4 steps for building a data-driven organization

4 steps for building a >As a data scientist, you have the power to change how your organization uses data.

Photo by Adeolu Eletu on Unsplash

When joining a start-up or any organization as a data science first-hire, there is a blank slate, and sure, the CEO has some priorities. Still, she probably hasn’t created a data science roadmap or data mission and vision statements. There is no centralized or distributed data science team structure because it’s just you. There are no managers. You generally report directly to the CEO or CTO.

You get limited direction. The consensus is that it’s your job to improve business efficiency, product quality, drive more customer leads, and increase client satisfaction. But, for the most part, no one can communicate to you how to do this.

You might want to panic at this point, but surprisingly, you are in a great position. The world is your oyster, and you have the opportunity to mold how data science is integrated within the entire organization. You are a de facto Data Science Leader.

“Leadership is the ability to amplify one’s capabilities through influence, supervision, governance, and inspiration to produce more significant impacts than what can be achieved as an individual.”

I was in a similar position, and here are the things I did to put the organization on a better footing. But, without surprise, all this is only possible if you have full support from the executive team. The following things are required to establish a />

Photo by krakenimages on Unsplash

Gain domain expertise through developing relationships.

My first step would be to gain domain expertise as fast as possible. Gaining expertise is essential because our purpose as data scientists is to solve problems. How can you develop the best solution if you don’t have a basic understanding of the field you are working in? Therefore, the acquisition of this knowledge should be rapid and complete. How to obtain this knowledge properly is where the concept of “learning to learn” comes into play (see my other blog), so that you can gain breadth and depth of understanding of complex topics within short periods.

The easiest way to gain domain expertise is to meet with the heads of every department to understand how they function and look for ways to make their processes smarter using Machine Learning, Data Analytics, or Automation. I mention all three because not all solutions require AI, and as the resident data expert, you need to communicate that difference while helping them find the correct answers. Only through fostering close relationships with internal stakeholders will you be able to move your organization toward being more />

Photo by Jen Theodore on Unsplash

Build a Data Science Roadmap

We need to formulate an attack plan to address the problems uncovered in our previous conversations with your colleagues. Our navigation tool will be a data science roadmap that defines the path to success. It will align all our priorities and give us direction. You should differentiate low-hanging fruit from the high impact yet highly complex strategic projects from your uncovered problems.

Let’s start with a complexity matrix (see below) and divide the projects you uncovered above into the correct quadrants based on complexity (effort) and impact (value). You want to focus on low complexity/high-impact projects to get easy wins that will turn heads in the C-suite and show department heads that you are working to unburden them.

The goal is to have a mix of 60% Leverage items, 25% Strategic items, 10% Bottleneck items, and 5% non-critical items per unit of time (month/quarter/year). Convert this into a visual representation and share it with key stakeholders. All this must also align with the larger strategic goal of the organization. Establishing a roadmap will provide a path forward for strategically improving the company’s bottom line.

Photo by Randy Fath on Unsplash

Build a data science playbook, then build the actual data products

It would be best if you established how data science is going to be performed within your organization. A playbook describes the roles and responsibilities (process) required for a data science team to deliver a successful data product. Model your playbook after the rules outlined in the blog post on “How (not) to fail at your data science project.” There are five phases: Scoping, Research, Development, Productization, and Peer Review. Taking the knowledge obtained above and your playbook, you should incorporate everything into an actual product that fits the stakeholder’s needs.

Photo by Cathy Mü on Unsplash

Nurture your >

Data Literacy training to promote a more analytically bent mindset

Teach technical workshops (such as python) during your lunch break

Have Lunch and Learns, where you discuss the latest developments in your department or field, as well as any excellent topic you can think of

Meet with department heads quarterly to sustain your repertoire

Ask colleagues their opinions on problems you are trying to solve

Final Thoughts

The article is essentially about exacting more influence on your organization than what is typically reserved for your position. This article applies to every employee. You wield more power than you think and should seize the opportunity to be a significant influence on your company’s bottom line. Even if you aren’t the first hire, you don’t have to hold a leadership position to do these or similar things. If you already have leaders with these responsibilities, you should find gaps in their leadership and fill them.

​​”Opportunities are usually disguised as hard work, so most people don’t recognize them.” — Ann Landers.

What do you think about this process? Do you think you can influence the overall trajectory of your organization? Leave a comment.

Please subscribe to receive emails about my latest articles. Also, if you liked this article, please give it a clap and share it with a friend.