Revenue administers multiple areas of taxation, customs and duties for individuals and businesses, along with import controls on behalf of the European Union.
Declining public revenues in many countries are increasing the need for public service organizations to work more efficiently.
Ireland’s Revenue Commissioners sought to improve Revenue’s ability to identify non-compliance within the tax system while maintaining high levels of service and efficiency. Existing systems for transaction processing contained checks and validations to detect fraudulent behaviors and erroneous claims, and then route such transactions for review. However, some cases of non-compliance were not detected until after refunds were sent or credits applied. Consequently, Revenue faced the challenge of recovering monies in a “look back” intervention or audit.
In addition, unnecessary delays resulted due to compliant transactions being incorrectly targeted for review.
Revenue identified opportunities to improve and strengthen its processing with real-time risk scoring, predictive analytics and intelligence-driven business rules. Real-time risk scoring would not only identify more non-compliant transactions in-flight, it would route fewer compliant transactions for review, thereby improving efficiency and service.
Revenue chose Accenture for its experience using predictive analytics to provide a scientific understanding of non-compliant behaviors. Embedding analytics within existing processes and systems would help reduce non-compliance, both deliberate and resulting from inadvertent errors. In addition, Accenture demonstrated a strong track record of helping public organizations worldwide improve productivity, enhance service quality and pursue high performance.
A decision was made to focus first on the income tax system for employees, known as Pay As You Earn (PAYE). The PAYE system updates employment records and tax-credit entitlements, collects taxes deducted by employers, calculates liability and issues tax refunds. Accenture Intelligent Processing Services provided a real-time risk framework. Accenture deployed a team of management scientists and data analysts from its global analytics network to work with the Revenue team. The team proceeded in three phases—laboratory, factory, and tracking and feedback—within a six-month timeframe.
Laboratory phase— The project team began analyzing data from data warehouses, including internal, external and third-party sources. Members of the project team then applied analytical modeling techniques to the key variables, resulting in the selection of the most appropriate predictive model. Factory phase— The project team implemented the predictive model and associated rules into back-office and public-facing Revenue systems. Transactions entering through multiple channels, including phone and Internet, could be assessed in real time.
Tracking and feedback phase— The project team developed dashboards to monitor the effectiveness of the model and risk rules to reduce the number of false positives and enhance treatment tactics. As a result, Revenue has been able to enhance its analytical models based on actual experience and make continual improvements.
Accenture and the Irish Office of the Revenue Commissioners delivered the PAYE Risk Project on schedule in only six months, leading to rapid achievements in savings for the government.
The PAYE project has assessed more than 1.2 million transactions, and Revenue has recorded a measurable reduction in the number of undeserved refunds and credits. As of mid-2012, the model and business rules have blocked fraudulent or erroneous refunds exceeding €2.5 million, thereby covering the costs of the project in a matter of months.
Revenue anticipates further savings resulting from increased voluntary compliance as word spreads about the new system.
“Customers can make genuine errors in their applications, which can result in the granting of incorrect refunds or tax credits," explains Declan Rigney, head of Revenue's Planning Division.
“Unfortunately, some taxpayers may misrepresent their circumstances to reduce the taxes they pay or to obtain refunds. We continually seek to improve our effectiveness in addressing non-compliance through the adoption of new processes, technologies and data sources. The inclusion of real-time predictive analytics within our systems is an important progression in approach.”
When the model or business rules identify a suspicious case, the file is now automatically directed to a case worker, who determines if the transaction should be allowed or prevented. The number of false positives has fallen, with the “hit rate” (i.e. percentage of cases deemed erroneous or fraudulent, as opposed to benign) rising 50 percent.