By Rachel Wells, Senior Manager, Center of Excellence, DataKind
Have you ever wanted a behind-the-scenes look at how we work at DataKind? Interested in learning best practices and lessons learned from almost a decade of doing Data Science and AI for Good? Do you have a passion for providing pro bono data science support to a local nonprofit, but just don’t know where to start? Well, you’ve come to the right place.
DataKind is thrilled to announce that we’re sharing our Playbook! DataKind’s Playbook is our globally-accessible, living knowledge base that enables the reader to design, implement, and follow-up on a Data Science and AI for Good project following DataKind’s approach. The Playbook exists because the brilliant people across our global network doing great work identified the value that could come from a centralized guide that captures all the knowledge gained from that experience and shares it with each other.
Volunteers played a critical role in the co-creation process for developing the Playbook. The Playbook provides direction on how to do a Data Science and AI for Good project -- and to do it well; it also codifies the lessons we’ve learned from our failures and successes for use on future projects. We share it to enable everyone across the DataKind Universe to do the same! DataKind believes that it’s important that we not only do the work ourselves, but also that we teach others to do the work, and do it well.
What is DataKind’s Playbook?
Based on DataKind’s a decade of experience, the Playbook is centered around these three guiding principles:
- A living document: DataKind is an organization that’s constantly learning and evolving; the Playbook is designed to accommodate new ideas and learnings over time. Vetted and trained volunteers have suggested edits and will continuously evolve the Playbook and DataKind’s processes.
- A set of standards for DataKind projects: In the Playbook, you’ll see many articles associated with a specific “stagegate.” Stagegates are our minimum quality standards, or tasks that we make sure to do for every DataKind project. We want to ensure that all of our project partners have the same experience with DataKind. We document standards to ensure that all of our projects are high quality technically and ethically, with equity, justice, anti-racism, and co-liberation top of mind across our entire global network. Still, we don’t require every Chapter or project at DataKind to do everything exactly the same because the method should vary based on the project’s cultural context.
- An answer to the question “how do I do this?”: Stagegates answer the question “what to do”, while the associated Playbook articles answer the question of “how to do it.” This is so important, as we often talk about “designing with, not for” and “data risk mitigation” - but what do these phrases actually mean? How do we practically implement these stagegates step-by-step? This is what the Playbook solves for. For the “what to do” that’s outlined in the stagegates, the “how to do it” is described in the Playbook article. Anyone in the Data Science and AI for Good space has and will have the question of HOW to do a data science engagement with DataKind. Recognizing that all projects are different and processes need to be flexible, this includes a variety of options for how so that the method can be adjusted to the cultural context and specific needs of the project.
Who is the Playbook for?
The primary audience for the Playbook is DataKind volunteers to enable them in their work. It’s also intended for our partner organizations and those who are considering working with DataKind to provide deeper insight into what to expect when working with DataKind. Additionally, the Playbook is useful for data professionals supporting social sector actors more broadly and other ecosystem actors. This is because we want to enable others in the Data Science and AI for Good space to follow best practices as we all learn and grow together.
How is the Playbook designed?
Next week, we will share a blog that dives deep into our co-creation process for the Playbook over the last 18 months. For now, the biggest themes and takeaways that came across in our design sprints were:
- Nobody reads long documents.
- There’s a need for user-specific structure based on the volunteer role.
- Integrate documentation with workflow where possible.
- Balance global standards and necessary flexibility.
- Prioritize vision, values, and culture alignment.
With these takeaways and more in mind, we developed the Playbook to meet everyone’s needs. The Playbook development process involved refining how we define our project process and each step. We landed on the following six stages in June 2020, and we’ve been testing this project process framework ever since:
DataKind's project process:
- Discover: We discover if a potential partner organization would be a mutual good fit for DataKind.
- Design: We explore the data, dive into the partner’s challenges and goals, and design a project accordingly.
- Prepare: We recruit and prepare the right fit team of data scientists to complete the project.
- Execute: The team and partner work together to execute on the project, starting with a prototype that’s adjusted based on feedback and needs.
- Share: The team delivers the final version and documentation so the partner can increase its impact.
- Evaluate: Checking back in later, we reflect on what worked and where we could have done better.
So, we organized the Playbook into the six stages and refined exactly what each stage encompasses. Additionally, we enabled the Playbook user to filter for the individual roles someone might be playing on a project, creating “user-specified structure based on volunteer role.”
We also developed an internal tool that provides the volunteer with the minimum quality standards, called stagegates, for each step of the project process within each of the six stages. These minimum quality standards are outlined in a checklist for each project, which volunteers can use as a tool to manage their projects and find access to the linked Playbook resources for each step. This addressed the finding that we needed to “integrate documentation with workflow where possible.” Each Playbook article addresses one specific stagegate, so that the articles don’t get too long and remain user-friendly. This also helps “balance global standards and necessary flexibility,” in that the stagegates are the global standards, but the best practices each team decides to implement to complete the stagegates are flexible to reflect the needs and differences of their cultural context. Finally, what’s chosen as a stagegate clearly reflects the prioritization of vision, values, and culture alignment.
How can I get started?
The best way to get to know the Playbook is to dive right in yourself! Create an account at playbook.datakind.org and explore.
A huge thanks to all the amazing volunteers, staff, and partners who’ve been involved in the process of co-creating the Playbook. We couldn’t have done it without you!
As the leader of the Center of Excellence, Rachel codifies processes, ensures all projects are executed with the highest quality standards, and creates structures to facilitate experimentation and sharing of learnings at DataKind.
If you want to see the social sector transformed, join us in advancing the use of data science and AI to support causes that can help make the world a better place.
- Interested in supporting our work? Donate here.
- Interested in sponsoring a project? Partner with us.
- Interested in volunteering with DataKind? Look no further.
- Interested in working at DataKind? Apply now!
- Interested in submitting a project? Go for it!
As always, thank you for your support of this critical work!