As soon as you begin sending your data, Anomaly will start learning from it. After this learning process is complete, it will be able to detect unusual patterns as they occur. It’s a smart detection system that adapts itself based on your metrics.
GET STARTED
To start the learning process, simply send your metrics to Anomaly. Our algorithms will start learning from your data and never stop, becoming ever more accurate over time. At first the system will notice simple patterns, such as minimums and maximums. Over time, it will start detecting much more complex patterns, such as trends, correlation, and much more.
This learning process uses cutting edge techniques such as machine learning, supervised learning, neural networks, and pattern recognition, as well as our own mathematical models.
We try every possibility to learn the most we can about your metrics. We transform each metric into many others through smoothing, aggregating, and other techniques, and try to learn from them. We even compare transformed metrics against each other, creating trillions or even quadrillions of possibilities.
Using all the patterns derived from your metrics, we can detect unusual behaviors. We watch every data point you send us in real time, and if anything unusual is triggered, we send an alert. The learning mechanism allows the detection process to become more or less flexible to anomalies, depending on how many times our forecasts are validated. When the predictions are often correct, the detection is very strict in triggering alerts. However, if the prediction is correct less often, more flexibility for deviation is allowed before an alert is generated.
Of course, detection is useless if you get don’t receive the information, or you get it too late. That’s why we notify you by email, text message, or using other alerting services such as PagerDuty. Alerts are a difficult subject; they need to be easy to understand in order to take quick action. We work hard to deliver understandable alerts in plain English.
If your system is undergoing maintenance, this will be detected as an anomaly. But is it really an anomaly? Technically yes, but from your point of view, it isn’t. Anomaly needs to be told when an anomaly is detected correctly or not. It will learn from your feedback and take it into account in future detections. Similarly, the system might find patterns that aren’t relevant to you, such as coincidences, incorrect trends or correlations that don’t make sense. When you get a false alert simply tell us what to do in the future. The more you tell Anomaly about what you need, the more accurate it will become.
© 2019 Anomaly.
sending...