New York State saves $889 million by optimizing audit case selection

The New York State Department of Taxation and Finance annually processes 24 million business and personal tax returns and collects more than $90 billion in state and local tax payments.

Closing the “tax gap”

Which refunds should not be paid? Which tax returns should be audited and investigated? For the Taxation and Finance Department of New York State, having the answers to these questions is key to maximizing the amount of tax revenue collected and to confirming fairness. The vast majority of taxpayers are entitled both to the refunds they request and to a prompt response from the state. However, tax evasion techniques were getting more sophisticated and the state was not keeping up with the challenge of identifying refunds that should be denied—the “tax gap” between what was owed and what was collected was growing.

“We wanted to provide a fair and equitable taxing authority to the citizens of New York State,” says Tim Gardinier, manager of the Data Warehousing Unit for the New York State Department of Taxation and Finance.

Focusing resources where it matters

With implementation of the IBM® Tax Audit and Compliance System, the New York State Department of Taxation and Finance has transformed its approach from “pay and chase” to “next best case”—the pending case most likely to be questionable. The system proactively identifies returns that are outliers and focuses constrained audit resources on those returns that seem most unusual. This sophisticated system grew out of a long-term project to leverage information to transform the department’s operations.

Initially, the team created a department-wide data warehouse focused on improving audit case selection that gave access to the history of each case and each return. This trusted source of information validates and improves the business rules that drive the audit process. Created and managed by business experts, these rules identify the cases that should be considered for audit.

Over time the department has been able to use this data warehouse to analytically derive new business rules, find new patterns of questionable returns, and add these rules to the system. The accessibility of the data and the analytic tools has allowed the department’s business users to develop and maintain these rules themselves. Most recently the data has been used to develop predictive models. These behavioral models predict how likely a return is to be questionable, allowing the system to prioritize the cases that are least likely to be eligible. The system automatically rejects those refund requests that are clearly ineligible, preventing these cases from even entering the audit process. It then uses predictive models and business rules to make a decision as to the priority of each remaining case. As case officers complete each case, the system presents the next best case. This helps to ensure that the department’s resources are always focused where they will make the most difference. With limited resources and a finite time in which to make a decision—refunds cannot be held up indefinitely—the system maximizes the effectiveness of audit officers.

The department has steadily increased the amount of data involved in the decision-making process. The data now includes taxpayer’s past interactions with the tax department—allowing a history of honest interactions to drive future decisions—as well as data on businesses and employees, and data from other internal information silos.

Smarter Government:	Identifying patterns to predict fraud
	Instrumented	Captures history of each case and each return and feeds it into predictive and behavioral models.
	Interconnected	Leverages information from across the department including taxpayers’ past interactions, data on businesses and employees, and data from other internal silos.
	Intelligent	The system analytically derives new business rules and identifies patterns of questionable returns to effectively prioritize audit cases and automatically reject ineligible refunds.