Rollbar Inc. today announced an upgrade to its error monitoring platform for developers called Adaptive Alerts that adds real-time error alerts to warn about unexpected issues in monitored applications using anomaly detection to reduce false positives and false negatives.

Ideally, developers and operations teams want to detect errors before they are deployed, but that’s not always possible. Therefore, error monitoring platforms are necessary for staying ahead of them. However, many operate on a threshold reporting level or metrics, which can lead to a lot of noise and extra work for developers.

“Adaptive Alerts is the next generation of trendline alerting in Rollbar,” said Brian Rue, co-founder and chief executive at Rollbar. “Compared to the previous generation, which Rollbar customers know as the 10^nth Occurrence and High Occurrence Rate, Adaptive Alerts sends 86% fewer notifications, thanks to automatically adjusting thresholds, and a broader exception-level view that effectively detects application-level trends.”

Enterprise applications exist in complex environments and can manifest unique errors, but many of them are not a priority to fix as long as they are not critical and do not occur at a significant rate. Developers do want to know if the rate of their occurrence has changed over a trend line so they can be prioritized.

Rollbar’s algorithm uses a continuous monitoring solution that learns from the past two weeks of historical data to create a trend for each application. Then the Adaptive Alerts uses that baseline to determine what is anomalous and sends alerts within minutes. Users can adjust minimum exception rates to alert and sensitivity to fine-tune anomaly detection in order to help cut through whatever noise remains.

By adapting its own thresholds to the actual activity of the product, Adaptive Alerts stays on top of a moving baseline and catches actual anomalies rather than a locked threshold. For example, if an application’s audience swells from thousands to hundreds of thousands, a minor but non-catastrophic database timeout — ordinarily handled in code — could rise from tens of instances to hundreds.

In a threshold-only alerting world this would produce constant alerts, which would take developer time to examine and then adjust the threshold. But with an adaptive baseline, it would be seen as a part of the new scale and no new alerts would be produced. However, if the same timeout suddenly spiked above that trend, that would be a clear anomaly and could be detected within minutes.

The new Adaptive Alerts feature will be automatically available to customers of Rollbar’s Advanced and Enterprise plans.

Image: Rollbar