Data science graduate diary: first rotation as a DS

Peak’s data science grad scheme

During the year of the graduate scheme we will be on four, three-month rotations across the data science (DS) team here at Peak. The four rotations are Operations Data Science in both retail and manufacturing, Insights and R&D.

Operations focuses on longer term projects with our customers. Any two Operations projects at Peak will be completely different. As Amy mentioned in her blog, the range of solutions is amazing, and the fact that we are working on all these projects with great companies makes it incredibly rewarding to be at Peak!

Insights is normally shorter term projects (typically one month) that aim to answer specific stakeholder questions, such as identifying alternative products to replace styles that are popular but out of stock, and looking at the popularity of stock items to help decide which ones to promote or discontinue. Insights also try to find potential value we can unlock to drive even more gains for existing customers.

R&D is the newest addition to Peak as a formal data science function, and it’s exactly what you’d expect; keeping up with the cutting edge of all things data science. R&D are also building some packages for both external and internal use, the plan is that some of these will eventually be made open source so everyone can benefit from their collective genius!

Diving back into retail

My first rotation is in retail ops, which I was really excited to get started on since I worked for a Food Box company as part of my Master’s dissertation; I knew I still had lots to learn in retail. I also wanted to see how Peak tackled retail problems – I could tell from my application process how smart everyone was and I wanted to see if they approached problems differently. 

For this rotation, I worked with a massive global sportswear retailer, I was so excited to get cracking and start delving into the data. And it didn’t disappoint! I absolutely loved my first rotation, and I’m sad that it’s coming to an end so soon. 

The team I’ve joined is the largest DS team at Peak, with five full time data scientists all working on one project. For context, most projects here at Peak have around one or two data scientists. I’ve loved being able to learn from so many people. 

As a team, we have an hour, three times a week set aside for a brainstorm. Here we either focus on a pre-discussed topic that one of us is struggling with, or if there’s nothing major we focus on small issues. This has been so helpful for my development, as everyone in the team is so knowledgeable about both data science and business.

This blend of expertise means I always come away from the sessions with more knowledge than I had at the beginning of them. I have just finished my Master’s in Business Analytics, so being able to apply that knowledge to the real world and build on it with actual examples has been great.

These brainstorm sessions have also helped me to learn that there’s always easier ways to code things. Apparently using 100’s of lines for loops can be hard to read, understand and can easily go wrong!! Who knew?! Whenever I show my code, it always comes away much thinner, more efficient and properly functional, which is probably the three most important things code should be! 

Although this isn’t an Insights rotation, I’m working on a more insights-style project. This means that I have also had the opportunity to present to stakeholders in the wider team at the company, which is such a great opportunity, especially so early in my time here at Peak. 

This meant that I had to build some slide decks, which is a completely novel experience for me as I’ve only ever had to make small presentations for uni before, not presentations that are trying to show the size of a problem for a massive customer, that could change the way a part of their business functions. I’d say the stakes are a lot higher now than before! Luckily, I wasn’t alone for this either, with colleagues on the project there to help me. Special shout out to Áine, our incredible Customer Success Manager who helped me refine my slides for the presentation, and keep it all on brand for both Peak and the customer. 

I got the opportunity to present to our main internal stakeholder, which I felt went really well, and helped us to focus more on the specifics of the problem so that we were prepared when presenting to the customer. I’m still working on the analysis as I write this, but I’m genuinely so interested to see the final answers – I’m really invested in the problem and how to best reflect it to the company.

My first Merge Monday!

Another of the great things about this rotation is being added to the Github repositories for the solution (I know, so glamorous!). Although I’m not working on the solution directly, I can still see the code. Being able to explore how the project has evolved over two years and understand the intricacies of solving a massive data science problem has been invaluable for me. 

One quirk of our unusually large team is Merge Monday, where we merge Pull Requests to the Dev Branch on Github (once they have been approved by the Squad!). I have just had my first pull request merged on our Merge Monday, which is a rite of passage for the project; although I’m still waiting for my first Monthly Main Merge Monday… (the same, but from Dev to Main)! 

This means I’ll always have my fingerprint on the solution, which is weirdly moving. It’s also great to see the actual code of the solution, as it can be hard at times to actually know what techniques to use for solutions, especially at the beginning of my journey as a data scientist. I now know the stages of one potential approach to this type of problem, and that’s very powerful for my own development, especially when I see the problem again in the future.

Getting to grips with the lingo

As we’re working with a large customer, I think the steepest part of my learning curve has been trying to understand their constantly changing landscape of acronyms! We have a three page document listing some of the most important ones, but this is almost a year old now, and everyone in the squad is regularly confused by one of the many new acronyms we hear in almost every meeting. But at least if I ever get quizzed on their acronyms in the future I might do marginally better than average!

Introducing…the ‘Spaghetti Diagram’

A similar learning point is working with masses of data, and trying not to destroy my laptop if I load in the customer’s whole inventory table by mistake. We have an internal document known as the “spaghetti diagram.” It lists all the roots of tables and how they’re all created, and, as you could probably guess, it’s pretty messy! 

I still don’t know what the difference in even some of the refined tables are, but it’s been very useful. At uni the majority of the data I had to work with was very clean, this meant it was easy to see where and why to use any given table, that’s not the case in industry. Data in the real world is hardly ever like that, so it’s been great to see what actual real world data looks like, even though the data we are given is still clean. 

All in all, I’ve absolutely loved my first rotation here at Peak, and I think I’ve learnt more in these three months than ever before, except potentially when I was learning to talk! It’s been great to interact with so many talented data scientists and to try and take on as much information from them as possible before I move on to R&D to learn even more.