Overview
Anomalies are unexpected spikes in your feedback data that provide valuable insights into your product. These data points represent significant deviations from standard patterns, manifesting as abrupt volume, sentiment, or content shifts.
Use Cases of Anomalies
Identifying product issues: Spot sudden negative feedback spikes to address potential problems.
Monitoring feature impact: Gauge user reception of new features or updates through feedback changes.
Assessing marketing effectiveness: Detect successful campaigns via increased positive feedback or tracked keyword mentions.
Detecting opportunities: Discover areas of excellence by identifying sudden increases in positive feedback or reasons.
Improving customer support: Evaluate support quality and identify improvement areas.
Discover Anomalies
Click on "Quantify" in the top navigation bar.
Click on "Trends + Anomalies" from the visualization options.
View the trend chart to identify any anomalies. Anomalies are represented as spikes or dips in the graph.
To control anomalies visualization, click the "Show Anomalies" toggle to enable or disable the view.
How do we define anomalies?
We use Z-Score to configure Anomalies.
A z-score is like a scorecard that tells us how different a specific data point is from the average, using a common unit known as standard deviations. It helps us determine if a data point is typical or unusual compared to the others.
Imagine user feedback as the speed of a car on a daily commute. The car usually travels at a consistent speed (average), but occasionally it goes faster or slower. A z-score of 2 means the car was two "steps" (standard deviations) faster than its average speed on a particular day, suggesting exceptional driving conditions that should be examined.
Using z-scores helps us identify interesting trends or patterns in user feedback, enabling us to make informed decisions and uncover new opportunities for enhancement.
Z-scores are a highly credible and widely used statistical tool, allowing for consistent and reliable comparisons across various datasets.
You can learn more about Z-Score from this Wikipedia article.
Configure Anomaly Detection
By default, Enterpret sets anomalies at a z-score of 1.5, but users can adjust the sensitivity of the anomaly detection to suit their needs better. Here are the steps to configure anomaly detection sensitivity:
Click on the gear icon next to the "Show Anomalies" toggle on the Trends + Anomalies page.
Select the "Custom" option from the menu.
Use the stepper to adjust the z-score threshold to your preferred level of sensitivity.
Click on the "Apply" button to save your changes.
You can always revert to the same settings by following the same steps.
By customizing the sensitivity of anomaly detection, you can better filter out noise or identify subtle changes in your user feedback data, providing you with more precise insights into user experience and product performance.
Setup Anomaly Reports
Subscribing to anomaly reports enables you to receive regular updates on unusual patterns or deviations in your data, automatically delivered right where you want it.
Create your Quantify Query.
Save your analysis by clicking the "Save" button.
Navigate to the desired dashboard to display the Anomalous Chart.
Click the "Add" button in the top right corner of the dashboard.
Choose "Add Existing Analysis" and select the appropriate analysis to include.
Complete the process by subscribing to the dashboard for regular updates.