Conor Jensen works for Domino Data Lab, provider of the well-known and well-respected data science platform, as a Customer Success Manger helping customers make the most out of the platform by working through their people and process that surrounds it. Jensen’s professional history includes stints in the insurance industry putting together and managing data science teams as large as 60 staff members. At ODSC West 2018, he delivered a talk “Managing Effective Data Science Teams” which discussed how to build and manage an effective data science team that will become an instrumental part of an organization. The slides for Jensen’s presentation can be found HERE.
[Related Article: Give Unicorns a Break, It’s Time for Data Science Teams]
As the field of data science continues its accelerated growth, the need for managers to develop and lead effective teams is quickly becoming more acute. The talk covers how to select and develop team members, create a self-sustaining pipeline of analytics projects, make appropriate technology selections, and deliver real value to an enterprise. The presentation provides the means to learn from real world examples that have gone well, and those that have gone poorly, in order to bring actionable insights back to your organization. Jensen reports that “95% of the companies putting together a data science initiative are spending money and pissing it down the drain.” Clearly there must be a better way to engage the benefits of data science, and this talk provides a sensible path toward achieving this.
I appreciated the talk’s premise because I am in the process of putting together and managing a data science team right now for one of my consulting clients. As a mentor to the host company, my job is to seek out and evaluate teams to make sure they have the skillset required to build a custom machine learning application. After interviewing a couple of dozen companies thus far, Jensen’s tips can effectively help out with the process. I totally agree with the notion in the presentation that data science is NOT a freshman level activity, but at its core, a graduate level activity. Getting into data science is not for beginners as there are many prerequisites for establishing success. Having interviewed many companies offering data science services I can say that many were not really prepared and were simply trying to move laterally into data science from an adjacent field like web development. The lecture goes into detail what these prerequisites need to be:
- People – the data scientists and the people they need around them to be successful are of tantamount importance. I particularly enjoyed the discussion of what it takes to be a good “manager” of a data science team.
- Projects pipeline – a systematic method of project selection is imperative.
- Technology – having the right tools is critical to success: flexibility, fast (enough), and centralized. Use a Data Science Platform, and a formalized process (see diagram below).
Jensen also makes the case for how the recent trend toward “agile data science” is different from “agile software development,” and how it’s important to adjust what you’re doing and be quick to kill what you’re doing as the results are bearing out. I’ve certainly seen data science teams fail in their pursuit in spite of the fact that the results they’re seeing do not support the initial hypothesis for the project (which is sort of antithetical to the scientific method). Project stakeholders must understand that you need to be open to pivot on a project if the goals for the project can’t be realized.
This talk should be a prerequisite for anyone managing a data science project or any stakeholder of such a project in order to gain a real-life perspective for critical elements that will lead to overall success. To take a deeper dive into how to put together an effective data science team, check out Jensen’s compelling talk from ODSC West 2018
[Related Article: Building Data Science Teams: What Do You Need to Know?]
Managing effective data science teams involves:
- How to effectively select and develop your team members
- How to create a self-sustaining pipeline of analytics projects
- How to make appropriate technology selections