Moving from university to industry is a big step for many, and finding a position as a junior in #datascience, or #Tech in general, can be a challenge. After three years of working in industry, I want to share my three key findings. Hopefully they help you on your journey. :)
TL;DR:
Many job advertisements in tech require you to have a minimum of three years of work experience. How do you ever get started just out of university, when there are few dedicated junior positions? And is working experience in academia (e.g. a PhD) not also real working experience? I’ve recently crossed this magical barrier of three years work experience in industry. Here are my top three learnings that might help you succeed faster.
When I first started to apply for jobs as data scientist one of the most common reasons for not being hired was the fact that I didn’t yet have the industry mindset. I didn’t know what that meant, but could only accept their verdict. After being exposed to the industry mindset, I realised that the industry mindset starts with two questions:
You generally never do anything just for the fun of it or because you were curious. There is always an intention behind every action. And there is always someone at the other end of your work that should benefit in one way or another. For example, my work as a data scientist should enable others to make decisions in a more informed manner, automate (partial) workflows, suggest ways to spend marketing money more profitably, etc. In my case, the people our team serves are primarily other employees in the company. That means if the people that should use our tools or predictions don’t believe in them, or don’t use them, we failed. It doesn’t matter how accurate our results were. This change in what is considered a success was a big learning point that shifted the way I approach my work.
To make this point a bit more explicit working in industry is the practice to your theory. The biggest difference to academia, both studies and research, is that in industry you are thrown into the large sea of life with all its imperfections and there is hardly a chance to create nice idealised conditions. A part of data science in industry is to take in that large variety and cluster it in ways to make meaningful and actionable predictions. Other ways in which data science and data-driven decision making has started to bring the rigour of science to the messiness of real work problems is by creating small proof of concept studies. Since in industry the show must go on, always, you create small pockets to test your new theories on the live business. This is exciting and, when it works, very rewarding.
The second point that I learned is that many processes are strongly interconnected. Therefore, seemingly small changes can have large consequences in processes that are impacted. The larger the company the bigger this issue. Your prediction might be great, but it could be hard to implement it in a short time. The better you understand how business processes are connected, the better you can translate the results into actionable recommendations. Luckily this is not the job of data scientists alone but also other team members, such as analytics translators, business owners, and others. However, business experience does give you a better context in which the data you use was gathered, what constraints you should keep in mind, and what bias to take into account.
Third and most relevant when you’re looking for your first position in industry, having working experience in industry is similar to a portfolio in the arts. It shows what you’re capable of, how you work under variable deadlines, how you collaborate with colleagues to achieve a goal, what you can deliver. It is often the easiest and best approximation the hiring manager has to evaluate your possible deliveries at the new company. That’s why many people in data science suggest working on different projects by yourself and putting them online, e.g. on Kaggle, Git, etc. This is not bad advise, but I would like to add that you can further expand this exercise by thinking about the use of your results in a real life scenario. Come up with a fictional business question and try answer it with the data you have, make recommendations, think what actions should be taken next, what results you would expect, who would use it? To make it easier, you could start with a question that interests you, find or generate some data that could answer it, and then write the analytics story with conclusion.
Overall, if you’re at the beginning of your career try to always think of the intention behind the business question that you are being asked. What does the answer to the question enable that saves time, money, resources or creates higher sales, customer retention, or customer satisfaction? What kind of action should follow your predictions? Your model will be part of a larger story and knowing this will help you put your work into perspective and make better decisions.
This article was first published on LinkedIN.