Jobs In Data logo

Jobs In Data

Agile in Data Science: A Pragmatic Approach

Updated on

Agile, a term inherited from the realm of software development, has enjoyed wide acceptance across different industries and disciplines, including the complex world of data science. While many sing praises of the Agile approach with its efficient, iterative processes, it's essential to remember that its success isn't universal. Like any tool, Agile is only as good as the hands that wield it.

The Significance of the Right Team

Surprisingly, many seasoned leaders might argue that with the right team, you don’t even need Agile. It's a simple equation - a great team can conquer projects irrespective of the methodologies used. In contrast, a poorly functioning team, regardless of how sophisticated the methods at their disposal are, will find ways to struggle.

Teams that aren't fully invested can misuse complex systems like Agile, using subjective story points or complicated workflows to hide inefficiencies. It is therefore crucial for leaders to focus on acquiring and retaining the best talent, creating an environment where potential problems are nipped in the bud, and the need for extensive systems like Agile and Scrum, with their associated meetings and coordination, is eliminated.

Practical Agile Approaches for Data Science Teams

That said, some Agile practices have found favor with data science teams, especially when adapted to suit the unique challenges this field poses. Here are some examples:

1. Quarterly/Weekly Working Rhythm:

  • Teams work towards quarterly goals, which are set in alignment with the product team.
  • Weekly meetings occur every Monday morning to sync on current topics.
  • Additional meetings are held to discuss research directions per project or address emerging urgent matters.
  • Sharing successful ideas and experiments is emphasized due to the typically low success rate in data science experiments.
  • A Kanban board is used to monitor progress and facilitate communication with stakeholders.

2. Generic Kanban for Data Science:

  • The project status is tracked using simple Kanban tags: "to-do", "in progress", "on hold", and "completed".
  • This system provides clarity to all stakeholders about the current stage of the project.

3. Kanban for Data Science Publications:

Various stages of writing data science publications are managed using Kanban. These stages include:

  • production
  • production QA
  • presentation
  • presentation
  • QA for data tables
  • production and production QA for charts
  • writing
  • writing QA
  • preview website QA for the online article divided into paragraphs for assignment

The Advantages of Agile in Remote Work

The transition to remote work environments has also demonstrated the benefits of Agile elements. Remote Data Scientists are reporting:

  • A positive response to the structure & rhythm of Scrum.
  • The value in maintaining a record of work done.
  • Clarity about future tasks and what they need to work on next.
  • The social benefit of morning stand-up meetings, which play a key role in reducing feelings of isolation in remote work settings.


In conclusion, while Agile methodologies like Scrum and Kanban can be advantageous for data science teams, their effectiveness is reliant on the team's quality. The right team can work efficiently irrespective of the methods in place, while a team that's not up to par will struggle regardless. As such, the focus should always be on building and nurturing the right team, who can then pick and choose the best from methodologies like Agile and make them their own. Remember, methodologies should serve teams, not the other way around. Agile can certainly be a great tool, but it's not a magic bullet. It's a part of the toolbox, to be used wisely by a talented and dedicated team.