New Relic has improved its Dynamic Baseline Alerts system by implementing auto-discovered seasonality and an ensemble algorithm that chooses the best fit for each time series data stream every minute. This allows for more accurate predictions and better performance, while also making it easier for customers to use the service without having to manually select algorithms or seasonality. The new system uses a technique called Fast Fourier Transforms (FFTs) to identify underlying frequency in time series data, and then evaluates candidates against historical metric data to determine the best fit. Additionally, the ensemble algorithm selects the algorithm with the best performance, using a weighted approach that considers recent performance more heavily than older data. This allows New Relic to provide more accurate and effective alerts for its customers, without requiring them to do manual configuration or tuning.