Using Data Analytics to Maintain Program Integrity

This blog post is an excerpt from GovLoop’s recent guide “Analytics in Action: How Government Tackles Critical Issues With Data.” In this guide, we share firsthand accounts from government employees at all levels who are using analytics to identify critical issues and find solutions. Download the full guide here.

For being a relatively small team, the Government Accountability Office’s Forensic Audits and Investigative Service (FAIS) has a mammoth task. The roughly 60-member division often reviews programs with multibillion-dollar budgets, providing Congress with forensic audits and investigations of fraud, waste and abuse; special investigations; and security and vulnerability assessments.

With a job so crucial and varied, FAIS demonstrates high proficiency with data analytics by necessity. Recently, GovLoop spoke with Seto Bagdoyan, Director of Audit Services at FAIS, to learn more about the nuances of the office.

GovLoop: What issues is your organization trying to tackle?

Bagdoyan: Our charge is to look at program integrity. Our work combination of performance audits, as well as investigative techniques, very frequently involves a data analytics component that helps inform our findings.

We look at pretty much anything and everything that involves the expenditure of federal dollars. We do focus a lot on means-tested benefits programs, and the FCC [Federal Communications Commission] Lifeline [program] is certainly one of those. It’s the most recently completed work. [Lifeline helps make communications services more a ordable for low-income consumers.]

GovLoop: How are you using data and analytics to address those issues?

Bagdoyan: We are dependent on having access to federal databases of various programs occasionally. We purchase access to proprietary commercial databases like medical licensure information, as well as the Postal Service’s comprehensive database of all the residential and commercial addresses in the country. We use that information in various ways to do data matching.

We also do data mining, which is essentially looking into the databases to see patterns. Our investigative techniques usually test controls to see whether the findings match up with our data analytics.

GovLoop: Can you share a time when your organization used analytics to drive audit findings?

Bagdoyan: Whatever we find in terms of the results of our data analytics, we usually make referrals to the agency that runs the program for them to take more in-depth looks at what we have and see if their review corroborates our initial suspicion that something untoward is happening.

Most recently we looked at the FCC Lifeline program. We got the list of current beneficiaries — roughly 12 million — and we matched them against selected databases of other benefits programs that they claimed in their applications as being their eligibility determinant, like Medicaid, or supplemental security income or food stamps.

GovLoop: What did you learn and what were the outcomes?

Bagdoyan: We found that at least 1.2 million subscribers did not appear on the programs that they claimed. So that to us is a significant flag that something might not be matching up appropriately, that these people may either [have] made an honest mistake, or they are not eligible for the benefit.

We’re in the process of making referrals to the FCC, and its inspector general, for them to take whatever action they deem appropriate with these individuals.

GovLoop: Can you share tips on how you got the data you needed and decided what tools/techniques to use?

Bagdoyan: By statute, GAO has access to any relevant data deemed essential to conducting an audit.

We work with the agencies, and in a matter of weeks — sometimes months — we get the data we need, and we start first and foremost doing data reliability assessments to see whether the data are even usable for our purposes. If they happen not to be usable because they’re deemed unreliable through our testing, then that itself becomes a story. If the data are of such poor quality, the problem for the agency is: How would they be able to oversee this program when they don’t have reliable data? We use things like SAS and Python and R, which are data analytics software programs, and we customize them to the specific needs of an audit. That can take time.

GovLoop: What was key to your success with analytics?

Bagdoyan: The willingness to do it certainly is one. It has to be part of a sound audit plan that has objectives, that also manages the scope and has the right people assigned, whether at the leadership level or at the subordinate level. At a minimum, I think it is imperative that if it is going to be a data analytics-heavy engagement, the analyst in charge be proficient in data analytics. Many of the audits that are starting up have significant data analytics components to them, so the team design is essentially the key aspect of success.

GovLoop: How are you creating a culture for analytics at your agency?

Bagdoyan: A couple of years ago, we had our road show for FAIS, where we went to each mission team and staff office at GAO, and we had an hour-long presentation of who we are, what we do, how we do it and a lot of our results. For most people, it seemed to be a well-kept secret.

Government auditors are very oriented to a routine approach, and they don’t shy too far away from tried-and-true audit approaches. So data analytics is a big disruptor. It may show aspects of a particular program that nobody really wants to know, so there are a lot of potentially ugly truths.

GovLoop: What advice do you have for others in government who want to use data analytics?

Bagdoyan: Don’t be afraid of it. It’s a powerful tool to be used in combination with other things.

Bring people along. Continue to show value [and] do the small wins.

GovLoop: What do you hope to do with data and analytics in the near term and long term that you cannot do today?

Bagdoyan: A data analyst’s dream is to be able to do more predictive work. Learning from the past and trying to project, for example, the prospect of fraudulent activity in program X over the next five years. Even retrospectively, we hope to continue to do deeper dives into the data that we have.

Leave a Comment

Leave a comment

Leave a Reply