It's easy to overlook planning and preparation when you're preoccupied with experimenting on the latest technologies and data! Nevertheless, the process of how you write scalable algorithms is just as important as the algorithms themselves. Therefore, it's crucial to understand the role of planning in your project and to choose an operating framework that allows you to respond to the demands of your goals. The first recommendation is to adopt an agile development methodology.
The distinctive ebb and flow of analytic authoring may mean that there is just no natural end to the project. By being disciplined and systematic with your approach, you can avoid many pitfalls that lead to an under performing project and poorly performing code. Conversely, no amount of innovative, open source software or copious corpus will rescue a project with no structure.
As every data science project is slightly different, there's no right or wrong answers when it comes to overall management...