Maturity Levels for Data Science Capability
Data science is a tricky profession as it requires overlapping skills varying between data engineering, analytics, AI/ML, research, development and DevOps. This makes creating new data science teams is challenging for leaders to find experienced scientists who cover the whole spectrum (we call them unicorns or superheroes). The other challenge is to build the right process with the right set of skills that can work on the pipeline I an efficient way to deliver the models required by business.
There are several maturity models for data science capability, but one commonly used framework is the Cross-Industry Standard Process for Data Mining (CRISP-DM) model, which includes five maturity levels. Here is a description of each level, along with the enabling factors, best practices, and pros and cons.
- Ad-Hoc (Chaiotic): The data science projects here rely on super heroes and data science unicorns who cover the whole spectrum of data science, including data engineering, analytics, modelling, DevOps and runtime operations. At the chaotic level, organizations use basic software tools and lack specialized or advanced data science tools. The technology platforms are inconsistent, and there is limited collaboration within the team, leading to isolated data science efforts and unclear roles and responsibilities. Automation is minimal, with manual…
