Predictive analytics for better government financial decision-making

Challenge

A federal agency trying to identify high-impact use cases for developing and implementing AI and machine learning tools across its operations turned to Guidehouse for assistance. After creating a vetting and scoring matrix to evaluate potential applications, our experts determined that a tool for forecasting unliquidated obligations (ULOs) could help the agency improve its financial decision-making.

At the time, decisions related to hundreds of millions of dollars’ worth of obligations were made manually on a quarterly basis—often relying on one individual to produce Excel-based reports. This created a single point of failure and limited decision-maker access to critical data. We proposed a solution using predictive analytics to create a process for repurposing funds more efficiently and with far greater transparency.

To be successful, we would need to access necessary data, understand existing agency processes and workflows, and build trust and confidence among end users and stakeholders—many of whom were unfamiliar with machine learning and skeptical of its reliability. Operating within the agency’s approved technology environment using tools like Databricks, Python, and Tableau would require careful planning and adherence to existing software policies. And we knew that the solution wouldn’t be adopted if stakeholders didn’t understand how it worked or questioned its accuracy.

Approach

To address these challenges, we adopted a highly collaborative, transparent approach with frequent touchpoints and structured reviews to fully engage stakeholders throughout the development process. Grounded in agile principles, our methodology included conducting interviews and database analyses, then cleaning client-specific data and migrating it into a modern engineering pipeline. Machine learning models were then validated and trained to produce accurate predictions to inform real-time decision-making.

The result was a ULO prediction tool designed to transform the agency’s reactive, quarterly analysis to a proactive, daily, data-driven approach. A new Microsoft Access database was also created to analyze tens of thousands of obligations. This process required high-level reconciliation across large disparate databases in addition to the manual work needed to investigate and categorize each obligation. 

Impact

With the new solution, more than 75 users gained access to on-demand data through Tableau dashboards organized into three views (probable, potential, and pending) that automatically refreshed every day. This shift from quarterly to daily data access dramatically improved the speed and cadence of decision-making by categorizing and ranking obligation transactions based on urgency. 

End users responded positively to the change, expressing their appreciation for how easy the tool was to use as they accessed specific records in real time and filtered data based on their needs. The dashboards provided a clear, intuitive interface that made complex financial data more accessible and actionable. Most importantly, this successful project implementation serves as a model for how predictive analytics can improve operational efficiency and financial stewardship in government.