Data scientists are taking on new importance as the challenges of turning raw data into an organizational asset become more and more daunting. Today, it is the data scientists rather than software developers who are likely to be called on for the task. They have the knowledge and the tools to convert our financial, operational, social and other data into useful information, which in turn helps federal agencies (as well as other public and private sector organizations) in many ways.
Examples of these benefits include faster processes enabled by intelligent automation; better detection of fraud, waste and abuse using machine learning; and greater tax compliance using insights into social behaviors to develop more effective messages.
To get the most value from the work of these experts, following are six best practices that can help you expand your organization’s capacity and capabilities and help you democratize data science across the agency:
- Create a culture and mindset that is focused on exploration. Increasing transparency and collaboration should be the target, as opposed to control. Providing quality data, analytical services and tools to the organization will improve the data scientists’ daily jobs. Consider implementing a citizen data science program that partners data scientists with domain experts and leverages self-service analytics to increase capacity and democratize analytics within your organization.
- Partner with IT early in the process. This will help the data science teams to better navigate the internal process to create a scalable solution once a model is ready for deployment. IT can help your data science teams leverage some of the DevOps tools and techniques that will help operationalize, maintain and operate the models.
- Demonstrate value to the organization, early and often. Data science projects should focus less on making cool discoveries, and more on demonstrating tangible results. While data science teams should be given the freedom to discover, there must also be a value for each use case at the start―preferably values that enable some quick wins.
- Operationalize data science with an eye to continuous improvement. Operationalizing goes beyond just scaling models; it also requires ensuring the processes they transform and support are sustainable and eventually institutionalized. The focus should be on enabling employees to make better decisions using the data-driven information provided by data science. Once models and processes are operationalized, periodic reevaluation is needed to ensure they continue to deliver value.
- Establish sustainable governance early. Governance success comes from the top and works its way down to the strategic, tactical and operational levels of an organization. Understanding and support from senior leaders is a must; a change management effort such as this requires shared responsibility and commitment across the organization. But governance should not be over-engineered—keeping policies and procedures simple and outcome-focused will reduce internal frustration and conflict. Try to leverage and integrate governance used on existing processes rather than establishing all new processes whenever possible.
- Elevate organizational maturity. Many organizations cannot get the full value from analytics until they reach certain maturity levels. This requires establishing the foundational practices for management, governance and education that will support your organizational big data, artificial intelligence, intelligent automation and other analytics projects. Once these practices are in place, data science teams can focus not only on predictive outcomes, but also on optimizing the future direction of the organization.
Following these best practices will help your data science teams become more efficient and focused, deliver greater value and spread data science principles across the organization.
For more insight on the value of data science to federal agencies, read my colleague Kevin Greer’s blog post, “Navigating the Federal Data Strategy.”