Data science has become increasingly popular due to the enormous value it presents to businesses and organisations. As a result, more people are gradually discovering the many opportunities available in a career in data science and are excited at the prospect of making a difference using data.
However, according to VentureBeat, about 87% of data science projects do not make it past production. It is therefore important to understand the various factors that can affect the success of a project and put in place strategies that can help to improve project success rates.
Understand the Risks and Requirements
The availability of training data is considered one of the very big risks when working on a data science project. A data science project ceases to be a “data science” project if there is no data available. Similarly, when the data does not match the problem that needs to be solved, the outcomes of the project would be quite unsatisfactory.
Not understanding the requirements of a project can be another risk. You may end up solving a problem that may not be the main problem or fail to solve a problem altogether. Therefore understanding and defining the problem, and knowing exactly what the expectations and evaluation metrics are will allow you to tell when the problem has been solved.
Always Keep the Dialogue Open
Before any project begins, it is important to note that stakeholders may have very high expectations of the project outcomes and it is important to take steps to manage that. Some stakeholders may have an understanding of the projects and some may not. Be aware of the context in which you are working and always keep the dialogue open with stakeholders and the people involved in the projects as well as those providing the data.
Having great communication skills enables you to explain the project scope, the approach, and the expected outcomes in a way that is unambiguous. For stakeholders that have domain knowledge, this is relatively easy, however, when the stakeholders are not familiar with data science and its peculiarities, it may require a bigger effort from the data scientist, to communicate the technicalities to the non-technical stakeholders/members.
Keep Stakeholders Actively Involved
Continuously check in with your stakeholders to make sure you are both on the same page around key milestones and decisions. State and restate the goal in order not to lose sight of it. You should do this consistently as you go through the project.
Another strategy that can help align expectations is to conduct the evaluation metrics for the projects with the stakeholders involved. Keeping them within the process will allow you to win their buy-in and understand what you are measuring and why.
Including all stakeholders in the evaluation will also reveal a lot of your understanding of the problem and allow you to clarify your understanding in the early stages. While it is obvious that there will be tradeoffs, it is still very important to talk about them openly.
In a nutshell: Understand the problem. Solving a very difficult problem well is not worth it if it is the wrong problem. Communicate. Keep communications open and transparent, and reiterate goals all the time. And lastly, take your stakeholder along the journey. Projects will only be successful if the key stakeholders are on board and satisfied with the outcomes.