Author name

Editor’s note: the post was originally published at Towards Data Science, a Medium publication sharing concepts, ideas, and codes.

Troy Shu is currently a data scientist at Lyft, a transportation company with over 23 million users. He’s located in New York, where he’s helping build out the bikes and scooters side of the business. His work includes collaborating with data engineers to build data pipelines in Airflow, creating dashboards, conducting A/B tests, and working with product managers on product analytics.

Before Lyft, he worked as a data scientist at Squarespace, ran his own data science consultancy, worked as a software engineer at Bond Street as well as a research analyst at AQR (Where Wes McKinney created and open-sourced the Pandas Python library). He graduated from the University of Pennsylvania with a dual-degree in Computer Science and Economics.

1. On Starting A New Role
When starting at a new company, Troy first recommends building domain expertise. The best way to build domain expertise is to talk to lots of people, both within your department as well as outside of it. When he first joined Lyft, he made sure to schedule 15mins with folks from a variety of departments to try to get a lay of the land. He supplemented these conversations by reading as much documentation as he could get his hands on. During these meetings, he focused on asking lots of questions in order to accelerate his learning.

Once you’ve established domain expertise, you can start identifying quick wins, projects that typically last less than 2 weeks, in order to establish credibility and trust. To make sure these projects are impactful, first learn what’s most important to your manager, your manager’s manager, and the company as a whole. Another way to source quick wins is to determine what phase your team is currently in. Are they trying to build something or are they trying to figure out what to build next? Answering these types of questions will help inform your decision on the type of quick win to pursue.

2. Becoming A Better Storyteller
In most cases, storytelling is one of the most essential skills that a data scientist can develop. The best storytellers are the ones that can create some sort of flow when presenting insights. This means avoiding non-sequiturs in your presentation and leveraging a top-down storytelling structure, which means starting with the most important insight (Don’t bury the lead!)

Creating fake user personas in order to demonstrate how they use the product is another effective technique in storytelling. Being able to humanize your data can make the insights much more compelling and memorable.

Another key component of storytelling is visualization. A good visualization can make or break a presentation. It may even end up being reused in other presentations throughout the company! Troy recommends reading Storytelling With Data, and the Chartify python library for those who want to improve their data visualization skills.

Finally, look out for opportunities to collaborate with UX researchers. Whether that’s by partnering on a project together or leveraging each others’ existing work. Including joint quantitative and qualitative insights in your story can make for a powerful combination.

3. Stakeholder Management
Working with various stakeholders means being able to take on their perspective. For example, if you’re working with a Product Manager, showing some product sense can be a sure way of building trust. Another technique that can be useful is taking the perspective of users, and leveraging human-centered design principles.

A trap to avoid when working with stakeholders is going down a rabbit hole, and coming out with a polished deliverable. This approach can easily backfire if it turns out that the work you’ve done is not what the stakeholder had in mind. The better approach is to be iterative and incremental. For example, leveraging lean startup methodology to build initial prototypes and getting feedback from stakeholders to make sure you’re on the right track.

Finally, be direct with recommendations, which should be concrete and clear so that the stakeholder can assess the options and make a decision. Try not to overwhelm them with the nitty-gritty analysis. Instead, opt to put that information in an appendix or omit it all together so that you can focus on the key elements of your recommendation.

4. Managing Up
One of the most important relationships that you’ll manage as a data scientist is the one with your boss. It’s important to learn the art of managing up so that you can set yourself up for success. Listening more than you speak, and making your manager feel heard is a great way to build rapport. There’s nothing worse for your relationship with your manager than getting their advice, only to do the complete opposite of what they recommended.

Figuring out what your manager is evaluated on, and then helping your manager look good is one of the best ways to build a good relationship with them. Ultimately the fate of your career rests in the hands of your manager, so be sure to find out what motivates them, and how to best gel with their management style.

5. How to Build a Consultancy
For those looking to take on part-time client work or to build a full-time consultancy, Troy has a variety of recommendations. The first is to be disciplined about setting your work hours and creating processes for yourself. This might mean renting out a coworking space where you can simulate a traditional work environment. Sticking to a schedule similar to salaried employees can help with building positive momentum and establishing a positive routine.

Another recommendation is to hone your ability to pitch and sell. A large portion of your time when you’re starting out will be trying to find client work. As you grow your reputation, you may start to receive inbound but in the initial stages, there will be a lot of legwork to get the first few projects.

When sourcing clients, the three main approaches are personal networks, community networks, and online portals such as Upwork. Your personal network is likely where you’ll have the most success in obtaining client work, whether it’s past colleagues or your college alumni network. The community network consists of physical communities such as those found on Meetup, or online communities such as Reddit. In the latter case, Troy was able to find some client work after doing a quick one-off data analysis for someone on a Reddit data science community. The third approach is the less lucrative of the three, as you’ll be competing with a global freelance talent base, many of which have no issue with putting in low-bids given their low cost of living.

Data Minds is a series that profiles professionals working with data. In this series, you’ll learn about their story, day-to-day, and advice for others.