Deloitte Consulting LLP The Power of "Nudge": Using Predictive Analytics and Behavioral Science to Drive Results for Government Deloitte Consulting LLP May 12, 2016
Presenters Rachel Frey Principal Deloitte Consulting LLP Rachel Frey is Principal in Deloitte’s state government health and human services practice. For the past fifteen years she has worked with state and local government agencies defining approaches and implement business and technology solutions to enhance service delivery, prevent and detect fraud, waste and abuse, streamline business operations, refine organizational structures and evaluate policies that drive these efforts. She has a Masters in Public Policy and Management from the Heinz School at Carnegie Mellon. Rachel currently leads Deloitte’s Human Services Analytics solution offerings. Lynnette Stern Specialist Leader Technology Systems Integration Deloitte Consulting LLP Lynnette Stern is Specialist Leader in Deloitte’s state government health and human services practice. She has 31 years of experience in public sector program and IT services, serving clients in the Workforce and Employment Market. She focuses on implementing innovative technology solutions designed to improve program performance and outcomes. These solutions include applying predictive analytics and behavioral economics as part of a strategy to reduce costs, improve operations, and increase customer satisfaction.
Nudging to Reduce Improper Payments – New Mexico Case Study: New Mexico Department of Workforce Solutions Overview Predictive analytics and behavioral “nudges” can reduce improper payments dramatically without impacting eligible claimants or putting additional burden on staff. Key Features Predictive models assess improper payment risk in real-time Behavioral nudges crafted to educated and inform Randomized Control Trials (RCT) employed to accurate measure benefits using a control group. Flexibility to “Test, Learn, Adapt” Benefits Up to 1.9x improvement of voluntary reporting of earnings 40% reduction in fraud Detecting more improper payments by working optimizing the queue to work “smarter not harder.” 3
Sources of Improper Payments Overview “If it’s in the news, don’t worry about it. The very definition of ‘news’ is ‘something that hardly ever happens.’ It’s when something isn’t in the news, when it’s so common that it’s no longer news … that you should start worrying.” – Bruce Schneier 95% - smaller issues 5% - serious schemes Work Search: claimant not performing required search for work Benefit Year Earnings: claimant doesn’t properly report earned income while on benefits Joy Mike to add color commentary Separation: claimant was not eligible due to reason for separation (e.g., terminated for cause) Able & Available Base Period Wages Other (including identity theft) Source: USDOL http://www.dol.gov/dol/maps/map-ipia.htm
The Challenge in UI Reduce overpayments without impacting eligible claimants or increasing time for staff. Enhance Program Integrity Enhance Program Integrity Joy Do not take benefits from eligible claimants Do not impact staff time
Combining Predictive Analytics and “Nudges” “When the ultimate goal is behavior change, predictive analytics and the science of behavioral nudges can serve as two parts of a greater, more effective whole.” A predictive model suggests who to target, but not how to effect the desired behavior change Predictive Analytics Behavioral Science “Push the worst, Nudge the rest” Customized behavioral interventions based on risk Mike Motivates one-size-fits-all intervention strategies http://dupress.com/articles/behavioral-economics-predictive-analytics/
Actions to Reduce Overpayments Once identified, actions are needed to reduce overpayments. A range of potential actions are possible. Prioritize Workload Outbound Call Open Investigation Stop Payment Broadcast Messaging Correspondence Joy Less Invasive More Invasive
Nudge through Pop-ups Pop-ups appear onscreen during the weekly certification in situations where overpayments are more likely to happen. Joy
uDetect™ increases reported earnings each week Reporting Earnings Messages delivered in real time are working, encouraging claimants to report more accurate earnings in real time: Top performing messages cause 1.9x more individuals to report earnings. Messages are 2.1x more effective when delivered by staff to claimants in person. Benefit Year Earnings Fraud: Once cross matches come in those seeing a pop-up 40% less likely to have committed fraud. Earnings Fraud Duration Average of all messages
Work Smarter Not Harder: Prioritization of BPC Workload Prioritization benefits: Overpayments are identified 8 weeks faster on average We forecast that 28% more overpayments will be identified in total: ~$750K per year. No additional workload on staff: the same number of issues are being worked on average
Application to other Programs Across HHS
Understanding our Population Population segmentation is a business strategy that involves dividing a broad target universe into distinct subsets based on a select set of population characteristics through domain knowledge and statistical methods Population Segmentation Analysis Outputs Individual/Household/ Organizational Insights Tailored Programs and Services Targeting Strategies Customer segmentation translates raw population data into meaningful service strategies, reducing the cost to serve certain segments while increasing efficiency and effectiveness of overall program
Which Part of our Population can we Influence/Nudge? Providers Employers Staff Clients Behavioral, Demographic, Needs, Lifestyle Analytics Public
Leveraging Analytics to Enhance Outreach and Engagement… Targeted outreach for client enrollment (e.g., WIC, Medicaid, Paid Family Leave, SNAP, Adult Protective Services) Targeted outreach for provider enrollment (e.g., child care providers, HCBS providers, foster care parents) Targeted outreach to support individual goals (e.g., employment opportunities, training opportunities, nearby farmer’s markets, preventive care reminders) Targeted outreach to request enhanced reporting of potential provider/recipient violations Tailored messaging to customer service workers when interacting with clients
Leveraging Analytics to Reduce Risk and Increase Compliance… Tailored notification language, emails and text messages to increase timeliness of change reporting, renewals and recertifications Guided navigation/behavioral messaging for clients, workers and providers to increase accuracy Tailored notifications to increase child support order compliance Targeted auditing and quality assurance sampling Generation of risk scores associated with risk of program errors, potential eligibility to receive or provide services, compliance with support orders and potential abuse and/or negligence
Leveraging Analytics to Tailor Services and Programs to meet Goals… Guided Individualized work responsibility plan creation and monitoring based on personal needs and employment goals for similar circumstances Tailored interventions that focus on timely and stable exits from foster care for children Tailored referrals (services and duration) to supportive services based on others’ successes Adjustment of programs and services based on program/service utilization and effectiveness
Presenter contact information Rachel Frey Principal, Deloitte Consulting LLP rfrey@deloitte.com Lynnette Stern Specialist Leader, Deloitte Consulting LLP lystern@deloitte.com
Q&A