Do Not Pay Business Center- Using Analytics to Help Agencies Prevent Improper Payments JFMIP May 2016

L EAD ∙ T RANSFORM ∙ D ELIVER Page 2 Moderator: Jeffrey Schramek, Assistant Commissioner, Debt Management Services, Bureau of the Fiscal Service, Department of the Treasury Dan Keenaghan, Policy Analyst, Office of Federal Financial Management, Office of Management and Budget Jenny Rone, Executive Director, Do Not Pay Business Center, Bureau of the Fiscal Service, Department of the Treasury Aaron Prose, Team Lead for Improper Payments and Debts, Fiscal Policy Division, Office of the Chief Financial Officer, Department of Agriculture Panel

L EAD ∙ T RANSFORM ∙ D ELIVER Page 3 Provide timely, accurate, and actionable information, about payees and payments, to assist agencies with eligibility verification in order to reduce improper payments. Provide clear and understandable information using data analytics, as appropriate, about the nature, causes, and magnitude of improper payments to a range of stakeholders to inform agency payment processing improvement. Provide timely, accurate, and actionable information about potential systemic improper payments to support activities by Federal oversight entities. Do Not Pay Business Center Goals

L EAD ∙ T RANSFORM ∙ D ELIVER Page 4 DNP’s Analytics Services CapabilityOutput Agency Payment Matching Matches of payments data with multiple data sources (including the application of program-specific business rules) to identify patterns, linkages, impropriety, and suspect data/payments Basic EligibilityComparisons of recipients/payees to entities included in a single data source (before award /payment or to monitor award/payment) Complex/Contingent Eligibility Comparisons recipients/payees to entities included on multiple data sources (before award /payment or to monitor award/payment) Anomaly DetectionReports describing payee or payment characteristics outside of normal (e.g., multiple payments to the same person or account; payment size significantly varies for similar payments) Fraud DetectionReports describing payment characteristics consistent with known or probable fraud scenarios Probabilistic Risk Assessment Risk scores created from the combined evaluation of eligibility, anomaly, and fraud analysis Descriptive modeling Smaller insights regarding payments and their attributes created from condensing “big data” Predictive modeling Models created to best predict the probability of payment impropriety and suspicion of fraud Prescriptive modeling Recommended courses of action (e.g., stop payments, data correction) showing the likely outcome of each decision created using descriptive and predictive models Advanced data matching algorithms Improved data matching via use of data standardization, predictive, deterministic, and probabilistic record linkage methods

L EAD ∙ T RANSFORM ∙ D ELIVER Page 5 Agency Insight Report (AIR) Report capturing key insights and findings from exploratory data analytics, which includes, but not limited to: –Data quality assessments –Data pattern analysis and profiling –Advanced matching –Anomaly detection and analysis Analytics focus project to assist in identifying potential in- depth analytics projects that could be of assistance to the agency –Approximately a three month turn around –Available upon request

L EAD ∙ T RANSFORM ∙ D ELIVER Page 6 Data Quality Assessment Data Quality is an assessment of the completeness and validity of payment data. The purpose of this section is to determine the root-cause of hindering factors in the matching process, including the presence of illegitimate TINs or missing names. Program Eligibility Data Source Total # of Additional Matches Total $ of Additional Matches Program ADMF1,524$1,075,117 Program B DMF635$537,559 SAM Exclusion254$322,535 Program CSAM Exclusion127$215,023 Total2,540$2,150,234 Additional Matches Using TIN-like Information from Payee Identifier

L EAD ∙ T RANSFORM ∙ D ELIVER Page 7 Data Pattern Analysis Pattern and Trend Analysis Program Total # of Matches Less Than $300 Total $ of Matches Less Than $300 % of Reduction in Total # of Matches % of Reduction in Total $ of Matches Program A3,769$456, %2.97% Program B2,386$276, %13.45% Program C1,231$121, %7.32% Total7,386854,046

L EAD ∙ T RANSFORM ∙ D ELIVER Page 8 Advanced Matching Advanced Matching applies advanced techniques to capture additional high quality matches that go beyond the standard matching process of Exact TIN + Name. The purpose of this section is to identify and perhaps score additional matches that may lead to the adjudication of additional improper or erroneous payments. Program Eligibility Data Source Exact TIN + Name Alteration or NicknamesSimilarity Score (>0.7)TIN Only Increase in Total # of Matches Increase in Total $ of Matches Increase in Total # of Matches Increase in Total $ of Matches Increase in Total # of Matches Increase in Total $ of Matches Program ADMF234$38,496509$150,4988,509$345,952 Program B DMF102$25,387267$70,6436,400$96,834 SAM Exclusion 56$207,49875$430,8654,618$789,214 Program C DMF103$1,467,964401$2,389,5679,843$4,234,078 SAM Exclusion 43$98,476275$124,8357,200$245,954 Total538$1,837,8211,527$3,166,40836,570$5,712,032

L EAD ∙ T RANSFORM ∙ D ELIVER Page 9 Anomaly Detection Anomaly Detection involves the detection of anomalistic patterns within the payment data to identify potential improper payments. The purpose of this section is to identify abnormal patterns and trends associated with payment duplication, multiple names associated with a distinct TIN, and other suspicious payment behavior.

L EAD ∙ T RANSFORM ∙ D ELIVER Page 10 Anomaly Detection Continued Program Same TIN with Two Different NamesThree Different NamesMore than Three Different Names # of Occurrences Total # of Associated Payments Total $ of Associated Payments # of Occurrences Total # of Associated Payments Total $ of Associated Payments # of Occurrences Total # of Associated Payments Total $ of Associated Payments Program A23,487210,853$245, ,698$12,578,936371,289$2,467,836 Program B10,36550,387$209, ,890$2,758, $994,387 Program C11,84685,005$278, ,357$4,172,075191,391$36,164 Total45,698346,245$734,0741,25411,945$19,509,387652,986$3,498,387 Program Total # of Associated Payments Total $ of Associated Payments % of All Payment Counts % of All Payment Amounts Program A994,395$47,302, %2.34% Program B4,592$23,493, %1.01% Program C598$503, %0.02% Total999,585$71,299,044 Multiple payee names associated with a distinct TIN Payments associated with potential duplication

L EAD ∙ T RANSFORM ∙ D ELIVER Page 11 Cross-Cutting Government Analytics Z Capability to look across the government payment landscape Can help identify potential overlapping/duplicate benefit payments

L EAD ∙ T RANSFORM ∙ D ELIVER Page 12 Cross-Cutting Results Number of ProgramsDistinct Individuals Total $ of Associated with Potential Payments 231,001$2,981,115, $831,264,114 48$53,210,101 # Individuals# Agency A Payments# Agency B Payments$ Agency A Payments$ Agency B Payments 1,1751,4251,715$35,277,457$6,455,877 Payments made to individuals (identified using TIN and full name) who were paid by both Agency A and Agency B Payments made to individuals (identified using TIN and full name) who were paid by multiple Agency Programs

L EAD ∙ T RANSFORM ∙ D ELIVER Page 13 Next Steps Following the AIR The Do Not Pay Business Center will provide recommendations for further investigation and possible next steps for continued research. Agencies may determine if findings and conclusions appear to be significant and require further investigation. Partnering with Payment Management with Fiscal Service further enhances research efforts performed and helps to build strong relationships that will only improve the detection of improper or erroneous payments in the future. Participation in information sharing and future projects is at the discretion of the agency. No additional payment information is needed.

L EAD ∙ T RANSFORM ∙ D ELIVER Page 14 Do Not Pay Business Center Contact Information donotpay.treas.gov Assistant Commissioner Executive Director Director, Outreach & Business Processes Director, Product & Development Senior Advisor, Data Analytics