Reporting from Spain and Canada, Juan Prida and Silvi Q. reveal the benefits of switching to Vista DnA’s data science chapter, and why data analysts make such good candidates for metamorphosis.
Visualize career change
Juan: Vista’s data mesh – decentralized data approach – means tight collaboration with data scientists. I was working as an embedded data analyst in the Pricing, Promotions and Personalization (PPP) domain when I realized machine learning (ML) models were my calling. There wasn’t and isn’t a set route to become a data scientist, but you do need passion – it’s hard work jumping into hours of self-study once you’ve finished the day job.
Find people to back you
Silvi: Having a mentor is essential. Credit goes to Senior Data Scientist Marta Dies, who made her own career leap from biophysics. In the run-up to formal change, Marta let me explore forecasting projects for Vista’s Manufacturing and Supply Chain (MSC) domain to figure out what the role entails. DnA’s open approach to professional development meant that our MSC domain lead, Roy Wildeman (blog), also stepped in to support, helping me to identify goals and create a transition plan.
The matrix difference
Juan: With multidisciplinary teams in a matrix organization under mesh architecture, it is complex, but the daily benefits, quality of interaction, and overall cohesiveness of our DnA teams pay off. In morning ‘stand up’ meetings, I’d hear which new feature people were working on – it was an inspiring part of the career discovery process. DnA’s inherent fluidity meant I was invited to collaboration sessions in the data science chapter and got to showcase the skills I’d been building up over time.
Why analysts make good data scientists
Silvi: The magic of data science is the more you know, the more you don’t know. It’s a fast-growing field academically and in the corporate world. Plus, there’s a lot of independence to pick out what to explore. The data analyst’s retrieval, analysis, and visualization skills are valuable to the data scientist who’ll focus on wrangling, modeling, and post-processing data. So an analyst’s descriptive skills are a springboard to the predictive skills a data scientist must develop.
The PhD question
Juan: The short answer is, no – a PhD isn’t a ‘must’ for data scientists – it’s the first thing I asked when scoping out change. You do need a good understanding of maths and statistics, with computer programming experience.
A series of firsts
Silvi: 2020 was a big year. I was the first data scientist in the MSC domain. And Vista’s matrix-led reorganization brought the domain into DnA’s data science chapter for the first time. I got to pilot everything by myself – even extracting data because we had no engineer at the time. It doubled the challenge, but also the learning. Kudos to the team of five I developed a forecasting data product with, which rolled out globally in 2021. Finance teams predict a seven-figure saving for Vista, so the rewards are exciting.
Juan: I wasn’t expecting to have the freedom to propose many model applications when I transitioned to the data science chapter six months ago as part of the PPP domain. I thought I’d be taking more direction from senior data scientists, but the first idea I proposed became part of this quarter’s objectives and key results (OKRs). It’s a model we piloted with the team to optimize discount recommendations against products – something you’d use in a marketing campaign.
World at your fingertips
Silvi: Vista’s remote-first approach to work has opened up internal career progression on an international level. It’s also made the working environment both more focused and flexible. I had a wonderful moment recently when I met my global team in Windsor, Ontario, where I’m based, for a tour of the local Vista manufacturing plant. Our camaraderie is the same online or on the ground.
Juan: DnA is tackling many complex, challenging and really interesting business problems. How will we deploy computer vision using the vast design data that we have? How should we transform Vista’s promotional strategy from price-based to brand-based? Can we leverage new ways to interact with small business customers using innovative, personalized promotional vehicles? There are always more questions than answers. That’s the fun part.
Silvi: One year into tapping data science capabilities for the MSC domain, it’s all about continuing the groundwork and advancing data products with increased data availability, gearing Vista for ‘Industry 4.0’. We take a proactive stance on manufacturing quality. Right now, with anomaly detection under proof of concept, we’re starting to predict when products, stocks and print presses will face issues.
Juan’s tips for would-be data scientists
- As a first step, spend hands-on time with ML problems. If you don’t have the scope within your company, try free resources like Kaggle.com.
- Get confident using various algorithms, then grasp the theory behind them. You’ll understand why and when different models are useful.
- Focus on the quality of your code: good coding practice becomes vital in a team scenario.