We Support your Mission with our advanced analytics
We support your mission and improve productivity through modeling, forecasting, and simulations. We enable your policy decisions with our proven digital modernization strategies and data-driven decision support tools. Here is our Capability Statement.
ANOMALY DETECTION: CASE STUDIES
SafeRock has broad experience in anomaly detection across multiple sectors. We can process millions of records and manage billions of data points to create automated management systems that help our clients identify outliers, anomalies, and areas of risk and potential threat.
- A vendor fund management department wanted to eliminate errors, duplicate entries, waste, and potential fraud. Management also wanted to identify emerging partners that over-performed their targets, in order to develop them as Key Strategic Partners.
- We created a unified system for 3 years of live transactions, representing 85,000+ items and vendor financial commitments for $300+ million. Our system identified anomalies as soon as data was received for 12,000 items daily, and separately grouped outlier values into items for a) supervised correction and approval, b) discarding, and c) additional research.
- Our system successfully detected and sorted anomalies and outliers. This increased staff productivity, eliminated duplicate record keeping, and reduced staff burden.
- Increased annual vendor fund capture by 12% and staff time savings valued at $2.4 million.
- DevOps, MLOps, AI, RPA, Vendor Fund SMEs, SQL, batch processing, SafeRock’s TOROS solution, relational database engine.
MODELING AND FORECASTING: CASE STUDIES
SafeRock’s robust mathematical modeling creates accurate, best-in-class forecasts for our clients. We have experience in predictive analytics across several sectors ranging from retail to E&P. Our successful track record is codified in sophisticated, reusable algorithms, models, and cloud-based solutions.
- A client requested an econometric analysis of COVID-19 correlates, as a way to improve their national business re-opening strategy as the work environment changes and recovers from the Pandemic.
- We analyzed variables associated with COVID-19 spread, including social vulnerability, income, wealth, education, and health, and integrated them into a robust predictive model. We used data for 3,100 counties covering 90%+ of the United States population, to achieve highly accurate, county-level forecasts for all counties in the United States.
Our models were accurate in forecasting COVID-19 incidence with 98% accuracy. The predictive factors were highly significant at the 99% level.
- DevOps, MLOps, supervised analysis and reporting, Statistics SMEs, SAS, Agile team, AWS, Python, R and Shiny.