“When you’re looking at an agency’s core strategic data, it actually doesn’t change that often. The measures by which any organization makes a lot of core mission decisions also don’t change that often, but there is a lot of other non-core strategic data that also needs to be managed” points out James Bench, vice president of technology consulting services at Maximus.
It’s why data governance must be the critical foundation for ensuring that agencies can make use of data as a strategic asset, he said. “It all starts with data governance.”
We asked Bench to share how agencies should build that governance plan. He walked through what matters in setting a data governance strategy but also detailed three additional factors that agencies should consider mapping out to take advantage of information to improve operations: data integration, analytics and security.
Data governance: Take time to get it right
Agencies should spend time and energy up front to define the roles, responsibilities, policies and procedures for managing the core data within their organizations, he said. Bench defined core data as the agency’s “true strategic information, the heart of your information.”
It’s critical to determine this information because it will be the data that requires the most stringent governance, he said.
As for the process, “start at the core and determine what is the absolute critical data that the organization must stand behind. With this core data, define the data, policies, standards, controls, usage and access. Start there and work your way outward until you have covered the core data assets that need to be governed,” Bench said.
Each agency’s conceptual data model will want to include levels of data risk, what each dataset means to the organization and then a general governance strategy for the datasets, he said. That governance strategy should involve identifying the decision makers for specific datasets, how these groups or councils will be formed and how any decisions about the datasets will be made.
The goal of the data governance plan is transparency, Bench said. That way, an organization can avoid circumstances where data or insights based on data might be taken out of context or used incorrectly.
“Do you have to have 100% data governance before you can start using something? No, you do not,” he said. “But what you do absolutely have to do is put the caveats and the explanations about how that data was consumed, how it was processed, so you can see that lineage of where it was sourced, how you got it, how you generated your analytics — and then that transparency will build trust.”
Data integration: Determine a path for making data available for decision-making
To make data actionable, it has to be accessible for analysis. That seems obvious, but it’s important to build out an integration plan so that the agency can get to the actual use of data to drive better decisions, Bench said. How can agencies gain integrated access to the right data?
“Data integration covers the lion’s share of development and collecting of the information,” he said. “How are you going to move the data from Point A to Point B? How is it meant to be combined? And how are you going to store it.”
The approach can be centralized or decentralized, but typically an organization will do one or the other, but not both, Bench said.
“A classic data warehouse is an example of where you’re taking a centralized approach to collecting all the information in one location, but it’s also curated and well organized (or structured). The data is cleansed, and it’s ready to use outside of the gates of the organization,” he said. Most reporting and analytics tools can consume data easily and quickly from a data warehouse, he added.
Another centralized approach is a data lake. In this case, the data is simply amassed centrally but will be combined and include both structured and unstructured data and often not be deeply organized. Some prep work will likely be necessary for reporting and analytics use of this data, Bench said.
New decentralized approaches involve the creation of data lake houses, which pair the benefits of a data warehouse and a data lake — think of multiple purpose-built warehouses taking advantage of the information stored in the data lake.
There is also the data mesh approach, he shared. A data mesh integrates data storage locations across the enterprise and then manages data by domain type, so in an agency that might include lines of business, human resources, specific government services and the like. It puts the data hygiene more on the organization that creates the data and less on the IT team.
Bench advised, “Try not to mix and match too much — only because your governance strategy must align with your data use processes. Plus, how you’re going to take care of or maintain your data varies and how you ensure data quality will vary. You don’t want to be spread too thin.”
Data analytics: Ask the right questions and lean into artificial intelligence
Increasingly, assistive technology that can augment the workforce will help agencies be able to take advantage of data, apply machine learning and artificial intelligence, and leverage more predictive analytics, he said. That’s because these advanced technologies have made it really easy to consume complex and large amounts of data easily, Bench said.
“All of that capability has hit the accelerator. There’s a lot going on,” he said. “It’s a fun time to be working in this field and helping agencies to uncover and work through their data governance strategies.”
But to take the best advantage of increasingly vast datasets, analyze the information and derive insights, agencies should also focus on asking the right questions for their analytics, Bench advised. What really helps, he said, is to define the correct use cases, which can help flesh out how the data is meant to be used.
Start by working your way backward from the mission, he said. “What are your mission drivers? What are the factors that show things are improving on the execution of mission? What data or information has helped support decisions so far? What data, if you had it, would make the decision process better? Which decisions can be made without a human in the loop?”
By taking this approach, an agency can help build the process muscle that it will need to be a smart user of its data, Bench said.
Data security: Develop more nuanced security controls
“The security of the data is, of course, absolutely critical,” he said, “because there are continual attacks attempting to extract government information.”
But the challenge in using federal data for decision-making comes from how it’s categorized because of privacy concerns or classified for Defense Department and Intelligence Community security concerns, Bench said. “For instance, in a large dataset, sometimes only a few elements or the combination of a couple elements is really what makes it classified.”
Agencies will need to tackle this challenge head on rather than simply locking down entire datasets if they want to take full advantage of data for decision-making, he said. A chief reason arises from the need to exchange data with other agencies to deliver services to the public effectively, Bench pointed out.
By de facto locking down entire datasets, “you’re really limiting what you can actually share with another agency or whether someone can draw really useful insights,” he said. Instead, Bench suggested that agencies develop more sophisticated risk profiles and set controls around specific data elements rather than entire datasets.
“Then, you need to make sure that from an execution standpoint — from a security, data security, privacy security standpoint — all of that is well understood and is all working together,” Bench said.
Part of embracing a data-centric decision-making culture requires thinking more about the challenge the agency is trying to address, and that’s a mindset change when it comes to data governance, integration, analytics and security, Bench said.
“Make sure that you’re staying focused on the right things versus saying, ‘This is the data I have, so this is the best I can do with what I have.’ That is a very limited way of trying to solve a problem.”