able to take and implement a decision, which can lead to real business impact within the organization. This can be tricky because data scientists can easily access data and produce analyses. But when it comes to acting, it often challenges the status quo. It’s not just about disagreements on order quantities but disagreements that go deeper. The biggest potential for failure lies when the data scientist is not in a position to truly act and deliver value to the business. That’s probably the main catch that I can see.

Kieran Chandler: You mentioned there’s a sort of push against the status quo. I can certainly have quite a lot of sympathy with some of the supply chain practitioners because they’ve been working with methods that have worked for decades. So, if you’ve got someone questioning what has worked and what has been working, I can understand why they approach things with a great deal of skepticism. You mentioned that at Lokad, we have supply chain scientists rather than data scientists. Could you possibly tell us a little bit more about them and why they go by a different name?

Joannes Vermorel: I think the different name reflects our approach to the problems. Our commitment lies with the supply chain. A supply chain scientist is someone who should generate real, actionable decisions, such as how many you need to order right now. The decisions should be actionable, practical, and profitable. It’s about someone who takes ownership in the business value of his or her propositions. This ownership entails quite a lot of things actually.

To put it in context, let’s walk backward. The decision is the end game, but if you start backward, it begins with data. The data comes from enterprise systems, but the data only makes sense through the eyes of the people who operate the software. So it’s not just software; it’s software plus people. The supply chain scientist needs to have a very good understanding of this. They need to comprehend the problem being solved, make sense of the extracted data, and then build an optimization model of some kind.

There’s a trade-off between complexity and precision. The real world is incredibly complex, and supply chains are no exception. It’s not possible to have perfect mathematical modeling, so you need to approximate and use heuristics, which are just recipes that work. The supply chain scientist needs to put all of these things together to ensure there are real savings, not just in percentages but in actual dollars. They need to commit to this whole picture. That’s what a supply chain scientist is about.

Kieran Chandler: Okay, but you’ve mentioned that a supply chain scientist should be responsible for the software and the people. What about the IT departments? Shouldn’t they be responsible for that? After all, they’re the ones who have put the software in place and often, they’re the people who build it.

It sounds like quite a lot of responsibility is being put on the shoulders of just one supply chain scientist. Are you expecting a miracle?

Joannes Vermorel: Yes, the responsibility is quite enormous. However, there is a significant difference. I believe that the core responsibility of IT is to ensure the system is operational. IT needs to handle the ongoing operations and ensure things work every second. The supply chain scientist has a different responsibility. This individual is not tasked with keeping everything up and running.

Kieran Chandler: So, what exactly is the responsibility of a supply chain scientist?

Joannes Vermorel: The responsibility of a supply chain scientist is to extract data and make sense of it. It’s a very different task. This person doesn’t have to deal with all the technicalities involved in keeping something up, running, and secure. That’s IT’s responsibility, which is indeed very difficult. The scientist’s commitment is about ensuring the understanding is correct. The business solution that emerges from this understanding needs to be profitable as a result of precisely identifying a problem that the business truly needs to solve.

Kieran Chandler: It seems that data extraction and preparation are critical tasks. However, aren’t data scientists inadequately trained in these aspects? Most data science courses and boot camps are about programming in languages like Python and R.

Joannes Vermorel: That’s an excellent question. Universities excel in certain areas and are weak in others. Let’s face reality: making sense of data requires actual data in the first place. Most large companies with sizeable supply chains do not share their data with universities. Therefore, universities use as training materials what they have access to. Accessing open-source software is much easier than accessing confidential supply chain data.

Kieran Chandler: There’s a lot of discussion about personal data, you know, similar to the GDPR in Europe. It’s requiring significant efforts from everyone to comply. So that’s incidental but complicates the situation. Universities, for instance, you want to train people in the hardest tasks, where they’ll be able to deliver the most value, but it’s difficult. So it’s much easier for universities to fall back on programming languages and statistical frameworks because they are more accessible, more mathematical. It’s also easier to test students on these subjects, which, as a professor, you need to both teach and evaluate your students. That requires teaching something where evaluation is possible. It’s a strange constraint, but it certainly influences what you can teach at a university.

Joannes Vermorel: Now, the main problem I see with this focus on statistical toolkits is that it can lead to overconfidence. It’s beneficial to know how to program, to be fluent in statistics. It’s certainly something that will help, it’s not a negative. But it comes with a subtle issue. It can make people overconfident, believing that knowing how to program, understanding statistics, understanding math, that this is the key to solving supply chain problems.

And here, there’s a certain wisdom to many supply chain practitioners who are often very low-tech. They try to stick to common sense, they stick to their Excel sheet. And there’s wisdom in that because they are sticking to what makes business sense. If the only reason why you’re sticking to common sense is that you lack knowledge about statistics and programming, that’s not ideal. But, on the other hand, if all you know is statistics and programming, that doesn’t make you an expert in supply chain.

Just because you’re proficient in these areas, it doesn’t automatically translate into solutions that will generate extra euros or dollars. So I believe that’s the biggest danger. We are now producing armies of people who frequently suffer from overconfidence. Programming is a means, not an end.

Kieran Chandler: That’s an insightful perspective. Thanks for shedding light on the subject of data scientists and, indeed, supply chain scientists. It’s a topic that’s becoming increasingly relevant given the staggering amount of data collected in our daily lives. Thanks for taking the time out today.

Joannes Vermorel: Thank you, Kieran.

Kieran Chandler: And thank you to our listeners for tuning in to today’s episode. We’ll be back very soon with another one. Until then, keep asking your questions and sending us your thoughts. Thanks very much for watching, and we’ll see you again very soon. Bye for now.