Merge pull request #198077 from mrbullwinkle/mrb_05_12_2022_ma

[Applied AI] [Metrics Advisor] auto-tuning

Author: ShannonLeavitt

articles/applied-ai-services/metrics-advisor/how-tos/configure-metrics.md

@@ -8,31 +8,94 @@ manager: nitinme
 ms.service: applied-ai-services
 ms.subservice: metrics-advisor
 ms.topic: how-to
-ms.date: 09/10/2020
+ms.date: 05/12/2022
 ms.author: mbullwin
 ---
 
-# How to: Configure metrics and fine tune detection configuration
+# Configure metrics and fine tune detection configuration
 
-Use this article to start configuring your Metrics Advisor instance using the web portal. To browse the metrics for a specific data feed, go to the **Data feeds** page and select one of the feeds. This will display a list of metrics associated with it.
+Use this article to start configuring your Metrics Advisor instance using the web portal and fine-tune the anomaly detection results.
+
+## Metrics
+
+To browse the metrics for a specific data feed, go to the **Data feeds** page and select one of the feeds. This will display a list of metrics associated with it.
 
 :::image type="content" source="../media/metrics/select-metric.png" alt-text="Select a metric" lightbox="../media/metrics/select-metric.png":::
 
-Select one of the metric names to see its details. In this detailed view, you can switch to another metric in the same data feed using the drop down list in the top right corner of the screen.
+Select one of the metric names to see its details. In this view, you can switch to another metric in the same data feed using the drop-down list in the top right corner of the screen.
 
-When you first view a metric's details, you can load a time series by letting Metrics Advisor choose one for you, or by specifying values to be included for each dimension. 
+When you first view a metric's details, you can load a time series by letting Metrics Advisor choose one for you, or by specifying values to be included for each dimension.
 
 You can also select time ranges, and change the layout of the page.
 
 > [!NOTE]
 > - The start time is inclusive.
-> - The end time is exclusive. 
+> - The end time is exclusive.
 
-You can click the **Incidents** tab to view anomalies, and find a link to the [Incident hub](diagnose-an-incident.md).
+You can select the **Incidents** tab to view anomalies, and find a link to the [Incident hub](diagnose-an-incident.md).
 
 ## Tune the detection configuration
 
-A metric can apply one or more detection configurations. There is a default configuration for each metric, which you can edit or add to, according to your monitoring needs.
+A metric can apply one or more detection configurations. There's a default configuration for each metric, which you can edit or add to, according to your monitoring needs.
+
+### Detection configuration auto-tuning based on anomaly preference
+
+Detection configuration auto-tuning is a new feature released in Metrics Advisor, to help address the following scenarios:
+
+- Depending on the use case, certain types of anomalies may be of more significant interest. Sometimes you may be interested in sudden spikes or dips, but other cases, spikes/dips or transient anomalies aren't critical. Previously it was hard to distinguish the configuration for different anomaly types. The new auto-tuning feature makes this distinction between anomaly types possible. As of now, there are five supported anomaly patterns:
+    * Spike
+    * Dip
+    * Increase
+    * Decrease
+    * Steady
+
+- Sometimes, there may be many dimensions within one metric, which will split the metric into hundreds, thousands, or even more time series to be monitored. However, often some of these dimensions aren't equally important. Take revenue as an example, the number for small regions or a niche product category might be quite small and therefore also not that stable. But at the same time not necessarily critical. The new auto-tuning feature has made it possible to fine tune configuration based on series value range.
+
+This allows you to not have to spend as much effort fine tuning your configuration again and again, and also reduces alert noise.
+
+> [!NOTE]
+> The auto-tuning feature is only applied on the 'Smart detection' method.
+
+#### Prerequisite for triggering auto-tuning
+
+After the metrics are onboarded to Metrics Advisor, the system will try to perform statistics on the metrics to categorize **anomaly pattern** types and **series value** distribution. By providing this functionality, you can further fine tune the configuration based on their specific preferences. At the beginning, it will show a status of **Initializing**.
+
+:::image type="content" source="../media/metrics/autotuning-initializing.png" alt-text="Screenshot of Metrics Advisor U I with Initializing auto-tuning text visible." lightbox="../media/metrics/autotuning-initializing.png":::
+
+#### Choose to enable auto-tuning on anomaly pattern and series value
+
+The feature enables you to tune detection configuration from two perspectives **anomaly pattern** and **series value**. Based on your specific use case, you can choose which one to enabled or enable both.
+
+- For the **anomaly pattern** option, the system will list out different anomaly patterns that were observed with the metric. You can choose which ones you're interested in and select them, the unselected patterns will have their sensitivity **reduced** by default.
+
+- For the **series value** option, your selection will depend on your specific use case. You'll have to decide if you want to use a higher sensitivity for series with higher values, and decrease sensitivity on low value ones, or vice versa. Then check the checkbox.
+
+:::image type="content" source="../media/metrics/autotuning-preference.png" alt-text="Screenshot with a toggle button for apply pattern preference and apply value preference selected." lightbox="../media/metrics/autotuning-preference.png":::
+
+#### Tune the configuration for selected anomaly patterns
+
+If specific anomaly patterns are chosen, the next step is to fine tune the configuration for each. There's a global **sensitivity** that is applied for all series. For each anomaly pattern, you can tune the **adjustment**, which is based on the global **sensitivity**.
+
+You must tune each anomaly pattern that has been chosen individually.
+
+:::image type="content" source="../media/metrics/autotuning-pattern.png" alt-text="Screenshot of auto-tuning pattern U I within Metrics Advisor." lightbox="../media/metrics/autotuning-pattern.png":::
+
+#### Tune the configuration for each series value group
+
+After the system generates statistics on all time series within the metric, several series value groups are created automatically. As described above, you can fine tune the **adjustment** for each series value group according to your specific business needs.
+
+There will be a default adjustment configured to get the best detection results, but it can be further tuned.
+
+:::image type="content" source="../media/metrics/autotuning-value.png" alt-text="Screenshot of pattern based sensitivity U I with adjustment for anomaly patterns, spike -30 highlighted on a slider with a range from -100 to 100." lightbox="../media/metrics/autotuning-value.png":::
+
+#### Set up alert rules
+
+Even once the detection configuration on capturing valid anomalies is tuned, it's still important to input **alert
+rules** to make sure the final alert rules can meet eventual business needs. There are a number of rules that can be set, like **filter rules** or **snooze continuous alert rules**.
+
+:::image type="content" source="../media/metrics/autotuning-alert.png" alt-text="Screenshot of setup alert rules UI within Metrics Advisor product." lightbox="../media/metrics/autotuning-alert.png":::
+
+After configuring all the settings described in the section above, the system will orchestrate them together and automatically detect anomalies based on your inputted preferences. The goal is to get the best configuration that works for each metric, which can be achieved much easier through use of the new **auto-tuning** capability.
 
 ### Tune the configuration for all series in current metric
 
@@ -44,7 +107,7 @@ There are additional parameters like **Direction**, and **Valid anomaly** that c
 
 ### Tune the configuration for a specific series or group
 
-Click **Advanced configuration** below the metric level configuration options to see the group level configuration.You can add a configuration for an individual series, or group of series by clicking the **+** icon in this window. The parameters are similar to the metric-level configuration parameters, but you may need to specify at least one dimension value for a group-level configuration to identify a group of series. And specify all dimension values for series-level configuration to identify a specific series. 
+Select **Advanced configuration** below the metric level configuration options to see the group level configuration.You can add a configuration for an individual series, or group of series by clicking the **+** icon in this window. The parameters are similar to the metric-level configuration parameters, but you may need to specify at least one dimension value for a group-level configuration to identify a group of series. And specify all dimension values for series-level configuration to identify a specific series. 
 
 This configuration will be applied to the group of series or specific series instead of the metric level configuration. After setting the conditions for this group, save it.
 
@@ -78,7 +141,7 @@ When the sensitivity is turned down, the expected value range will be wider, and
 Change threshold is normally used when metric data generally stays around a certain range. The threshold is set according to **Change percentage**. The **Change threshold** mode is able to detect anomalies in the scenarios:
 
 * Your data is normally stable and smooth. You want to be notified when there are fluctuations.
-* Your data is normally quite unstable and fluctuates a lot. You want to be notified when it becomes too stable or flat.
+* Your data is normally unstable and fluctuates a lot. You want to be notified when it becomes too stable or flat.
 
 Use the following steps to use this mode:
 
@@ -111,8 +174,8 @@ Sometimes, expected events and occurrences (such as holidays) can generate anoma
 > [!Note]
 > Preset event configuration will take holidays into consideration during anomaly detection, and may change your results. It will be applied to the data points ingested after you save the configuration. 
 
-Click the **Configure Preset Event** button next to the metrics drop-down list on each metric details page.
- 
+Select the **Configure Preset Event** button next to the metrics drop-down list on each metric details page.
+
 :::image type="content" source="../media/metrics/preset-event-button.png" alt-text="preset event button":::
 
 In the window that appears, configure the options according to your usage. Make sure **Enable holiday event** is selected to use the configuration. 
@@ -126,17 +189,16 @@ There are several other values you can configure:
 
 |Option  |Description  |
 |---------|---------|
-|**Choose one dimension as country**     | Choose a dimension that contains country information. For example a country code.         |
+|**Choose one dimension as country**     | Choose a dimension that contains country information. For example, a country code.         |
 |**Country code mapping**     | The mapping between a standard [country code](https://wikipedia.org/wiki/ISO_3166-1_alpha-2), and chosen dimension's country data.        |
 |**Holiday options**    | Whether to take into account all holidays, only PTO (Paid Time Off) holidays, or only Non-PTO holidays.         |
 |**Days to expand**    |  The impacted days before and after a holiday.        |
 
+The **Cycle event** section can be used in some scenarios to help reduce unnecessary alerts by using cyclic patterns in the data. For example:
 
-The **Cycle event** section can be used in some scenarios to help reduce unnecessary alerts by using cyclic patterns in the data. For example: 
+- Metrics that have multiple patterns or cycles, such as both a weekly and monthly pattern.
+- Metrics that don't have a clear pattern, but the data is comparable Year over Year (YoY), Month over Month (MoM), Week Over Week (WoW), or Day Over Day (DoD).
 
-- Metrics that have multiple patterns or cycles, such as both a weekly and monthly pattern. 
-- Metrics that do not have a clear pattern, but the data is comparable Year over Year (YoY), Month over Month (MoM), Week Over Week (WoW), or Day Over Day (DoD).
- 
 Not all options are selectable for every granularity. The available options per granularity are below (✔ for available, X for unavailable):
 
 | Granularity | YoY | MoM | WoW | DoD |
@@ -150,7 +212,6 @@ Not all options are selectable for every granularity. The available options per
 | Secondly | X | X | X | X |
 | Custom* | ✔ | ✔ | ✔ | ✔ |
 
-  
 When using a custom granularity in seconds, only available if the metric is longer than one hour and less than one day.
 
 Cycle event is used to reduce anomalies if they follow a cyclic pattern, but it will report an anomaly if multiple data points don't follow the pattern. **Strict mode** is used to enable anomaly reporting if even one data point doesn't follow the pattern. 
@@ -159,14 +220,13 @@ Cycle event is used to reduce anomalies if they follow a cyclic pattern, but it
 
 ## View recent incidents
 
-Metrics Advisor detects anomalies on all your time series data as they're ingested. However, not all anomalies need to be escalated, because they might not have a big impact. Aggregation will be performed on anomalies to group related ones into incidents. You can view these incidents from the **Incident** tab in metrics details page. 
+Metrics Advisor detects anomalies on all your time series data as they're ingested. However, not all anomalies need to be escalated, because they might not have a significant impact. Aggregation will be performed on anomalies to group related ones into incidents. You can view these incidents from the **Incident** tab in metrics details page. 
 
-Click on an incident to go to the **Incidents analysis** page where you can see more details about it. Click on **Manage incidents in new Incident hub**, to find the [Incident hub](diagnose-an-incident.md) page where you can find all incidents under the specific metric. 
+Select an incident to go to the **Incidents analysis** page where you can see more details about it. Select **Manage incidents in new Incident hub**, to find the [Incident hub](diagnose-an-incident.md) page where you can find all incidents under the specific metric.
 
 ## Subscribe anomalies for notification
 
-If you'd like to get notified whenever an anomaly is detected, you can subscribe to alerts for the metric, using a hook. See [Configure alerts and get notifications using a hook](alerts.md) for more information.
-
+If you'd like to get notified whenever an anomaly is detected, you can subscribe to alerts for the metric, using a hook. For more information, see [configure alerts and get notifications using a hook](alerts.md) for more information.
 
 ## Next steps 
 - [Configure alerts and get notifications using a hook](alerts.md)

articles/applied-ai-services/metrics-advisor/index.yml

@@ -78,3 +78,18 @@ landingContent:
     links:
       - text: Support and help options
         url: ../../cognitive-services/cognitive-services-support-options.md?context=/azure/cognitive-services/metrics-advisor/context/context
+
+- title: Blog posts
+  linkLists:
+    - linkListType: reference
+      links:
+        - text: Introducing Metrics Advisor - A new Cognitive Service
+          url: https://techcommunity.microsoft.com/t5/azure-ai/introducing-metrics-advisor-a-new-cognitive-service/ba-p/1668025
+        - text: Protect your organization's growth by using Azure Metrics Advisor
+          url: https://techcommunity.microsoft.com/t5/azure-ai/protect-your-organization-s-growth-by-using-azure-metrics/ba-p/2564682
+        - text: Supply Chain Anomaly Detection and Root Cause Analysis with Azure Metric Advisor
+          url: https://techcommunity.microsoft.com/t5/azure-ai/supply-chain-anomaly-detection-and-root-cause-analysis-with/ba-p/2871920
+        - text: 'Customer support: How Azure Metrics Advisor can help improve customer satisfaction'
+          url: https://techcommunity.microsoft.com/t5/azure-ai-blog/customer-support-how-azure-metrics-advisor-can-help-improve/ba-p/3038907
+        - text: Detecting Methane Leaks using Azure Metrics Advisor
+          url: https://techcommunity.microsoft.com/t5/ai-cognitive-services-blog/detecting-methane-leaks-using-azure-metrics-advisor/ba-p/3254005