GCP Timeseries API Explorations: Finding Signals Amongst The Noise

One of the powerful aspects of GCP's new Timeseries Insights API is its ability to discern signals amongst the noise. Asked to find major anomalies from 6 to 7PM tonight, the API identified "National Anthem" as strongly anomalous. Looking closely at its timeline over the past 72 hours, it is extremely bursty, with sharp increases and decreases alongside macro-level trends. Yet, simply eyeballing the timeline below, it is clear that the 6-7 PM period at the end stands alone over the past three days as a unique vertical surge without precedent over the comparison period.

The API offers a number of parameters that can be tuned to adjust precisely what constitutes an "anomaly" for a given application and dataset, but as the timeline below shows, even with a generic set of parameters, the API is able to look over an extremely bursty timeseries and identify when it truly diverges from its baseline.