Photo by Skye Studios on Unsplash
How to Build a Data Science Portfolio
Creating and implementing an effective data science strategy has become a must for business success and survival. The challenge lies in choosing the right projects to build an impactful project portfolio a business's data science team is capable and prepared to implement.
Building an effective data science portfolio consists of four distinct and sequential phases: challenges and opportunities, pitch, scope, and plan.
Building a Data Science Portfolio, Phase 1:
Identifying Challenges & Opportunities
During Phase One, all relevant stakeholders generate numerous ideas to uncover new opportunities and challenges. It's essential to think about a potential data science portfolio in terms of the problem needing solving, not the solution. Present the challenge or opportunity the company is facing and allow the team to generate data science-based approaches.
Building a Data Science Portfolio, Phase 2:
Once Phase One is complete, stakeholders move to the pitch phase of portfolio planning, which consists of three sub-phases: preliminary scoping, pitching, and winnowing.
The pitch creator(s) begin preliminarily analyzing the pitch’s impact, the impact hypothesis (how the data science will transform into business impact), and the risks involved with the process.
All stakeholders present their pitches that have undergone preliminary scoping. Aptly titled the “elevator pitch party,” it is meant to be positive and exciting, where stakeholders feel free to share any pitches they have.
After receiving all pitches, business managers and primary stakeholders decide which pitches have made the cut and will move on to the next phase in the data science portfolio creation process.
Building a Data Science Portfolio, Phase 3:
The Scoping Process
Each pitch that's made the initial cut is analyzed technically, non-technically, or somewhere in between.
Technical Data Science Scoping
Technical scoping reveals potential data-based solutions and performance estimations based on data science metrics (e.g. accuracy or precision).
Nontechnical Data Science Scoping
Stakeholders formulate impact hypotheses for all relevant pitches, identify internal or external impacted parties, and assess the negative and positive impacts of both success and failure.
Somewhere In-Between Scoping
Some pitches will require input from all parties technical and nontechnical alike to estimate the cost of failure and the benefits of success.
Document, Share, and Save
During the long scoping process, new opportunities or challenges may present themselves. If you've documented the entire process and all pitches involved, your team won't have to reinvent the wheel to redo scoping that is already complete.
Building a Data Science Portfolio, Phase 4:
Finish building your data science portfolio by plotting the cost/benefit of each project, building a path by connecting projects, then selecting a course that achieves your goals.
Plot projects using a traditional cost vs. benefit analysis — add a range, examine size, and place risk or unknownness at critical points. It’s also very important to add time and cost saving dependencies, where one project can springboard to another easily.
Once all projects have been visualized it's time to chart out potential paths by connecting several projects. One or many paths can represent a project portfolio, where the member paths have a cumulative benefit and possibly cumulative cost.
The path you select will become your data science portfolio. Paths that consist of numerous intermediate projects are more flexible. Approaches that have fewer intermediate projects are more rigid.
Building a Data Science Portfolio, Phase 5:
Evolve Your Data Science Portfolio
The last step in the process is to evolve, which means staying aware of the changing environment and standing ready to pivot your portfolio as required.