Best Practices for Managing Data Science at Scale

best practicesmanaging data science
Share

We recently published a practical guide for data science management intended to help current and aspiring managers learn from the challenges and successes of industry leaders. This blog post provides a distilled summary of the guide.

How to Advance Data Science at Your Organization

We at Domino are fortunate to have the opportunity to work with leading data science teams large and small—organizations from Audubon to Zurich. From this perspective, we get to see the 30,000-foot industry trends that capture headlines while also witnessing the less glamorous (but essential) work that happens in the trenches. I call this the “guts and glory” of data science today.

It’s an exciting time and we wanted to share some industry learnings which we distilled into a practical guide based on our experience with data science teams over the last few years. More than anything, we consistently hear a yearning to know what other people are doing in the space, what’s working, and where not to tread.

This guide represents our perspective on the sources of many of the industry’s challenges and the best practices honed by the leaders. Hopefully this demystifies the state-of-the-art and inspires you to further advance data science in your organization.

Below is the executive summary of the guide.

The Practical Guide to Managing Data Science at Scale: Executive Summary

The ability to manage, scale, and accelerate an entire data science discipline increasingly separates successful organizations from those falling victim to hype and disillusionment. While data scientists may have the sexiest job of the 21st century, data science managers probably have the most important and least understood job of the 21st century. This paper aims to demystify and elevate the current state of data science management. We identify consistent struggles around stakeholder alignment, the pace of model delivery, and the measurement of impact. The root cause of these challenges can be traced to a set of particular cultural issues, gaps in process and organizational structure, and inadequate technology.

Based on 4+ years of working with leaders in data science, such as Allstate, Monsanto, and Moody’s, we have observed that the best solution is a holistic approach to the entire project lifecycle from ideation to delivery and monitoring. Organizations that are able to develop a disciplined practice of iterative delivery of business value and self-measurement, while utilizing data science platform technology to support a hub-and-spoke organizational structure, can scale data science to a core capability, and accelerate the delivery of robust portfolio of models. While a complete transformation can take years, we suggest a “crawl, walk, run” approach to build momentum towards the ultimate vision.

The full guide is available for download here.