My first three months as a data scientist
I landed my role as a Junior Product Data Scientist at Motorway in November 2021 — nearly 3 months ago! It’s been a whirlwind of a time, with lots learned and many mistakes made. Some expected, some not, but all of them valuable experiences. In this blog, I’ll share some of the things I’ve learned from my first dip in the world of data here at Motorway.
Lesson 1: Simplicity vs complexity
When I started in the field, I had a perception in my mind of building these crazy machine learning pipelines, using deep learning to do some next level analytics and predicting things that could cut costs by 95%. It all sounded very exciting, but what I’ve come to understand is that each problem has different needs and the machine learning aspect of data science is not always required. When it comes down to it, our roles are to help solve problems using data. Whether it’s a case of building a model or improving the quality of data people receive, it doesn’t matter.
I remember one time at university, there was a specific classification problem that involved complex email domain names. As a data science student, I was pumped to use cool algorithms and models to solve problems. I spent days developing a model, refining parameters and ended up with a model that had an accuracy of ~65%. It wasn’t amazing, but it was the best I could do. However I was advised that the 35% false positives would most likely result in overheads that wouldn’t fly in industry. I showed one of my peers, who looked at my model, understood the problem and went off to have a go. A few hours later, much to my surprise, he comes back and tells me he’s solved it. Confused, I had a look at his solution. He had basically built a solution using a set of regexes, with a >95% accuracy. I felt quite humbled that day.
As a fresh data science novice, the learning curve is constant. Tasks sometimes result in endless Googling to figure out a solution or solving bugs you didn’t even know existed. Navigating the complexities of the data and producing meaningful insights is a continuous learning process. However I’m learning that it’s a trap to search for solutions in search of problems. A common theme has emerged for me, that the smartest people are those who spend time really understanding the problem and then coming up with the most simple solution.
Lesson 2: University ≠ Industry
Another misconception I had when moving into industry, is that data would be clean, with nothing missing and no duplicates. I couldn’t have been more wrong. Gone were the days of perfect public, Kaggle like datasets. At Motorway, we use dbt to transform data, enable testing and produce documentation. Learning about the ELT workflow in general and tools such as this, highlighted the importance of all the steps that occur before getting a perfect data set. Data is generally messy and data collection and integrity are important parts of the process, which shouldn’t be overlooked.
Lesson 3: You’re not working alone
In addition to analysing data and building models, there was a major part of data science that I was missing. I was thinking too much about the tech, not the business. This required a major shift in mindset coming from a purely academic background. The role has no meaning if we can’t provide value or bring some profit to the business.
What I’ve realised is that SQL, Python and dashboarding will only get you so far. The story you tell and interpretability matters the most. Ultimately, people are at the core of what we do. People have emotions and biases and no tool or technique can overcome that. Data doesn’t automatically create impact. Stakeholders are not necessarily interested in the specific statistics behind the data, cool visualisations or even the tools used. What matters is the takeaway, the insights, and the actionable outcomes this data can provide. It could be in the form of insights on a feature release or helping define the right metrics.
Lesson 4: Culture matters
Admittedly less of a data science specific related lesson, but working at Motorway has highlighted to me the importance of culture. Working in a team which has helped me figure out my way as I enter this field has been invaluable. The hub and spoke model that we use here has offered great exposure to how data science can provide impacts in different parts of the business. Alongside this, having regular guild meetings has fostered a culture of openness and collaboration amongst the data team, whether it’s discussing new tech or different ways of working.
In addition to the work practices, there are many other great elements of the culture here. I remember my first week here involved a company trip to Brighton to check out the offices there and listen in on some of the agents take calls. That was in addition to the announcement that the company had reached unicorn status, and having a party to celebrate that! Throughout all the mistakes and learnings, it’s been incredibly fun to be part of such a fast growing and exciting team and I look forward to what’s ahead.
It’s been a great ride so far and I’m grateful to work here at Motorway. If you’re interested in joining the team check out our careers page to see what’s available.
