There are many articles about the skills needed to be a data scientist vs. a data analyst but there are few that tell you the skills needed to be successful – whether it is getting an exceptional performance review, praise from management, a raise, a promotion, or all of the above. Today I’d like to share my firsthand experience as a data scientist vs. a data analyst and what I learned to become successful.
I was fortunate enough to be offered a data scientist position without any experience in data science. How I managed this is a story for another time and my point here is that I only had a vague idea of what a data scientist did before accepting the job.
I was hired for my experience building data pipelines because of my prior role as a data engineer where I had developed the predictive analytics datamart used by the data science team.
My first year as a data scientist involved building data pipelines to train machine learning models and deploy them in production. I kept a low profile and wasn’t involved in many meetings with the marketing stakeholders that were the end users of the models.
In my second year, the data science manager who was the liaison with marketing left the company. From then on I was the point person and became more involved in model development and project timeline discussions.
As I interacted more with stakeholders I realized that data science was a vague concept that people heard about but didn’t quite understand, especially senior management.
I built more than a hundred models but because I didn’t know how to show the value, only one-third had ever been used even though the models had been requested by marketing in the first place.
A member of my team spent months developing a model that upper management thought would showcase the value of the data science team. The idea was to pitch this model around the organization after it was developed and encourage marketing teams to use it.
This turned out to be an utter failure because no one understood what a machine learning model was and couldn’t understand the value of using it. The result was months of time wasted on something no one wanted.
This brings me back to the lessons I took away to become a successful data scientist:
1. Set yourself up for success by choosing the right company.
When interviewing at a company ask about their data culture and how many machine learning models have been adopted and used in making decisions. Ask for examples.
Find out if the data infrastructure been set up for you to start modeling. If you spend 90% of your time trying to pull raw data and cleaning it you’ll barely have time to build any models to demonstrate your value as a data scientist.
Beware if you’re the first data science hire. This can be a good or bad depending on the data culture. You may find more resistance in adoption if senior management is only hiring a data scientist because they want to be known as a company using data science to make better decisions and have no clue what that really means. Alternatively, if you find a company that is data driven then you grow as the company grows.
2. Know the data and key performance indicators ( KPIs )
In the beginning I mentioned I had built the analytics datamart for the data science team when I was a data engineer. When I transitioned into my data scientist role I was able to find new features that improved the accuracy of the models because I had worked intensively with the raw data in my prior role.
When I presented campaign results I was able to show the models generated higher conversion rates, one of the KPIs campaigns were measured on. This demonstrated the value a model had on Business performance that marketing could relate to.
3. Ensure model adoption by showing the value to stakeholders
You will never be successful as a data scientist if your stakeholders never use your models to make business decisions.
One way to ensure adoption is to find a pain point for the business and show how a model can help.
I realized after talking to our sales team that there were 2 reps working full time manually going through millions of users in the company database to identify single license users more likely to upgrade to team licenses. There were a set of criteria used in the selection process but it was time consuming because reps had to look up one user at a time. With the model I developed, reps could now select users with the highest probability to purchase and increase the chance of conversion in less time. This translated to a more efficient use of time by increasing conversion rates for a KPI the sales team could relate to.
A few years went by and I had been developing the same type of models repeatedly and felt I was no longer learning anything new. I decided to look for a different kind of role and ended up getting a position as a data analyst.
The difference in responsibilities could not have been more different compared to when I was a data scientist even though I again supported marketing.
It was the first time I analyzed A/B experiments and discovered all **** the ways an experiment could go wrong. As a data scientist I didn’t get to work on A/B testing at all because that was reserved for the experimentation team.
I worked on a wide variety of analysis that marketing had an impact on – from increasing premium conversion rates to user engagement to churn prevention. I learned many different ways to look at data and spent a great deal of time compiling results and presenting them to stakeholders and senior leadership. As a data scientist I had mainly worked on one type of model and rarely gave presentations.
My data analyst experience was very different from my time as a data scientist where the focus was developing models to optimize user conversion compared to the breadth of projects I worked on as a data analyst.
Fast forward a few years and these are the skills I learned to become a successful data analyst:
1. Learn how to tell stories with data
Don’t look at KPIs in isolation. Tie them together by looking at the business from a holistic point of view. This will allow you to identify areas that influence each other. Senior leadership looks at the business from a 30,000 feet lens and a person demonstrating this ability gets noticed when it comes time to decide on promotions.
2. Provide actionable insights
Provide actionable insights that is a solution to a problem. It’s even better if you are proactive to offer this solution without being told it was a problem in the first place.
For example if you said to marketing – "I noticed a recent decline in monthly visitors to the website". This is trend they may have noticed on a dashboard and you would’ve offered no value as an analyst because you were stating an observation and not providing a solution.
Instead, you can look into the data to find the cause and offer a solution. A better example to marketing would be – "I noticed we had a drop recently in visitors to our website. I found the source is from organic search due to changes we made recently that caused us to drop in Google search rankings." This demonstrated you were tracking company KPIs, noticed a change, researched the cause, and offered a solution to the problem.
3. Become a trusted advisor
You need to become the first person your stakeholder goes to for recommendations or questions about the business line you support.
There is no shortcut because it takes time to demonstrate these abilities.
The key is to consistently deliver high quality analysis with very little mistakes. Any error in calculations will cost you credibility points because the next time you present an analysis they may wonder if you were wrong last time are you wrong this time. Always double check your work. It also doesn’t hurt to ask your manager or colleague to look over your numbers before presenting if you have any doubts about your analysis.
4. Learn how to communicate complex results in a clear manner
Again there is no shortcut to be an effective communicator. This takes practice and you’ll get better over time.
The key is to figure out the main points you want to make and recommend any actions your stakeholders can make to improve the business as a result of your analysis. The more senior you get in the organization the more important it is to communicate well. The ability to convey complex results is an important skill to demonstrate.
I spent years learning the secrets to success as a data scientist and a data analyst. We define success in many ways and being described as an "amazing" and an "all-star" analyst is success in my eyes. Now that you know these secrets I hope your path to success, however you define it, comes sooner rather than later.
You might also like…
How I Used a Machine Learning Model to Generate Actionable Insights
My Unbelievable Move From Data Engineer to Data Scientist Without Any Prior Experience