To New Data Scientists: Work at Mature Companies to Free Your Creativity
This is post 2/2 from our interview with Emil. You can find the 1st post here
As Director of Product Analytics at IKEA, Emil S. Jørgensen knows a thing or two about the world of Data Science today. His team is responsible for infusing data and analytics into the expanding range of customer-facing digital solutions at IKEA, including omnichannel usage of machine learning (ML) to improve business performance. If Data Science plays a role, Emil also plays a role.
But modern Data Science can be a daunting field to go into as a young engineer or scientist. The field is less than 20 years old, and has seen almost unprecedented growth in demand world-wide alongside the popularity of the internet (in particular, social media), which generates vast amounts of data to work with. This has caused a wide range of responsibilities to be placed under the Data Science “umbrella”, resulting in huge variation between roles at different companies.
To get a better grasp of what Data Science is today, I spoke to Emil to learn about his view on the matter. Below, we dive into his own experiences and learnings from various Data Science roles and hear his thoughts on why well-established companies typically offer a superior learning platform with regards to Data Science, rather than e.g. consultancies and start-ups.
Career advice is always subjective and these choices are, at their core, personal preferences. However, there is still tremendous value to be found in mimicking and learning from interesting people with interesting careers – especially those whose story you can see yourself in.
Conflux Insights is a new series of articles where we learn key insights and findings from interesting Conflux mentors from the Danish STEM industries and share them with you.
Emil’s career in Data Science
In 2014, Emil graduated with an MSc in Mathematical Modelling and Computation from DTU. To further deepen his mathematical understanding he chose to pursue a Ph.D. in Probability Theory at the University of Copenhagen, before entering the Data Science world.
He joined KAYAK, one of the world’s largest meta-search engines in the travel space, as a Junior Data Scientist in 2017. Why KAYAK? “I was searching for a company with the right open-minded and casual work culture that large technology companies are known for. Due to a recent acquisition of the Danish brand Momondo, KAYAK had a new office in Copenhagen, and provided a great opportunity for personal growth and the chance to travel to our Tech HQ in Boston too during my stay”, Emil says.
As an insider, Emil’s inner statistics nerd was pleasantly surprised to see the role of data and analytics for decision-making purposes at KAYAK. This was a data-centric company; they constantly ran large A/B testing suites as part of product development and had built their own resilient, easy-to-use infrastructure to enable it and empower teams. Along with it, self-service dashboards were providing real-time insights into product performance across all platforms, allowing for rapid hypothesis development and error mitigation.
It was a fascinating place to work for Emil, and he enjoyed his time there fully. But after about a year at KAYAK, an irresistible offer landed on his table: the chance to join a newly created Data & Analytics organization in IKEA as one of the first employees, and thereby getting the opportunity to shape its growth right from the get-go.
Looking back, Emil is very conscious of how the engineering mentality and leadership culture at KAYAK shaped his thinking on how to successfully lead a Data Science team.
“The most important thing I learned at KAYAK is an understanding of what good looks like’. At KAYAK, we had a meticulous focus on ensuring that things were done right. Tools were often built in-house to ensure control and flexibility when it comes to infrastructure. Moreover, I 100% bought into their agile and experimental culture and values – values that I was able to bring with me to IKEA.”
His new position at IKEA meant a shift in what his responsibilities were, as he had to transform from being an Individual Contributor, leading the implementation of ML applications, to building and managing a team of 10-12 Data Scientists. And this shift turned out to suit Emil’s skillset quite well.
Data Science today
From his former role as Data Science Manager, Emil has his finger on the pulse of the industry. I asked him if he could provide some examples of what Data Scientists typically end up working with:
“Some focus on what I would consider the core of the field. They develop and improve ML applications and systems, closely analyzing the impact of changes to input features or algorithmic modifications on performance metrics. And the business potential in this area is indeed also huge! Combining the increased access to data sourced from the internet revolution with ever-growing cloud-computing possibilities, this core is now being brought forward as a collaboration between academia and large R&D departments in private companies.”
According to Emil, we’re witnessing an interesting change when it comes down to who has the upper hand around cutting-edge ML research: around the 2000s, this area was heavily dominated by academia, but with ownership of large datasets solidly placed within the industry, they find themselves in a position to not only do great research but change the world with applied AI. Great examples include the wave of research happening inside Facebook Research, Google DeepMind, and OpenAI, all of which are pioneering substantial progress in modern applications of AI. To learn more, check out our previous post with Emil, in which he explains why joining a large R&D department is strongly worth considering over a purely academic career path.
Besides fundamental contributions to learning theory, an area that is arguably going through the most rapid development today is tooling. This progress is central for lowering the entry barrier into Data Science day by day – something that Emil believes is key for truly democratizing applied AI and making it accessible for a large audience. You can tap into state-of-the-art research using PyTorch (developed by Facebook) and Tensorflow (developed by Google) without having a deep theoretical understanding of the models you use. “You don’t necessarily need to pursue a Ph.D. in machine learning or statistics to enter this field, as the growing availability of simple-to-use tools does a lot of the dirty work for you. Yes, there is still quite a steep learning curve and a lot of theory to understand today, but the barrier of entry gets lower over time. In particular, I think companies that are willing to invest time and effort into upskilling software engineers and allow them to learn the basics of statistical reasoning will gain a strong advantage over those with a small, central team of data experts.”
According to Emil, the true value of Data Science lies not in the application and interpretation of advanced algorithms themselves, but in the enhanced decision-making options they unlock. “One, if not the, most important responsibilities for my team at IKEA is to ensure that we correctly evaluate and make decisions based on our internal analyses or outcomes of experiments”, he points out. “The focus has to be on deriving the right conclusions, and not on applying a particular class of algorithms, such as recent deep learning techniques.” And this can be disappointing for new university graduates, who are used to spending a lot of their time getting up to speed with modern research literature and trying stuff out locally on their laptops. Once within the industry, you must take a step back and realize the greater scope, you are now actively contributing to:
“If you are hired as a Data Scientist at IKEA, you should not expect to go straight to the metal and start coding away. Instead, you need to learn how we at IKEA make decisions, and who’s involved around the table. Only when you understand the bigger picture, you can start coding. This ensures that we drive impact.”, Emil explains.
When to reap the benefits
As a rule of thumb, Emil recommends newly graduated engineers and scientists to search towards well-established, mature companies early in their careers, instead of joining a 10 person start-up as Chief Data Scientist:
“As a company, you need to reach some kind of stationary state and market fit before the power of Data Science can be felt. Not only does it require the right engineering infrastructure, but it also requires heartfelt commitment and dedication to high-quality performance measurements and visualization first. Moreover, quantitative insights to support early-stage product development tends to have much higher interest and ROI, rather than ML applications themselves. Data Science is intimately connected to business optimization, and the biggest results are naturally associated with a certain scale. As a creative thinker within data science, the complexity that almost always surrounds large-scale problems in the industry is an amazing playground and one to seek out.”
A more basic reason to be cautious of joining a small company is that you, as one among few or potentially even the only Data Scientist in the company, are likely to have to take a vast responsibility for not only the development but also the operation of your ML system. This includes
- maintaining and improving the data infrastructure and setup,
- collecting, storing, and preprocessing raw data into curated datasets, and
- developing and maintaining dashboards with the right engineering and performance metrics.
This is a big responsibility to take on, especially as you enter the field. At more mature companies, a lot of this foundational work has hopefully already been thought of and partially taken care of – meaning that you get your hands on exciting analytics projects faster.
“When done right, Data Science is an integral component of cross-functional development. That means that Data Analysts and Scientists don’t sit in the corner and take requests, but work side by side with team members from disciplines like Engineering, or UX Design. During my time at both KAYAK and IKEA, I have seen the outcome of working like this firsthand, with impressive results.”
What about Data Science consulting? Almost all of the large consultancy firms now offer data and analytics advice or services to their clients. And before joining KAYAK, Emil was also seriously considering going down this path. So why didn’t he seize that opportunity?
“If you want to advise others, learn the craft first. And while I see the value and role of consulting around topics like data and analytics strategy and PoC development for clients across industries, Data Science is a marathon, not a sprint. To me, the only path to becoming a genuine expert worth listening to is to be part of in-house engineering culture and learn the basics.”
5 quick takeaways
Data Science is a rapidly developing and exciting field with great career possibilities. Based on my conversation with Emil, a few pointers to keep in mind are that Data Science is:
- very loosely defined today. When applying for a role, be mindful of whether the role description matches your skills and what you want to work with.
- only as valuable as its ability to influence and improve decision-making. Therefore, great Data Scientists are not only theoretically strong but also need to build a firm understanding of the domain in which they operate.
- predominantly good for optimizing applications and processes. Hence, the business case for companies who have reached market fit to invest in Data Science is apparent. A start-up will often benefit much more from qualitative or quantitative insights to provide a strong foundation for business and product development, but less from in-house built ML systems. Try to assess the companies’ needs before joining.
- the tip of the iceberg. To continuously generate value, Data Science requires an extensive and well-functioning infrastructure to support the creative and agile work you will be doing. Choosing a mature, data-driven company can have a clear edge here.
- slowly taking over many aspects of cutting-edge ML research, especially computational ones. Large R&D departments at private companies are worth keeping a close eye on as alternatives to academia.
If you have any comments or questions about the article or Conflux Insights in general, please reach out at email@example.com.
Thanks to Emil S. Jørgensen for sharing his experiences and career history with us.
This interview was performed and written by Jakob from Conflux.