Skip to main content Skip to Footer


Fraud Analytics for San Bernardino County, California

Fraud analytics is helping maximize compliance and minimize abusive behavior among participants in the CalFresh program.


The San Bernardino County Transitional Assistance Department (TAD) chose to work with Accenture to gain greater efficiency and effectiveness in identifying fraud and non-compliance among CalFresh participants. The team built analytic models that would strengthen detection, improve prevention and maximize insights. For example, the team would run models on the all cases, score and rank the cases and look deeper at the top 10 percent to see if there was a high number of known fraud cases appearing at the top of the list. After refining the model and changing algorithms based on the findings, the team saw a major leap in the “hit rate” of fraudulent cases—moving from 6.1 percent to 20.8 percent.

“The jump in hit rate illustrates the efficiency gains we can achieve by using analytics to detect fraud. We look forward to operationalizing the model and achieving even better outcomes in our program,”
Nancy Swanson, Director
San Bernardino County


The San Bernardino County Transitional Assistance Department (TAD) wanted to maximize program compliance while minimizing the fraudulent and abusive behavior among participants in the CalFresh program, formerly known as Food Stamps and federally known as the Supplemental Nutrition Assistance Program (SNAP). TAD was spending valuable time and resources on investigating fraud cases that were often not severe enough to pursue prosecution. The County recognized that analytics could help make the process more efficient, and thereby make the investigations unit more efficient in finding fraudulent cases.

TAD chose to work with Accenture on identifying fraudulent activities related to CalFresh for several reasons—Accenture’s human services experience (for example, on the Statewide Automated Welfare System Consortium-IV program), its robust analytics capabilities and knowledge of public service in California.


Download PDFDownload the full case study [PDF 252 KB]


TAD and Accenture set out to bolster non-compliance and error identification processes by:

  • Strengthening detection. Better detection of fraud would reduce the cost of non-compliance and error to TAD, complement existing detection capabilities to better target known patterns, increase hit rates and reduce investigation time.

  • Improving prevention. Enhanced prevention would help reduce costs for TAD by identifying non-compliance and error earlier in the lifecycle of a claim, while maintaining a high standard of customer service.

  • Maximizing insights. Sharing investigation results across key business units would help to increase the understanding of customer behaviors and non-compliance patterns across TAD.

The effort to create a model to better predict and identify known fraud began with Accenture meeting with the client to understand how TAD currently identified fraud and tracked investigations. Using that knowledge, the team developed analytic models and tested those models on the entire universe of cases, including fraudulent ones, to identify which model was performing the strongest.


As a result of Accenture’s Intelligent Processing Services pilot at TAD, the team has achieved the following:

Improved hit rate
A baseline of 6.1 percent was calculated based on investigations completed from March. The objective was to improve the fraud identification rate to ~6.7 percent (a 10 percent improvement). After running the new analytical model, the observed hit rate, based on the validation, was 20.8 percent.

Earlier identification of fraud cases to investigate
The current average is approximately nine months on CalFresh program aid before investigation begins on a fraud case. Based on preliminary results, the model identifies fraud at least three months earlier.

Process automation
Previously, TAD investigators would have to manually pull data from the statewide EBT system to track fraud. Now there is a batch process that auto-pulls data from the system, loads it into a database, runs analytics through a software-as-a-service model and then generates a scored list of cases. Accenture Intelligent Processing Services also provides a periodically refreshed list of likely non-compliant cases and basic investigative support data, along with a measurement process, to assess the lift and value gained.

Based on the success of the pilot analytical model, TAD intends to operationalize the model and embed it in current fraud investigation processes and procedures.