Data Science project management I

Data Science project management I

According to the Project Management Institute, there are over 16 million project managers in the world. That’s a lot! If you are engaging a new Data Science team there are a lot of chances that one long time developer speaks fluent project management. So, scope, requirements, visio graphs and documentation of the progress might land on the table of your newly built team.

If you work for a big organisation or even, not big, in one that in the past performed well with project management old rules, is not odd to think that the DS team might be forced to follow the same path. But that is not going to work out! Let me explain why…

A Data science team is primarily exploratory. That’s the Science in Data Science.

The whole purpose of exploring is that you don’t know what you’re going to find. Project management is a defined process. It requires that you fully understand your deliverable before you begin. Can you answer the business question at the beginning of a DS project? The answer is no, you cannot know the answer to the question, not even when you might be able to answer it.

A DS needs to react to data as exploration goes. Exploration might discover that the business question is wrong and if the business question is wrong, how a DS is able to answer it? Is like when you are looking for a place to it in a street full of good restaurants, you might walk by all the restaurants along the street, checking the menus, the environment, prices,…at the end you will eat in the most convenient restaurant according to your mood and starvation level. But if someone had approached you at the beginning of your walking to ask where were you were going to eat? you might not have been able to answer.

If and old-fashion management forces a DS team to create narrow scopes of projects and answer very specific answers, the team might focus on completing their mandatory milestones (wrongly or well). The team will be discouraged to focus on being data explorers, looking for interesting new patterns. So, at the end of the project, they and the company might have confirmation of already known or suspected obvious patterns but will ignore the possibility of new insights and new value.

A project goal must always give value to the company. If not, why are we doing it?