Federal agencies have access to massive amounts of data that could transform how they interact with and support the citizens they serve. But before agencies can unleash the full potential of their data, it is essential that agency leaders establish comprehensive strategies for effective data use and management.
In its review of the Federal Data Strategy, the U.S. Chief Information Officers Council notes that government leaders need “a robust, integrated approach to using data to deliver on mission, serve customers, and steward resources while respecting privacy and confidentiality.” By focusing on five key areas, each of which requires its own established strategy, federal technology leaders can lay the groundwork for an enterprise-wide data strategy that will power evidence-based decisions, improve operational efficiencies, enhance customer service, and more.
5 Strategies to Unlock the Potential of Agency Data
Leveraging Data as a Strategic Asset
An agency’s data governance strategy is the essential framework that will guide its overall data management. In developing a comprehensive data governance strategy, agency leaders are encouraged to assess all roles, responsibilities, and policies related to how their organization handles data. Overly complex or rigid strategies can create roadblocks.
“There can be so many layers associated with governance that it becomes tough to navigate, which could stagnate your ability to move data forward. You need to make sure you have good decision-making strategies in place to avoid stagnation while also maintaining the flexibility to be nimble in your processes,” says James Bench, vice president of technology consulting services at Maximus. "And you need to have the right decision-makers who will take into consideration key issues that could put an organization at risk, especially regarding regulatory and data privacy concerns.”
To mitigate regulatory, privacy, and legal concerns, agency leaders should consider how to support metrics around data quality indicators, how to manage data risk and privacy metrics, and what data elements are most important to the agency’s mission. Such questions are at the heart of Maximus’ approach to data governance, which focuses on implementing and maintaining scalable, secure data architectures while keeping an organization’s long-term goals and technical strategy in mind.
“Having clear, mission-focused principles to guide governance decisions is critical so that value and outcomes don’t get overridden by oppressive processes,” Bench says, “because not every best practice is the right practice in every situation.”
Since the early days of modern data management, many organizations have relied on data warehouses to store information. The essential concept of a data warehouse — a term coined by IBM researchers in the 1980s — is a place to centralize your structured data. By creating this single source of truth, an organization could ensure that anyone building analysis or reporting based on that structured data knew that the quality was dependable.
While this way of maintaining data quality was standard for years, the sheer amount of data and the number of sources that organizations juggle in the present era of “big data,” has reduced the efficiency of data warehouses. Their structured nature leads to a level of rigidity that can make maintenance and upgrades difficult and costly. These complications have led organizations to seek other options for standardizing quality.
“You shouldn’t have to wait for a full, laborious requirements process or cleansing process to get data in your hands,” Bench says. “But that also shouldn’t preclude you from being able to come up with a good data quality strategy for the strategic information that you want to maintain and keep as one version of that particular truth.”
Essential data quality considerations include: Is the data accurate? Is it complete? Is it consistent? Is it timely? Is it unique?
“You can still tag data quality metrics to every single data source, and that allows for data consumers throughout your organization to at least have a good understanding about what they’re about to consume,” Bench says.
Even if data isn’t perfectly clean, “having quality metrics puts the right caveats in place, so it’s taken with an eyes-wide-open understanding,” Bench says. This includes accepting that data isn’t objective and knowing that potential biases or ethical skew in the source data may produce similarly biased analyses.
“As people consume that piece of information, there should be layers of understanding about what they are looking at,” Bench says. “For example, is it skewed? How was it created? Where was it originally created from? And what has happened to that data between when we sourced it and when data consumers access it?”
After assessing data quality protocols, the next step is to review existing data integration strategies. For example, where will data be combined and stored across the organization?
“Is it going to be a data warehouse? Is it a data lake? Is it a data lake house or a data mesh?” Bench asks. “Ultimately, the data integration strategy should be based on what your enterprise needs and whether you want centralized or decentralized ownership of data.”
Consider the flow of data through an organization as a pipeline. If data integration informs the flow of data into your chosen repository — whether warehouse, lake or mesh — the data analytics process happens when that data leaves the repository. At a high level, an analytics strategy involves establishing what business decisions an organization should make based on its data.
“Historically, analytics have primarily been used for reporting, providing insights through operational reports, or trending and analysis reports,” Bench says. “Today, many agencies use data for statistics work, AI, and machine learning projects.”
Of course, the quality of analytic outcomes will depend on an agency’s data quality. No matter how good an analytic model is, bad data will result in flawed decision-making. A good data analytics strategy will outline prerequisites, such as traceability, source transparency and clear communication about any potential caveats or biases in the data.
“The key is to question validity constantly. You need to understand the accuracy of your analytics when making decisions.” Bench says. This is where associating quantitative metrics is beneficial; you can see how your analytics are trending over time.
Security is ensuring that you understand both the risk profiles of your data and your access control policies really well. This is why good auditing practices are important; you can see who accessed the data
and when, as well as understand the state the data was in at that time it was accessed.
James Bench, Vice President of Technology Consulting Services
Learn how Maximus is assisting federal agencies to establish data strategies that improve efficiency, drive interoperability and support their missions.
are a centralized place to store structured and curated data and serve as a single source of truth. Data warehouses typically follow extract, transform, and load (ETL) data integration processes.
like warehouses, are centralized places to store data. However, unlike warehouses, data lakes store all forms of data, including raw data in its original format. Data lakes offer increased flexibility and agility than data warehouses, but potentially less quality assurance. Data lake integration processes follow the same steps as data warehouses in a different order: extract, load, transform (ELT).
Data lake houses
are a hybrid concept that combines the best of both worlds. It marries a data lake's ability to store both structured and unstructured data with many of the data governance and cleansing protocols associated with data warehouses.
are a more recent concept that takes a decentralized approach to data management. Rather than an enterprise-wide repository, individual business domains host their own data. Each domain is responsible for maintaining its data and making it accessible to the rest of the organization.
A strong data integration strategy doesn’t necessarily require an organization to pick just one of these options.
In a data mesh approach, for example, individual business units could still employ their own data lakes
“In most cases, you need a mix,” Bench says. “But there are tradeoffs you have to consider when developing your data integration strategy, because the investment in time to get value varies wildly between which integration strategy you build.”
“If only two elements of privacy concern are blocking a whole data source from being shared, that is a missed opportunity,” Bench says. Luckily, data management tools have reached a level of precision that makes it possible to access and use data in targeted ways.
Technology leaders can now build policies and implement tools that are sophisticated enough to maintain that optimal balance of security and flexibility by granting a range of very specific levels of access.
“Security is ensuring that you understand both the risk profiles of your data and your access control policies really well,” Bench says. “This is why good auditing practices are important; you can see who accessed the data and when, as well as understand the state the data was in at that time it was accessed.”
What data elements, or what combinations of elements, present privacy concerns? For example, one piece of information may not be sensitive by itself but becomes sensitive when coupled with other data.
Security is always a top priority for government agencies, and the best security practices strike a balance between protecting sensitive information and maximizing informed decision-making.
“The most secure system is the system that is not plugged into a wall, but once it’s plugged in, everything becomes accepting a certain amount of risk,” Bench says. “Data is the same way, in that the safest way to protect it is to ensure no one has access to it.”
However, data security is not about eliminating risk but rather it’s about determining how much risk is acceptable and for what gain.
Risk analysis starts with establishing the “what” and the “who” of data access:
Who is allowed to access each element? What is the criteria for this decision?
Leveraging Data to Break Down Silos
While each federal agency is at a different point on the road to using data as a strategic asset, they share a common goal of increasing collaboration between agencies. In fact, the Federal Data Strategy advises agencies to “assess and proactively address the procedural, regulatory, legal and cultural barriers to sharing data within and across federal agencies, as well as with external partners.”
Working with partners like Maximus on creating a data governance strategy, understanding the quality of data and analytics, and being deliberate about integration and security choices can assist agencies in breaking down those inter-agency silos. However, Bench also acknowledges that hurdles to interoperability are not necessarily technical.
“It's a regulatory issue, it's a privacy issue,” he says. “Better labeling of data, identifying which areas are actually sensitive and not sensitive, and creating interoperability so you can share that data, those are areas that could move a lot further and faster in the public sector.”
James Bench, Vice President of Technology Consulting Services
There can be so many layers associated with governance that it becomes tough to navigate, which could stagnate your ability to move data forward. You need to make sure you have good decision-making strategies in place to avoid stagnation while also maintaining the flexibility to be nimble in your processes.