This document discusses Twitter's approach to detecting breakouts or changes in usage patterns from time series data in order to better plan system capacity. The approach uses a change point detection algorithm that first applies a rolling median to smooth the time series data and reduce noise. It then performs hypothesis testing on the smoothed data to identify any statistically significant changes that could indicate hot or cold services. This approach is evaluated on Twitter usage data and other time series data and is shown to outperform existing methods like MACD and Donchian channels in detecting meaningful breaks or shifts while minimizing false positives.