Resource
SNAP Quality Control Resources for States
Highlights
The One Big Beautiful Bill (HR1) passed in July, 2025 dramatically altered how the supplemental nutrition assistance program (SNAP) is implemented by states. Among the changes, HR1 requires states to pay large penalties based on their SNAP payment error rate (PER). Based on existing error rates, 43 states will face penalties, with two thirds of states facing severe penalties of over $100 million a year.
In an effort to reduce penalties associated with SNAP error rates, Better Government Lab members are working collaboratively with a non-profit coalition to assist states in their response to HR1. Part of this effort includes leveraging analytic models built from historical state SNAP Quality Control (QC) data, to help agencies identify process or system improvements and prioritize cases for review - thereby improving SNAP error rates and reducing impending financial penalties.
This resource page includes data visualization tools, modeling tips, and lessons gathered from States to support the work of QC data modeling nationwide.
SNAP Data Modeling Workshops (Safety Net Response Network)
The Better Government Lab, SCALE Lab at Yale, and Aspen Financial Security Program are convening a group of research and data analytics staff involved in SNAP QC data modeling to foster collaboration and share modeling practices among state agencies in response to HR1.
March 2, 2026 - PER Data Workshop: Predictive Analytics for SNAP PER Reduction
March 30, 2026 - PER Data Workshop: Feature Engineering for Prioritizing Case Reviews
SNAP QC Error Viewer
This application visualizes Quality Control (QC) errors in Supplemental Nutrition Assistance Program (SNAP) cases, which are sampled and reviewed by state and federal auditors. Note that the underlying data from SNAPQCdata.net excludes cases where the household was found to be ineligible, which means that many errors are not represented here.
Use the Error Pathways tab to examine how errors occurred and what caused them, Error Demographics to see which groups were most affected, and Error Severity to assess the financial impact of different error categories. The Pivot Table and Base Rates tabs provides a customizable table that summarizes SNAP QC errors by selected categories.
Click here to use the SNAP QC Error viewer (will open new page).
SNAP Error Prediction with Regression Trees
Open source code implementing regression trees to find rules at the state and national level that can be optimized using internal state data. The national data, which is probably most helpful for finding general rules, focuses on over- and under-issuance income errors. Modeling "buffers" (i.e., how much a case can change before it has an error), can help to increase precision by excluding cases and examples of that can be found here (see end of notes from March 30 workshop for details).
SNAP PER Cost Share Projection: Impact of Removing the QC Tolerance Threshold QC Error Viewer
Congress is considering removing the SNAP QC tolerance threshold, which excludes small errors (currently $58 or less) from a state's payment error rate (PER). Beginning in FY28, states with error rates exceeding 6%, 8%, and 10% will need to cover 5%, 10%, and 15% of benefits issued, respectively. This tool uses the most recent available data (FY23 and FY24) to project the potential impact of removing the tolerance threshold in terms of those cost shares.
Click here to use the SNAP PER Cost Share Projections viewer (will open new page).