Elite Analytics Develops Predictive Models Resulting in $400 Million Increase in Tax Revenue
A state tax board in the United States implements a predictive analytics solution that helps increase revenue by USD400 million during the first two years when it works with IBM Business Partner Elite Analytics, LLC and uses modeling software from IBM.
Part of the California Government Operations Agency, the California Franchise Tax Board (FTB) collects state personal income tax and corporate income tax in California. It is composed of the California State Controller, the director of the California Department of Finance, and the chair of the California Board of Equalization.
Each year, the California Franchise Tax Board (FTB) handles 2.4 million collections cases, 900,000 nonfiler cases and 250,000 audit cases. The complexity of these cases varies widely in terms of how much money is at stake and how likely the case is to result in payment. In the past, the California FTB had relied on the staff’s experience and general observations to determine which cases to investigate first, but this process involved prioritization that was often based on anecdotal evidence or observations that had become out of date over the years. With billions of dollars at stake, the California FTB needed a better way to prioritize nonpayment cases so that it could collect as much of the money it was owed as possible, as quickly as possible.
When faced with a list of more than three million delinquent or nonfiling taxpayers, it’s hard to know where to start trying to collect. Sorting the list alphabetically or by date seems too simplistic. And while starting with the cases that represent the largest sums owed appears to be a good idea, chasing down one large figure might take more time than securing an equal amount of money from five or ten other taxpayers. Faced with ever-increasing budget problems and an ever-expanding overdue tax bill, the California FTB brought analytics into the equation. By developing a set of models and predictors and feeding them into advanced modeling algorithms, the board can predict which cases are most likely to result in payment so that it can prioritize its workload accordingly.
As part of an overall tax-modernization project, the California FTB worked with Elite Analytics to develop an analytics solution that uses IBM SPSS Modeler and IBM SPSS Collaboration and Deployment Services software to help reduce the gap between what taxpayers owe and what the state receives. The SPSS Modeler software mines data from numerous sources, including internal applications used by the California FTB’s audit department, fraud-prevention department, collections department and nonfiler department. This information now comes together with data drawn from county records, motor-vehicle department systems and federal tax systems.
The client uses the newly correlated data to develop models and define important predictors, such as whether an individual or company is filing for the first time or whether the subject has filed reliably in the past. The solution feeds these predictors into several modeling algorithms in the SPSS Modeler software, which then identifies and scores the combinations that are most likely to result in payment. After the modeling software scores and prioritizes the cases, the SPSS Collaboration and Deployment Services software automates their management so that the California FTB can optimize investigative efforts and revenue collection. In the past, the board may have identified a wage-based case as having strong potential for repayment; however, the new solution takes this information a step further, adding other predictive attributes – location, for example – to wage-based cases to help determine much more definitively which cases the department should pursue first.
Real Business Results
- Generated a USD400 million increase in revenue during the first two years after implementation
- Led to a 300 percent improvement in the success rate for conducting business-entity non-filers
- Maximizes the department’s efficiency, so that it spends the majority of its time on cases with a high potential for remuneration
The solution pulls together data from a variety of sources, including the department’s own applications, motor-vehicle department records, databases owned by other state agencies and information submitted on federal tax returns.
The solution provides the department with a thorough understanding of each delinquent filer and the attributes that can influence a filer’s potential for making payment.
By applying the solution’s sophisticated algorithms to a wide range of data, the department can prioritize its workload according to which cases are most likely to lead to the successful recovery of unpaid taxes.
“The value of this project for us comes down to dollars and cents. We’re a revenue-generating agency, and we measure ourselves on that. In that regard, we’ve had excellent results.” — Jeff McTygue, Manager, Business Intelligence and Data Services Section, California Franchise Tax Board.