This article is an excerpt from GovLoop’s recent guide, “How You Can Effectively Use Data Analytics.” Download the full guide here.
Accumulation of data is common for any organization — especially those in state and local governments — and can be difficult to translate into a data science project that is well-defined and properly implemented.
Through its DataSF program, the city of San Francisco strives to break down barriers to accessing important data by focusing on the structure, design and visual aspects of data programs. DataSF assists agencies within the city with using analytics to improve operations and offer better services. In an interview with GovLoop, Joy Bonaguro, the first Chief Data Officer (CDO) for the city and County of San Francisco, identified common challenges for data leaders in harnessing a data science program and best practices to combat those challenges.
When Bonaguro joined DataSF in 2014, the CDO role was poorly defined in the public and private sectors. To better refine the focus of San Francisco’s data program, Bonaguro turned the emphasis to overcoming obstacles to data use. “Our data strategy has been to systematically attack barriers to data use,” she said. “We didn’t jump into a data science program, because if we did that, we would have been met with skepticism.”
Instead, Bonaguro stressed that CDOs should understand their local context. “San Francisco has many analysts distributed across our departments,” Bonaguro said. “We have 52 or more and many of those have strong policy, planning and analysis skills.”
To fully harness those skills and understand the everyday barriers to data use, DataSF instituted an annual survey, beginning in 2014.
One barrier is differences in the level of core data skills among city analysts. To reduce these discrepancies, DataSF created a training program called Data Academy. “We co-founded Data Academy with our Controller’s Office to lift and level data literacy skills across the city,” Bonaguro said. “We offer about 20 peer-taught classes that are free and that we regularly teach.”
Bonaguro found a way to make these classes accessible for any analyst. “We don’t charge for them because competency with use of data shouldn’t be a function of your department’s budget,” she said. DataSF minimizes expenses through the participation of volunteer teachers, using consistent training materials, and light administration and management of the system. “We’re saving about $1.7 million a year in saved staff time due to improved skills,” Bonaguro said.
“We have a gateway to additional work and projects,” Bonaguro said. “It gives you an opportunity to work with and touch lots of different people across the organization. When they see you as someone who’s creating value for them, they want to work with you on other things.” As Data Academy gained traction, other jurisdictions started to create their own versions of the school and reached out to DataSF for help.
Another common obstacle to data is what Bonaguro called “the Sisyphean reporting cycle,” or a barrier to an organization’s capacity for more advanced data use given the opportunity cost.
“You’re constantly writing and authorizing reports, and when you have analysts exhausting their time on them, they’re not doing high-level analysis,” Bonaguro said. “An open data portal can help with some of that. With our self-service dashboarding strategy, we can empower analysts to automate the reporting requirements in a way that’s flexible and can address ad hoc data requests.”
Once analysts free up time by automating certain processes, they have more availability for advanced tasks like active performance management and high-level analytics.
To support improved metrics and performance management, Bonaguro’s team collaborated with the Controller’s Office to create tools and guidebooks about what it means to have useful and valuable metrics. “We put together tools and guidebooks into a resource collection on performance management and metrics,” Bonaguro said.
When considering whether to offer data science services, Bonaguro noticed that a common issue in the field was implementing changes based on data science insights. “A pattern that our team observed was that a lot of projects struggle to get implemented and completed,” she said. “Anyone can model and run cool stats on a large dataset. The hard part is meaningfully answering an operational question and then helping that group implement the data insight you provided.
At DataSF, the emphasis is on user-centered design in the data science engagement. Before assisting clients, such as county and city departments and agencies, DataSF takes the time to understand their current business processes to help fuel ideas about implementing change. “It’s not about an academic report. You’re trying to implement a service change,” Bonaguro said. “We develop a project charter and constantly talk about the service change.”
Bonaguro and her team ensure that through the project charter and other solicitation materials, clients know from day one that the DataSF team will work with them to implement a service change based on the data science insights.
Lastly, DataSF helps clients implement service and data changes in a way that is understandable and feasible. “While we may use advanced statistics and modeling in the project, we tailor the end product to meet the needs of the end users. Sometimes that’s a script or model and sometimes it’s a user-friendly workbook that [the clients] can use in their business process.”
Following the implementation process, DataSF hosts a showcase where they present the work they completed with their clients and client testimonials using easy visuals such as Microsoft PowerPoint presentations and online articles.
Overall, Bonaguro believes that creating a product that is understandable and usable is key. “Being able to translate your insights into something that’s a useful tool or a useful business process is really essential,” Bonaguro concluded.