How Does Anomaly Detection Work?

This new beta feature is crafted to ensure you stay informed about key changes in your data—effortlessly.

Anomaly Detection is designed to keep you informed about significant changes in your data metrics through timely notifications. With Anomaly Detection, you can now gain invaluable insights into sudden spikes, incidents, outages, and out-of-the-ordinary trends.

Sentisum uses advanced machine learning to monitor customer interactions and feedback in real-time to spot issues, trends, and opportunities as they happen, so you can be better equipped to address them. Think of this as an early warning system that allows you to keep a close eye on the metrics that matter most, even when you’re away from the platform or dashboard.

How We Detect These Anomalies?

Anomaly Detection is intelligent—it learns from your historical data to establish what’s “normal” for each metric. Here’s how it defines and detects anomalies:

• Learning Phase: The system analyses past data to understand typical patterns and fluctuations in each metric.
• Normal Range Establishment: Based on historical insights, it sets a baseline or 'normal' range for each metric.
• Anomaly Detection: When new data deviates from this established normal range—either increasing or decreasing significantly—the system identifies this as an anomaly.

How It Works?

Anomaly Detection operates on two primary frequencies:

• Daily Anomaly Alerts:
  • Analyses historical data over the past 100 days to establish a “normal” baseline.
  • Anomaly Detection collects and analyses the entire previous day's data, detects anomalies and sends notifications right to your inbox.
• Hourly Anomaly Alerts:
  • Continuously monitors ticket volumes and data changes every hour.
  • Sends out alerts promptly after processing and detecting any anomalies.

How to Use Anomaly Detection?

1. Setting Up Alerts:
   • Go to the Anomaly Detection section in your dashboard.

1. • Go to the Alert configuration section in your dashboard.

1. • Click “Setup New Alert” to start configuring.
     
     
2. Configuring Alerts:
   • Name your alert for easy identification.
   • Choose your preferred frequency: Daily or Hourly.
   • Select the metric you want to monitor, such as volume or percentage of ticket.
   • Choose relevant dimensions to specify the criteria (e.g., tags like Sentisum or DIY).
     
3. Adding Email Notifications:
   • Enter the email addresses where alerts should be sent. You can add multiple recipients if necessary.
4. Editing or Deleting Alerts:
   • After creating alerts, you can modify their settings if your monitoring needs change.
     

1. • You can also turn alerts on or off or delete alerts from the management page if they are no longer needed.

2. Receiving Notifications:

   • For Daily Alerts, you will receive a single email in the morning summarising anomalies from the previous day.
   • For Hourly Alerts, expect multiple email notifications within a few hours if anomalies are detected.

Providing Feedback:

Direct Feedback: Each anomaly detected on your dashboard comes with a "thumbs up" and "thumbs down" option. By selecting one of these, you can quickly tell us how relevant and useful each anomaly is to you. The system learns from the feedback actively.

Troubleshooting Common Scenarios

Not Receiving Expected Alerts?

• Ensure you have opted in for notifications on the Alert Configuration Page.
• Check your email settings to make sure alerts aren’t going to your spam folder

Receiving duplicate alerts:

• Check if you are a part of multiple email alerts with the same few topics.

Receiving too many alerts?

• Reach out to the team to increase the sensitivity of your alert detection.
• Toggle the email alert off

FAQs

Q: What frequency of alerts should I choose?

A: Select daily alerts if you prefer a summary at the end of the day. Choose hourly alerts for immediate updates on critical metrics.

Q: Can I create multiple alerts for the same metric?

A: Yes, you can set up multiple alerts for the same metric.

Q: What happens if I try to set a duplicate alert?

A: If you attempt to create an identical alert, the system will notify you that a similar alert already exists.

Q: How is the normal baseline used in alerts?

A: The normal or the baseline is the Z score that the Ai identifies. The Z-score is a statistical measure that indicates how far a data point deviates from the average. It is vital for accurately identifying anomalies in your data.

Q: Can I edit the conditions of my alerts after they are set up?

A: Absolutely! You can modify alerts at any time to adjust their criteria, metrics, or email notifications.

Q: How can I provide feedback about my alerts?

A: After receiving notifications, you can provide feedback using the thumbs up or thumbs down options to indicate whether the alerts were helpful.

If you have any further questions or need assistance with the Anomaly Detection feature, please reach out to our support team. We’re here for you.