You just landed yourself as the sole data scientist/engineer/analyst at a company and are a little nervous, eh? Or maybe you have had this role for a couple of weeks and are beginning to realize that the paths forward are incredibly murky? Well, I empathize with you. I have worked at a handful of startups/projects/mid-sized companies over the past few years as the only data scientist. The opinions of this article are my own and do not represent the experiences of any one company but rather a holistic gathering of thoughts and opinions.
The TL;DR of this article:
1. Understand where your work fits in on the analyst/engineer/scientist spectrum.
2. Make it a goal to create a data-driven culture.
3. Establish a work-request process.
4. Don't shy away from meetings.
5. Stay relevant
Understand where your work fits in on the analyst/engineer/scientist spectrum.
People often assume that your role encompasses everything under the data umbrella. And the truth is, when you are the sole person with a data title at the place you work, it does. Still, it is important to understand the hierarchy of data needs.
One of the first steps I take when I get to a company is assess what in the above pyramid is working well and what isn't. Sometimes it looks more like an inverted funnel, sometimes a square, sometimes a top-heavy pyramid. Having this diagram in your head allows you to identify bottlenecks in relationship to company goals.
Make it a goal to create a data-driven culture
I take it on myself when I start a new job to create a mission statement. Most times it will be a variation on the following:
"Everyday, I work toward a data-driven culture where people have access to the information they need to make the right decision."
As part of this, it is my job to be a defender of good data. This means that I regularly ask: "What data was your decision based on?"
This has become my superpower question. I want to get people in the habit of people thinking what data they need to make a decision. This is step one. The step beyond that is getting people to be their own data analyst. Shifting the dynamic to become the subject-matter expert who will help guide a coworker is very different from the data code monkey that just answers people's questions for them. The more people looking at data, the better. However, this means that you have to have a good data infrastructure set up where people can actually access the data they need when they need it. Doing the legwork here has compounding returns and is hugely beneficial.
Establish a work-request process
Whether you are a company of 3 people or a company of 1000, it is important to understand where you can make maximum impact and how you ought to spend your time. You should have a handful of top priorities where you spend your time. If you are stretched too thin, you make no impact. If you are only working on one project, you aren't spreading your data prowess appropriately.
If you are supporting every department under the sun, I recommend setting up some "Data Office Hours" two or three times a week where you can time-box all requests that aren't the priority. This way people are still getting the data they need, but you have a dedicated time to do it.
Don't shy away from meetings
One of the biggest mistakes I see companies make again and again is putting their data scientist in a corner.Luckily this has never happened to me personally, but I have seen it happen to a lot of my friends. How can you expect someone to understand the data if they are not really involved in the process? Don't get me wrong, I don't think that the data scientist should be in every meeting, but they should have enough ownership and understanding to contribute to the project in a meaningful way.
Meetings are also the perfect place to raise concerns about: "What does success look like here?", "How will we get the data?", "If this new feature performs worse, will we really choose not to implement it?"
If it is a one-on-one meeting, I always like to end with the question: "What one piece of feedback do you have for me?" Asking this question gives the space for open communication.
"Data science is hard enough without bad communication. Don't make it harder on yourself."
My trailing tidbit of advice is to stay up to date in the field. If you are the only data scientist, no one is going to talk about the coolest new RNN out there. If you are the only data engineer, Airflow won't ever casually come up in conversation. It is important to find both mentors and peers so you can grow technically.
Most of my mentors and peers are all people I have met online. I read material and take time to reach out to the author and give compliments and feedback.
Here are some resources that I have found helpful along my journey. Get involved in the online community:
If you find yourself in a similar situation or ever just want to chat, I would love to help. All my social media is available on this site. Ping me and let's chat about data!