The Opportunity Youth Research Project attempted to collect data from different organizations across Detroit who work with Opportunity Youth (OY) and use it to understand where OY are located. OY are typically defined as young adults 16-24 years old who are neither working nor enrolled in school or a vocational training program. Using traditional data sources to understand OY is difficult because they are inherently disconnected from systems that would normally collect data about them like schools.
We used work from a previous project with Microsoft that built out a structure for a future data collaborative called the Metro Detroit Data Alliance. One of D3’s long term goals is to develop a local data collaborative. Overall, we learned that this approach could help us achieve the goal of understanding OY, but it will require additional support to help organizations build consistent data processes.
Our goals were to:
- Improve the understanding of the OY in Detroit so that decision-makers can be more informed when planning programs.
- Test out the data collaborative structure using data from local organizations to find out if it is possible to use it on a larger scale and identify potential challenges to using it.
Throughout the project, we created a number of resources for organizations to use immediately in their work:
- This resource map can be used by other organizations to understand who is providing services to Opportunity Youth and what services they’re providing.
- To learn more about the process we used to make the resource map, check out this blog post.
- The D3Anonymizer helps organizations anonymize the data while maintaining a unique ID for each individual.
- A final report that details the entire process as well as findings from our analysis.
From the resource map, we identified 13 potential data partners. In the end, we received data from two organizations who work with OY. After receiving the data, we uploaded it to the Metro Detroit Data Alliance and began our analysis. Since we only received data from two organizations, the analysis couldn’t answer our original questions about the true number of opportunity youth in Detroit. However, we were able to answer two questions for the participating partners:
- Which areas of Detroit seem to be served most by the providers who contributed data?
- Which areas could benefit from additional outreach by these organizations?
In the process of answering these questions, we found five zip codes with high rates of opportunity youth but low levels of coverage by the two participating organizations. We detail these findings and the process in the final report.
Through this process of obtaining confidential data from partners, aggregating it, and analyzing it, we learned six important lessons:
- Bigger isn’t always better: While large organizations are working with more youth and theoretically have larger data systems and larger datasets, the bureaucratic processes of legal reviews and signoffs often limited their ability to share data.
- Align with existing efforts to assist organizations: Many organizations needed to spend additional time cleaning their data so we could use it. Organizations that already have grants to upgrade their internal processes can be ideal partners for this type of data collaborative.
- Knowledge about data and technology varies widely: This might seem like a no-brainer, but during the project we learned that not having a shared vocabulary can cause confusion. For example, we described the D3 anonymizer as an application, which it is, but that particular word led some partners to think they were uploading confidential data to the internet instead of running a desktop process that didn’t compromise the anonymity of their data. Expanding programs like Data University and AskD3 to build shared vocabulary and understanding in the nonprofit sector could help smooth data sharing.
- Data collection processes are very important: Or as the saying goes: “garbage in, garbage out”. It’s critical for the data being collected by staff on the front lines to be collected consistently for a data collaborative to be successful. For example, we found addresses that were collected in a variety of different formats which complicated data processing.
- Be patient: Data sharing is a very slow process when done properly and since we didn’t provide funding, we had to convince partners of the value in data sharing and be patient while they continued doing their own very important work. Even a small investment of time to make this data sharing happen might have to be spread out over multiple weeks or months since it was unfunded work.
- Support consistency: In addition to Lesson #4 about data entry, many of these organizations are working to balance the very different reporting demands of federal, state, local, and philanthropic funders. Supporting consistency within and across organizations is an attractive argument for partners to join in a data collaborative.
Overall, we discovered a number of challenges, particularly in the current ability of organizations to provide data to a data collaborative and their internal capacity to build out new systems to contribute data more effectively. However, with more data partners, there are promising new types of analysis that could be done across different organizations. For example, prior to this analysis, our data partners weren’t aware of the where other partners worked and which areas weren’t being served as much. At a larger scale, organizations could identify places that are underserved by all organizations and focus their resources in those communities for a larger impact. We are excited to pursue additional tests of the data collaborative in the future.