How to Frame a Government Data Analysis Project
Data is becoming a bigger part of every government information workers’ day-to-day job. According to several studies, about 30 percent of all government employees deal with data on a daily basis. And given that a recent GovLoop study cites 96 percent of public sector employees believe their agency is experiencing a skills gap, it’s imperative that more and more government information workers learn the fundamentals of data analysis to use in their daily work. It may take years to learn advanced statistics to do data analysis, but the framework below provides a jump-start on performing data analysis the right way.
The first (and usually most important) part of doing a data analysis project is to actually identify the problem. This is a key and often overlooked step. It’s easy as an individual contributor to take a request from a manager at face-value: time is often constrained and makes it almost a necessity. But the greatest skill an effective public servant can have is clarifying the problem that is trying to be solved and the intended outcome. Only then can data effectively be used to inform decision-making. Key components of a clear problem statement include:
- A concise problem statement
- List of stakeholders
- Scope limitations
- Deliverable format
Exploring a problem area can be the most fun part. After establishing what problem needs to be solved, an analyst can then spend time building out their understanding of the problem space. This could involve:
- Online research into how similar problems have been solved
- Interviewing subject matter experts on how the different parts of the problem interact with each other
- Discussing with colleagues what trends have been affecting the problem recently
It’s important to note that very little time is spent on data in this section. The goal of exploring the problem is to build out a conceptual model of the problem.
After establishing a firm understanding of the problem being addressed and the context around it, the next major step is to form a clear hypothesis to test. Although the quintessential “If…then…” structure will work, it’s not critical that a scientific hypothesis is constructed. Rather, you just need to establish a concept that you can test using data. This is key. The potential solution (or hypothesis) must be clear, address the problem statement, and be able to be tested with data.
Finally, the actual data analysis happens. This is only after the problem has been clarified, research has been done around the problem statement, and a testable hypothesis has been built out. Analysis could involve descriptive statistics, predictive statistics, exploratory analysis, or a variety of other activities. The input to this section of the project should involve data, and the output is insight on whether the hypothesis is supported or not. You should be able to make a confident statement on the decision you’ve made and why it’s supported by data.
Following all of this, you’ll need to create communications, visualizations, and reports using the insight from analysis to convince others of the right decision to make or action to take. It’s key here to think about the rhetorical triangle:
- Who are you speaking to?
- How are you communicating to them?
- How will you appeal to them?
You can then use data and insights from analysis to drive emotions through storytelling.
Performing data analysis can seem like a cumbersome, ambiguous task. The time and resource constraints on a project can often make it seem even more so. However, being diligent about executing each of the components of the framework can lead to faster, more effective data analysis work, even if you’re not a professional analyst.
Want to learn more data analysis skills? Sign up for our free online course through the Socrata Data Academy.