How to Be a Great Data Scientist

how to be a great data scientist

You put in the time learning the skills, you took the right approach to your job applications, and now you’ve found yourself with a job in data science. That’s great! But now that you’ve got that job, how do you keep it and excel at it? How can you prove your value as a great data scientist?

That’s a question that isn’t always as easy as it sounds. Having the skills is key, but a successful data scientist also needs to know how those skills can be applied to produce a meaningful business impact — one that justifies the high salaries data scientists command.

Your analysis has to be good. But more importantly, it has to bolster your company’s bottom line.

At larger companies, making this happen is often pretty straightforward. A big company is likely to have an established data science process and a team that you fit into, with clearly-defined goals that management knows will generate a return on their data science investment.

But at smaller companies, it can be considerably more difficult. You may be the company’s first data science hire, and there may not be a clear vision for how your role will help the company succeed. You might be tasked with “finding opportunities” and left to your own devices.

This kind of freedom can be a blessing, but it can also be a curse if you’re not sure where you can make an impact. And it can be really difficult to handle when it’s combined with the impostor syndrome that often comes when stepping into a new role at a new company.

Don’t despair, though — you can do this! There are lots of ways to prove your value to your company as a great data scientist. Here are some of them.

Embed Yourself on Business Teams

If you’re the only “data person” at your company, don’t be an island! Your first move should be to spend some time working closely with core business teams (sales, marketing, product, etc.) to get an understanding of what their biggest priorities are, what data is most important to them, and what information you might be able to provide that could help them.

At this early point, it’s important to stay open and flexible. "The first thing I recommend is to abandon a task-driven thought model,” says Daniel Jadczak, a senior developer at Netguru. “This is a process full of discovery where, in the early stages, the industry doesn’t really know how it can benefit from machine learning. That means you have to be ready to propose solutions, iterate on ideas quickly and stay flexible.” As you embed with teams, experiment!

Chances are that simply embedding with a business team is going to reveal some opportunities to you. It will also probably generate some requests from team members. Take note of everything, but remember to prioritize. At a small company, there will be tons of opportunities for data science work, so the key is finding the ones that can make the biggest impact and focusing on those first.

Which projects are likely to have the biggest impact? That depends on what you find while investigating how your company makes decisions.

Pay Attention to Decision-Making

As you’re working with different teams, you should also be paying close attention to how, when, and by whom important decisions at the company are made. Does the CEO define the marketing strategy? Does strategy come from the director of Marketing? Is strategy evaluated every month? Every quarter?

You’ll also want to think about what kinds of data inform those decisions, what other data might be relevant, and how success and failure can be measured.

Knowing the answers to these questions will make it easier for you to impact the company, because you can focus your efforts on data projects that inform critical decisions. Building a marketing dashboard that helps the team execute their current strategy, for example, may be less impactful than using predictive analytics to demonstrate how a different strategy could yield even better results.

Find Valuable Data You Can Collect

It may be a bit cliché, but it’s also true: your analysis is only as good as your data.

“If a company doesn't have access [to] good, large data sets, then no matter how sophisticated their applied machine learning algorithms are, their models and interesting research derived from those models will only be as good as their data sets,” says Nordigen data scientist Indra Ikauniece.

“Lots of innovation in data science comes from applying existing methods to new types of data, rather than inventing new machine learning models.”

If you’re working in data at a small company, chances are there are opportunities — probably lots of opportunities — for data collection that nobody has taken advantage of. On a company level, other team members may lack the skill set or the time needed to collect and properly store data for analysis, and there may also be valuable industry data out there that your company isn’t paying attention to.

Your embeds with business teams should give you a good idea of what data the company already has and how it’s being used, so the next step is to ask questions like:

  • What customer data are we not collecting?
  • What company data are we not collecting?
  • What industry data are we not collecting?
  • What data are we collecting but not using?

Again, at a small company where you’re part of a small data team, or perhaps the only data analyst or data scientist on the payroll, there are likely to be a lot of answers to these questions, so your challenge is to prioritize.

This is why embedding with business teams is so helpful and why it’s a great first step — once you’ve done that, you should have a good idea of what the most critical priorities are, and that should help you decide what kind of data to focus on first.

Learn Customer Behavior

One specific area of data collection and analysis that’s worth your time at almost any company is customer behavior.

Chances are your company is collecting some data about what customers do and how they interact with your company and its products. At a small company, chances are also good that this data is being under-used, and that it could contain valuable insights.

Exploration and experimentation are key here, because there are probably insights in this data that no one at the company has thought to look for. Your goal should be simply to dig into the data to try to understand more about how customers interact with your company and its products.

Try not to give much credence to assumptions about what customers do or why they do it before you’ve looked at the data. What’s considered “common sense” at the company could still be wrong, and it is wrong more often than you might expect. 

Analyze Competitors

At small companies in particular, it’s possible that nobody has spent much time thinking about the competition. When there’s lots of work to be done and few people to do it, people tend to focus on internal execution. They may not be thinking about competitors at all.

This is a potential “blind spot” that, as a data scientist, you should keep an eye out for. It’s unlikely you’ll have access to internal metrics from any competitors, but there may be public data, industry data, social media data, etc. that’s still available to you. And from this data, you might be able to identify strategies that work well for your own company to copy, weaknesses in a competitor for your company to exploit, or other opportunities.

Even if you don’t find anything, digging into external industry and competitor data will give you a more well-rounded understanding of your company’s context within its business niche, and this will likely make you a more effective data scientist in the long term.

Look Out For Problems

When you’re working closely with the business teams at your company, you’re likely to hear about the problems each team faces. One of the key ways a data scientist can add value is by identifying the causes of problems so that they can be resolved.

While you’re exploring the data, it’s also quite possible you will find problems that nobody was aware of, or that your models will predict future problems the team hasn’t seen coming. These are also great ways to add value, assuming you have the communication skills required to convince management to do something about the problems.

Encourage Data-Driven Culture

If you’re on a small data team, it’s likely that one of your roles will be as a de facto data educator for other folks in your company. You can add value by showing them how using data to make decisions benefits their teams. You can also help them set up regular data reporting and analytical systems so that they’re empowered to use data on their own, without having to rely on you.

One particularly important thing you can do is help translate company goals into metrics that can be reliably measured with data. You are likely to know better than anyone else at the company what data is actually available and in what ways it can be meaningfully analyzed, so you’re often going to be the best-positioned person to take broad directives, turn them into measurable metrics, and then measure them and predict future performance.

Keep Developing Your Skills

Finally, don’t rest on your laurels! Data science is a fast-changing field. If you want to keep your company on the cutting edge (and set yourself up well for the next job when the time comes) you need to keep your skills sharp. You need to stay engaged with the data science community to learn about new research and new tools that might help your job performance.

At Dataquest we have a large and fast-growing directory of interactive online data science courses that can help you add to your skill set by (for example) picking up some R in addition to your Python, or adding some Data Engineering expertise to your resume.

We also have a weekly newsletter that’ll keep you on top of the latest happenings in data science. Sign up for a free account to get a copy.

You should also keep in touch with the broader data science community by doing things like participating in social media discussions, attending conferences and meetups. And of course, you should always be working on and sharing data science projects in your spare time as you find projects that interest you or skills you’d like to get more practice with.

(Note: this article is focused on helping you find valuable work to do as a data scientist if you don’t have much guidance from management about what they need. You may also encounter problems with communication and office politics, and we’ve offered solutions for those in a different article).

Free Data Science Resources

Sign up for free to get our weekly newsletter with data science, Python, R, and SQL resource links. Plus, you get access to our free, interactive online courses!