A data-savvy staff that works to meet the requirements of program managers will deliver the insights agencies need to make sense of their data, a local government agency chief data officer says.
Disclaimer: The expressed opinions belong to the author and do not represent the opinion of the government of the District of Columbia.
With the advent of greater capabilities to collect data from smart phones, enterprise application software, smart sensors, etc., public sector agencies own ever increasing stores of information. “Government has reached a tipping point for data driven decisions,” according to the Deloitte report, "Data-Driven Decision Making in Government." To enable and promote such decision-making, many public organizations have been carving out data teams to lead the charge.
The focus of this article is on data teams that support local, state and federal level organizations and help them deliver better services by enabling data driven decision making. The following tips are based on my experience as a chief performance officer at the Washington, D.C., Department of Public Works and will be helpful to public-sector data team leaders in building and running successful teams.
Here are three considerations essential for a data team success:
1. Recruit people who are good with data -- not necessarily proficient in coding or a specific tool
When recruiting for data team members consider hiring candidates who are comfortable with data but may not be experienced in custom coding with R and Python or familiar with a business intelligence (BI) tools, such as Tableau, Power BI or Qlik, for example.
Budgetary, human resources and procurement constraints of the public sector do not always allow agencies to attract data scientists. According to Gartner's Staffing Data Science Teams report, “data scientists are currently difficult to find, and are commanding steep salaries.” However, “[i]n many instances, businesses do not require hard core data scientists because tool-based solutions [i.e. BI tools] have begun to automate certain repeatable patterns”
Team leaders may come across data-savvy candidates who may not have experience with the agency's BI tool of choice or may have no BI knowledge whatsoever. They should give such candidates a chance because modern BI tools are simple enough to be used by typical business /program /management analysts and can be learned. When recruiting, they should focus on the candidates who are 1) comfortable with data and math and 2) have good problem-solving skills. As a screening suggestion, have candidates interpret data visualizations and complete a data analysis exercise (and, yes, it can be in Excel) to answer questions.
The next step is coaching and challenging a new team member to grow. Data- and technology-savvy team members learn quickly. In few months they can be proficient and ready to produce high-quality reports and dashboards. We hired five senior- and junior-level analysts. None of them had experience with Tableau, but today they are experts who produce fully automated and hybrid dashboards of various complexity.
2. Deliver quick wins
Running a successful data team is about building an enduring data system with robust data collection tools and processes, delivering accessible databases and warehouses as well as developing agile and automated reporting capabilities. It is also as much about quick wins and short-term demonstrable results. A grandiose effort happening in the backend may not be visible to program or agency leaders.
Data warehouse development, data cataloguing or data cleansing automation might be great endeavors, but they may not be easily understood and appreciated. While running a long-term game, data team leaders should also find quick ways to create value and showcase their work to sustain goodwill and support of the executives and other colleagues. Such a quick win could be creating a visually appealing dashboard around an issue area that the organization has data on and actively promoting it among peers and the agency leaders. Case in point, at my current organization we deployed a dashboard on the 311 citywide public-works-related service requests. We had robust Salesforce data and a need to quality control service request completion. The dashboard caught on because it provided actionable insights into important service area that everyone in the agency cares about.
3. Engage, engage and engage
Data teams only exist to support programs of an agency. Team leaders should always keep this in mind. Visualization, reports and dashboards should be creating value by enabling data-driven decision-making to save time and money and improve customer satisfaction. (Of course there are reports for compliance and accountability, but those are not the focus here.) These data products must become integrated into regular work process. Program managers and staff should be using reports and dashboards to make informed and effective decisions.
Often data teams develop swanky dashboards with elaborate back-end architecture that sit unused collecting virtual dust. To avoid this waste and increase adoption, data teams must actively engage program managers and staff by promoting use of reports and dashboards. They can:
- Work with program managers and staff to identify questions they are trying to answer and then develop a draft dashboard or report. Have program partners review and provide feedback. When program staff are involved in the report or dashboard development process, their sense of ownership and familiarity with the tool will make them more likely to use it.
- Have data team members take ownership over the adoption phase for reports and dashboards along with the design, development and deployment phases. Data team leaders should set clear expectations making the team members responsible for engaging program managers and staff and collaboratively finding ways to use the reports and dashboards.
- Turn team members from business /program /management analysts to data advocates. Set clear goals for them to create dashboard/report use cases. It is better to build a handful of reports and dashboards that are frequently used than to produce a large number that will have no impact on an agency’s operations.