Fast-track anomaly detection and analysis
Fast-track anomaly detection and analysis
Spotting unusual behaviors in deep data layers
Tina is an IT manager for an online retail company. One of the key metrics to understand the health of their website is the number of active users; however, using traditional threshold based techniques to identify anomalies led to alert fatigue as they failed to accommodate daily, weekly, and seasonal change.
She turns to Metrics Advisor to provide intelligent anomaly detection.
Tina is an IT manager for an online retail company. One of the key metrics to understand the health of their website is the number of active users; however, using traditional threshold based techniques to identify anomalies led to alert fatigue as they failed to accommodate daily, weekly, and seasonal change.
She turns to Metrics Advisor to provide intelligent anomaly detection.
Bring data from one of many supported sources
Tina pulls her active users time series data from a Azure SQL database, one of many supported data sources. She then identifies the dimensions and metrics within the data before ingesting it.
Ingest historic and real-time data
From the Data Feeds section, she can monitor and configure the ongoing ingestion of historical and new data before drilling into a specific metric for analysis.
Intelligently analyze metrics across dimensions
From the metrics page, Tina can see the aggregate active user activity; but with a few clicks, she can inspect top anomalies or drill down to analyze specific regions or referral channels.
Gain insight into anomalies through multi-dimensional analysis
With Metrics Advisor, Tina can quickly ingest her active users' data, identify anomalies, and then drill into different dimensions to identify contributing factors. Smart detection and rich analytical capabilities make it easy to understand the health of their online business.
Metrics Advisor empowers you with automated multi-dimensional analysis to analyze your time series data.
With Metrics Advisor, Tina can quickly ingest her active users' data, identify anomalies, and then drill into different dimensions to identify contributing factors. Smart detection and rich analytical capabilities make it easy to understand the health of their online business.
Metrics Advisor empowers you with automated multi-dimensional analysis to analyze your time series data.
Bring domain knowledge to rapidly improve accuracy
Bring domain knowledge to rapidly improve accuracy
Improving detection accuracy with domain expertise
Tina is an IT manager for an online retail company. By monitoring active users on the website with Metrics Advisor, she’s able to identify unusual activity; however, she’s spotted some false positives and negatives which can be explained by specific business trends and events.
Tina knows Metrics Advisor provides fine tuning capabilities for its models, so she decides to add some domain knowledge to the solution.
Tina is an IT manager for an online retail company. By monitoring active users on the website with Metrics Advisor, she’s able to identify unusual activity; however, she’s spotted some false positives and negatives which can be explained by specific business trends and events.
Tina knows Metrics Advisor provides fine tuning capabilities for its models, so she decides to add some domain knowledge to the solution.
Adjust detection thresholds at the metric level
Tina knows active users aren’t constant throughout the week and instead follow patterns influenced by work hours and weekends. She moves from hard thresholds to smart detection, an AI-powered detection approach, and tunes it to ensure more minor incidents are correctly flagged.
Custom tune detection thresholds for specific dimensions
She also knows that her company has only recently expanded into the Brazil region. As a result, sales are lower and variance is higher. Fortunately she can reduce the smart detection sensitivity for specific series.
Provide feedback to improve performance
Tina’s efforts have improved detection accuracy considerably. She finishes by tagging a few specific incidents as false positives and by adding local holidays to allow the model to account for increased usage.
Fine tune anomaly detection models for your scenario
With Metrics Advisor, Tina can quickly bring her business expertise to improve the accuracy of anomaly detection. By tuning smart detection at the metric and series levels as well as informing the model of false positives and upcoming events, the solution learns to fit her business.
Metric Advisor combines automated model building with rich customization capabilities to maximize accuracy.
With Metrics Advisor, Tina can quickly bring her business expertise to improve the accuracy of anomaly detection. By tuning smart detection at the metric and series levels as well as informing the model of false positives and upcoming events, the solution learns to fit her business.
Metric Advisor combines automated model building with rich customization capabilities to maximize accuracy.
Gain insight into the driving factors behind each anomaly
Gain insight into the driving factors behind each anomaly
Identifying the root cause of a business anomaly
Tina is an IT manager for an online retail company. By monitoring active users on the website with a customized model, she’s able to accurately identify issues in near real-time; however, diagnosing and solving these issues is still a slow and time consuming process.
She decides to explore Metrics Advisor’s root cause analysis capabilities to get to the bottom of anomalies faster.
Tina is an IT manager for an online retail company. By monitoring active users on the website with a customized model, she’s able to accurately identify issues in near real-time; however, diagnosing and solving these issues is still a slow and time consuming process.
She decides to explore Metrics Advisor’s root cause analysis capabilities to get to the bottom of anomalies faster.
Reduce time to diagnosis with intelligent suggestions
From the incident page, she can see Metrics Advisor correctly identified a drop in active users. It has also identified the key contributors to this change - United States-based channels - which she can easily overlay.
Explore data through visualization tools
Tina can also use the interactive tree to drill down into the contributing regions and channels. The most prominent contributor has been identified in one specific region, and by inspecting further, she finds that each channel shows similar levels of impact. Tina then concludes that this is likely to be an issue with the site in that specific region.
See data in a broader business context
To dig even deeper, she uses the Metrics Graph to draw relationships between other metrics she’s monitoring, like latency and CPU usage, allowing her to correlate this incident back to a technical root cause in the database tier.
Drill into dimensions across metrics to quickly identify root cause
With Metrics Advisor, Tina receives an instant prediction of the root cause of each incident. She also has the tools to drill into contributing factors, both within the current metric as well as across related data sets. With rich analysis capabilities, anomaly detection is no longer a black box.
Metrics Advisor helps explain anomalous behavior within each aspect of your business.
With Metrics Advisor, Tina receives an instant prediction of the root cause of each incident. She also has the tools to drill into contributing factors, both within the current metric as well as across related data sets. With rich analysis capabilities, anomaly detection is no longer a black box.
Metrics Advisor helps explain anomalous behavior within each aspect of your business.