Without well-known, comprehensive project management approaches specific for data science, teams often combine elements from two or more approaches to cater to their own needs. Although combining multiple approaches can be challenging, they allow users to design approaches that fit their individual needs. Two common hybrid approaches in data science are bimodal and research and development.
Bimodal
Also known as Waterfall-agile, bimodal data science combines elements of waterfall and agile. At first glance, why not take the best elements of two different processes and combine them? A deeper dive shows that this could be challenging and results might suffer. However, there are specific circumstances, such as projects involving medical equipment data processing, where such an approach might make sense.
Research and development
Research and development data science approaches feature two general phases: a research phase which is largely unstructured and a development phase which tends to leverage standard software engineering approaches. As data science is often seen as a research endeavor, this approach could be a natural fit, especially for teams with research backgrounds. Yet, managing such projects can be challenging as this loose “research” framework can mirror ad hoc approaches.
