# Recommended fallback strategy — Hot/Trending recall Hot/Trending recall is a recommendation strategy based on global user behavior statistics, aimed at recommending the most popular items within a specific time period to users. ### Applicable scenarios This strategy primarily addresses the following two issues: 1. **Recommended fallback**: When other personalized recommendation strategies (such as "You may like") fail to recall enough items, Hot/Trending recall supplements with globally popular items to ensure the recommendation is not empty and improve user experience. 2. **New user cold start**: For new users or those with sparse behavior data, the system cannot perform effective personalized recommendations. At this time, Hot/Trending recall can recommend generally popular items as a starting point for discovery and exploration. --- This article will guide you on how to configure the Hot/Trending recall strategy. ### Configuration instructions You can configure the calculation logic for the "most popular" items. The platform calculates popular items based on your defined rules and uses the results for fallback recommendations and cold start scenarios for new users. 1. In the console's left navigation bar, select ** Strategy** \> **Fallback Strategy**

2. This will open a configuration window for the new fallback strategy.

3. In the pop\-up window, set parameters according to your business needs, then click the "**OK**" button to complete the configuration. |**Configuration Items** |**Input Restrictions** |**Description** | |---|---|---| |**Time Window** |Minimum: 5 minutes|Time range for counting popular items.|\ | |Maximum: 7 days ||\ | | |* **Explanation**: defines how long the platform "reviews" user behavior to calculate item popularity. For example, if set to "7 days," the system counts data from the past 7 days.|\ | | |* **Optional units**: minutes, hours, days. | |**Behavior type** |Must be within the behavior types defined in the behavior dataset |User behavior used to measure item "popularity."|\ | | ||\ | | |* **Explanation**: defines which behavior can accumulate item popularity.|\ | | |* **Configuration suggestion**: it is recommended to select behaviors that strongly represent user preference, such as `purchase`, `collect`, `share`, and avoid weakly related behaviors like `impression` to ensure the quality of the popularity ranking. |

Viking AI Search