Guardian in Salesforce Marketing Cloud Personalization (formerly Interaction Studio) is an anomaly-detection feature that monitors key metrics in your Personalization environment (e.g., impressions, clicks, add-to-cart events, revenue). Guardian compares real-time data against expected ranges to alert you if a potential anomaly is detected.

Below is how it determines the expected range:

Historical Baseline and Machine Learning

Guardian leverages historical data for each metric and applies machine learning algorithms to learn typical patterns. This includes seasonality, general traffic trends, and cyclical behaviors.

As data is collected over time, Guardian refines the upper and lower thresholds for each monitored metric based on these learned patterns.

Automated Threshold Adjustments

Because Guardian is continuously learning, it adapts to new patterns in user behavior over time. If your site or campaign sees increased traffic due to a seasonal event or marketing push, Guardian will eventually absorb these changes into its baseline, allowing for more accurate anomaly detection.

Real-Time Monitoring

Guardian then uses these learned thresholds in real time. When a metric falls outside its expected bounds (too high or too low), Guardian flags this as a potential anomaly and can notify administrators or other stakeholders.

Salesforce Documentation References

Salesforce Help: Monitor Metrics with Guardian

Describes how Guardian uses machine learning to establish metric thresholds and detect anomalies.

Salesforce Help: Analyzing Key Metrics

Explains various ways to analyze metrics in Personalization, including how Guardian can highlight anomalies.

Why the Other Options Are Not Correct

B. Guardian comes with pre-built ranges for each metric, which cannot be configured

Incorrect. Guardian does not rely on unchanging static thresholds; it dynamically learns from your data.

C. Guardian uses upper and lower bounds set by the user for each metric

Partially correct in a custom scenario where manual thresholds can be set, but by default, Guardian’s key benefit is its automated, machine-learning-driven approach.

D. Guardian queries the Data Warehouse to establish logical expected ranges

While Guardian does rely on your platform’s data, it’s not just a raw query. It uses machine learning models to understand patterns and anomalies rather than simply performing manual logic-based queries.