Every data science project ends up being a software system that generates scheduled reports or operates online. The world of software engineering already provides us with a multitude of software project management methodologies, so why do we need to reinvent a special approach for data science projects? The answer is that data science projects require much more experimentation and have to tolerate far more failures than software engineering projects.
To see the difference between a traditional software system and a system with predictive algorithms, let's look at the common causes of failure for data science projects:
- Dependence on data: A robust customer relationship management (CRM) system that organizes the sales process will work well in many organizations, independent of their business. A system that predicts the outcome of...
