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.