Alert Group Types in SAS Field Quality Analytics
There are a few workflows within SAS Field Quality Analytics. You can start with either an alert group, or a data selection. Alert groups can be a good place to start if you want to see which problems exist in your data, and a data selection can be a good place to start if you have a specific product or event that you want to analyze. After you have an alert group or data selection you can dive into it deeper by performing specific analyses. Additionally at any time you can create a launch for the data selection and use it in other SAS Applications.
Select any image to see a larger version.
Mobile users: To view the images, select the "Full" version at the bottom of the page.
The purpose of this post is talk more about the different types of alert groups within SAS Field Quality Analytics. Future posts will dive into data selections, the different types of analyses, and launches.
Alert groups, provide early warnings for field and quality issues as early as possible. Alert groups can be executed after each data refresh. These processes generate alerts and related graphics that you can view in the Alert workspace. There are three types of alert groups: emerging issues, on-demand, and threshold. The emerging issues, and on-demand alert groups use an analytic process to detect statistically significant shifts in claims or other activity to flag these events for investigation. The threshold alert group compares actual event activity to user-specified targets, and flags instances where targets are exceeded.
Analytic Methods
Emerging issues and on-demand alert groups apply two different statistical methodologies to detect upward shifts in event activity: production period and event period. The production period method monitors event activity for production periods at different time-in-service periods. The event period monitors event activity across calendar periods.
The production period method takes three-time dimensions into account. Two of the dimensions appear on the graph produced: Build period (production date), and periods in service. Red cells in the matrix indicate that event activity is significantly higher than expected for units in the given build and period in service.
The event period method monitors event activity across calendar periods. When the method identifies higher than expected event activity for a given group of units and events, it produces a graphic like the example below. The graphic displays three lines: actual count in blue, critical value in red, and expected count in green. When the actual count exceeds the critical value, the conclusion is that event activity is significantly higher than expected for that calendar period.
Emerging Issues
Emerging issues alert groups can have one or more product reporting variables and one or more event reporting variables. They must include at least one product reporting variable and one event reporting variable. For emerging issues alert groups, you can select whether to run the production period method only, the event period method only, or run both methods. The emerging issues alert group then creates groups of all the combinations of the reporting variables selected and runs each group with the analytic method selected. Emerging issues alert groups are typically configured by administrators and used to identify most issues.
In the event period graph shown above an alert was generated when the model code was Gemini, the county code was 840 (US), and the primary labor code is I-003.
On-Demand
On-demand alert groups can have either one product reporting variable or one event reporting variable. The one reporting variable is then run for each analytic method. This alert group is good for targeted, limited scope, and one-time investigations.
In the production period graph shown above an alert was generated when the primary code was C-007. The on-demand alert group only includes one reporting variable at a time.
Threshold
Threshold alert groups calculate actual event activity for each value of a specified reporting variable within the data selection. The results are compared to a user-specified target, and instances where the target is exceeded are flagged. Threshold alert groups can be useful for safety and regulatory issues when a known benchmark can be used as a target.
Once an alert group is set up alerts will begin appearing. Future posts will delve more into working with alerts, data selections, and specific analyses. Additionally, recently the SAS® Field Quality Analytics on SAS® Viya® course has been added to the SAS Internet of Things Learning Subscription.
Find more articles from SAS Global Enablement and Learning here.
