Building a Data Team in Local Government

Using data and analytics in local government has gained traction in the past several years. As cities look to build and then scale their data operations, investment in people is required. But, just hiring a bunch of data scientists probably won’t work.

Because you likely won’t be able to hire dozens of new staff, hiring people who can do a little bit of everything I detail below is important. You’ll hire a data engineer, but that person should know a bit about project management. Your project manager should be able to build a data visualization. All of the people you hire, especially initially, should be passionate about the work. Doing this work is fascinating but also brings a lot of challenges. The subject matter expertise is also critical. So, finding someone who reads blogs about city planning isn’t essential, but it helps. Last, being compassionate and empathetic to work colleagues and people in the community is essential — you shouldn’t hire someone who will treat this as just another job. They won’t be successful.

Pardis Noorzad wrote a really interesting post entitled Models for integrating data science teams within organizations that I’ve referenced regularly when thinking about my team. We’ve built a mostly consultant model. It is centralized, works with departments based on needs and requests, sometimes builds products for the enterprise, and also serves as an evangelist for better use of data.

I’m going to focus on the skill sets and job functions I think are important for a data science team in government — and part of what I tried to put together on my own team. I covered a bit of my approach for getting buy in to build a team in this blog post from last year.

Key Pieces

In government, the assumption should be that the data is not clean nor necessarily easy to access. Most data, also, is location based (where is that pothole? At what address was that crime committed? Which restaurant has a permit for outdoor seating?) which means mapping will be important. Especially if you’re just building up a data science team, your government may not be quite ready to delve into predictive analytics, and instead just might want help counting things consistently. Hiring people who are excited to work on these kinds of projects will be key.

Data Engineer

When I first started with the City of Syracuse, I was a data analyst, and my first hire was also a data analyst. Realistically, most of the work was focused on pulling data from different systems and getting the data into a format others could use (like an Excel spreadsheet). Hiring someone who is well versed in preparing data for analysis — a Data Engineer — is now my recommendation as a good first hire. As stated above, hopefully this person can also do some simple analysis and visualization, but there will initially be much more focus on just getting data available for departments to see, outside of their source systems.

Skills to look for:

• SQL as it will be the key tool to extract data from different databases.
• Programming language like Python to allow for building the extract-transform-and load (ETL) processes.
• Business intelligence tools like PowerBI or Tableau to ensure understanding of how data needs to be formatted in order to be visualized.
• Database and data warehouse experience and additionally useful if there is some experience in the cloud and patience for legacy systems.
• Interpersonal skills as this person will need to work with data analysts to prepare data in the appropriate way, as well as the client departments who know exactly what the data means. Someone who isn’t willing to listen and learn won’t be successful, because they won’t understand the needs and requirements.

Data Analyst

Naturally, once data is extracted, made usable, and shared, there will be a desire for it to be analyzed. Initially the data engineer (and you!) can do some of this analysis, but the requests will pile up and the questions may be too challenging for staff in individual departments to answer. The data analyst can work to understand the questions and develop answers. This person will likely need to do their own data cleaning in addition to what the data engineer provides, and will need to give guidance to the data engineer about what they need to see from the data.

Skills to look for:

SQL because the data engineer cannot extract all of the data and it is a basic skill for any data staff.

Programming language like Python or R. I don’t care which and it is good to know at least a bit of both. Use these to build the analysis and share results.

Business Intelligence tools like PowerBI or Tableau to share results of the analyses. Building custom tools in R or Python is also suitable, though there can be limitations in some organizations.

Data visualization experience because telling a good story about the analysis is a critical part of the process, but a bad data visualization will confuse, overwhelm, and potentially share the wrong message.

Interpersonal skills as this person will need to work with the data engineer and project manager, as well as with departments that are asking the questions. This person may also need to help departments develop the questions and teach them how to understand the analysis and use the data tools provided in the deliverables.

Project Manager

This position becomes more important if you are able to build a partnership with external organizations that can provide interns, but you as the Chief Data Officer will quickly have too much on your plate to manage every piece of all of the data projects currently happening. A project manager can help to scope a project, ensure the team and the respective departments have a mutual understanding of timelines and deliverables, as well as the questions being asked. In our case, building a program where dozens of interns were working on projects at any given time, a project manager helps to keep track of everything going on and brings up concerns with the Chief Data Officer as well as the individual project teams.

Skills to look for:

Interpersonal skills are critical here because the project manager ends up needing to work with everyone at every step of the process and often times needs to have difficult conversations while still keeping the project moving forward in a positive way.

Project management experience is an obvious piece here as there are specific approaches and standards that help to ensure a project is moving forward smoothly. A certificate is useful, but not essential.

Some kind of data background is important so the project manager can speak the language of the team but also think about how departmental questions translate to data projects.

GIS Analyst

This is the one position I’m recommending that I was not able to add to my team — though it would have been the next step. Almost everything related to city data has a location associated with it, and mapping and spatial analysis will always be a need in any analysis. Your data engineer can probably geocode addresses or work with geojson files to help build a spatial dataset. Your data analyst can probably build an attractive looking map. But GIS is about more than that and is a special skill set. Just understanding how different projections impact your analysis is something worth having a GIS Analyst for anyway.

Skills to look for

Data visualization is important because information displayed on a map can very easily skew a person’s opinion. Think about how Presidential election results are displayed and then realize that the amount of space versus the density can drastically change the look of a map. Someone who knows how to present spatial information coherently and responsibly is very important.

GIS data tool experience is obviously important and the person should probably know how to use ESRI’s tools as every government I know of has a license for ArcGIS. Use of open source tools is also useful and important as sometimes ESRI is not the right solution.

Interpersonal skills, just like with every other person on the team, are important. The GIS analyst will work with the other members of the team and will need to interact with departments and present to senior staff. Communicating analyses is important because otherwise what value is being provided?

Partner with Universities

Very few, if any municipal Chief Data Officers, will ever be able to hire as many staff as they really need to do all the work that is required or could be done. One way to solve this that I pursued is establishing a formal relationship with one of our local universities — Syracuse University. I have been an adjunct professor for both the School of Information Studies and the Maxwell School of Citizenship and Public Affairs. Students would always ask if they could intern in my office and help work on projects. I was thrilled to have the interest, but also realistic that I did not have the bandwidth to supervise dozens of students all at once. In the program we set up, I hired a data project manager who served as the main point of contact with the University. The University provided several student project managers who worked with the data project manager on my team to scope and manage projects. The student project managers also oversaw those dozens of students. While not perfect, it has provided us with additional capacity and the students all get real-world experience working with messy city data and get to build data tools that are actually delivered and used by city departments.

Conclusion

Ultimately, with a team lead (like a Chief Data Officer) and these four full-time positions, plus a roster of interns, you can get a ton accomplished. Adding staff in the order listed above, I think, will help you get quick wins and continue to build as well. In bigger cities, the staff sizes are larger, but I think the skill sets would still fit in the buckets as listed above.

What are your thoughts on how I’ve built my team and would you do it differently?