Analytics Case Study
Income Change Detection Algorithm
After investigating several available algorithms, Sparkfish created a new algorithm that improved upon the existing options.
Means-tested programs provide benefits to individuals, and families, whose income and financial resources fall below certain eligibility requirements. Because the ongoing verification of changes in income is difficult, our client wanted to analyze streams of income data, based on historical data, to determine whether material changes in income had occurred that might impact eligibility.
Our team investigated several available algorithms, such as the Bollinger Bands, Cumulative Sum, and Exponentially Weighted Moving Average, to analyze streams of payments to identify changes in an individual’s income based on current and historical data. The results were compared against whether a human would tag them as “Change” or “No Change” when presented with the streams of income data.
Not satisfied with the accuracy of the results, Sparkfish created a new algorithm that improved over the existing algorithm options. The new algorithm compared daily average income across the entire reference period, and identified any increases above the threshold limit. The algorithm worked well across multiple datasets and, over a two-month period, identified more than 5,400 cases in which the income of current beneficiaries had grown to exceed the eligibility requirement limit, resulting in a savings of $1.9 million to our client.
- Developing a new, custom algorithm that improved over existing ones
- Validating data
- Visualizing and interpreting data
- Complying with a growing array of State and Federal regulations
- Adhering to stringent privacy and security requirements
- Providing complete audit functionality
- Immediate, cost-effective solution
- Speedier eligibility determination
- Increased productivity and accuracy