Accessing Recommendation Engine Settings
The Recommendation Engine Settings are configured in the Admin module. These settings define which videos are eligible for recommendations, how content is constrained, and what is promoted.
- Navigate to the Admin module.
- Select Recommendation Engine Settings.
Eligible Content Rules
These rules define the account-level recommendation pool. Only videos matching these criteria will be considered for recommendations. Live channels and events are not eligible.
Look-back Period
Defines the oldest published date eligible for recommendations. Videos published before this timeframe are excluded.
| Option | Description |
|---|---|
| Past Month | Only videos published within the last month |
| Past 3 Months | Only videos published within the last 3 months |
| Past 6 Months | Only videos published within the last 6 months |
| Past Year | Only videos published within the last year |
| Past 2 Years | Only videos published within the last 2 years |
Min Video Duration
Sets the minimum length of recommended videos. Videos shorter than this threshold will not be recommended. The default is No Limit.
Max Video Duration
Sets the maximum length of recommended videos. Videos longer than this threshold will not be recommended. The default is 30 minutes.
Trending Now Timeframe
Controls how virality is calculated for the Trending Now recommendation strategy.
- Daily — Favors fast spikes in viewership. Best for breaking news or event-driven content.
- Weekly — Favors sustained growth. Best for evergreen content strategies.
Expand Recommendations
When enabled, the system will expand recommendations beyond your selected look-back period and duration constraints when no eligible matches are found. Restrictions are still enforced even when this option is active.
Excluded Recommendation Tags
Use excluded recommendation tags to prevent specific content from appearing in recommendations. Videos with any of these tags will be excluded from all recommendation results.
This is useful for:
- Time-sensitive content that should no longer be recommended after a specific date
- Outdated content that is still in the catalog but should not be surfaced
- Sensitive content that should not appear as a recommendation
To add excluded tags, enter the tag values in the Excluded Recommendation Tags field. For example: no_rec, internal.
Recommendation Restrictions
Restrictions enforce editorial lanes by requiring recommended videos to share specific attributes with the initial video. Adding restrictions narrows recommendations to videos that meet your selected criteria.
Restrictions are evaluated against the initial video only — not subsequent recommended videos in a session.
Creating a Restriction
- In the Recommendations Restrictions section, click .
- Fill in the restriction settings:
- Name (required) — A descriptive name for the restriction.
- Description — An optional description of what the restriction does.
- Type — Choose Tag or Custom Field.
- Value — The tag value or custom field value to match.
- Schedule Restriction (optional) — Set a start and end date to apply the restriction only during a specific timeframe.
- Click .
Tag-based Restrictions
When a tag-based restriction is active, the system calculates the intersection of the initial video's tags and the restriction's tags. Recommended videos must contain all tags in that intersection.
Custom Field Restrictions
Custom field restrictions require recommended videos to match a specific custom field value. This is useful when your content taxonomy relies on custom fields rather than tags.
Common Use Cases
- Kids content — Create a restriction on the
kidstag to ensure kids content is never mixed with other categories. If no more eligible kids videos exist, recommendations stop rather than showing non-kids content. - Travel content — Keep
travelrecommendations within the travel lane. - Sports / Politics separation — Prevent sports content from being mixed with politics.
- Campaign windows — Use scheduled restrictions during special coverage periods (e.g.,
elections week,transfer window) to confine recommendations to those topics.
Promoted Content
Promoted content increases the visibility of selected videos within recommendations. Use this feature to highlight new releases, sponsored content, or priority videos.
Adding Promoted Videos
- In the Promoted Videos section, click .
- Select the video to promote and click .
- Configure the promotion settings and click .
Promotion Settings
| Setting | Description |
|---|---|
| Promotion Frequency (%) | The percentage of recommendation requests in which the promoted video will appear. For each request, a random number determines inclusion — the video appears in roughly this percentage of requests. |
| Views Goal | The total number of views after which the promoted video will no longer be included. Once the goal is reached, the video becomes ineligible for promotion. |
| Recommendation Position | The slot in the recommendation list where the promoted video will appear. This position is 0-indexed and does not count the initial video. For example, position 3 means the viewer watches the initial video, then three organic recommendations, then the promoted video. |
| Start Date | The date and time when the promotion begins. |
| End Date | The date and time when the promotion ends. |
Tagging Best Practices
Strong tagging directly improves recommendation quality. The recommendation engine uses your tags alongside AI-based content analysis, so well-structured tags significantly boost precision.
Minimum Viable Tagging
At a minimum, each video should have:
- At least one section/category tag (e.g.,
Sports,News,Business) - At least one entity tag (e.g.,
Real Madrid,Apple,MoMA)
Tips
- Use a controlled vocabulary — Pick canonical tag names and document them. Avoid near-duplicates (use
Soccer, not bothSoccerandFootball). - Use 3–7 high-signal tags — Over-tagging dilutes relevance; under-tagging limits match rates. Prioritize section, topic, entities, format, and geography.
- Align tags with titles — Put primary entities in both the title/description and as tags. This helps both the AI similarity engine and SEO.
- Mark special content with explicit tags — Use tags like
kids,opinion, orsensitiveso restrictions can filter them effectively.
Anti-patterns to Avoid
- Synonyms or variants for the same entity (e.g.,
Ninersvs49ersvsSan Francisco 49ers) - Overly broad single tags with no specifics (e.g.,
Sportsalone with no sport or team) - Low-signal tags that add noise (e.g.,
viral,must-watch) - Mixing taxonomy styles (plural vs singular, local language vs English)
Saving Changes
After configuring your settings, click to apply the changes. Changes take effect immediately for new recommendation requests.