Rich Metadata
Add detailed genres, tags, and descriptions to all content. The algorithm relies on metadata for content-based matching.
Recommendations
Vidori’s recommendation engine uses machine learning to deliver personalized content suggestions to each viewer, increasing engagement and watch time.
How Recommendations Work
Section titled “How Recommendations Work”Vidori uses a hybrid recommendation algorithm that combines:
- Collaborative Filtering - Learns from viewing patterns across all users
- Content-Based Filtering - Matches content features (genres, tags, actors)
- Contextual Signals - Considers time of day, device, and viewing history
Recommendation Types
Section titled “Recommendation Types”Personalized For You
Section titled “Personalized For You”Content tailored to each individual viewer based on their watch history, favorites, and engagement patterns.
Where it appears:
- Home screen “Recommended For You” swimlane
- Personalized rows throughout the app
More Like This
Section titled “More Like This”Similar content suggestions shown on video detail pages, based on:
- Shared genres and tags
- Similar content groups
- Viewing patterns of users who watched the same content
Continue Watching
Section titled “Continue Watching”Videos the user has started but not finished, ordered by recency.
Behavior:
- Appears when a video is >5% and <95% complete
- Shows resume position timestamp
- Automatically removes when video is completed
Because You Watched
Section titled “Because You Watched”Recommendations based on a specific recently watched video.
Example: “Because you watched Documentary Title” → shows similar documentaries
Recommendation Swimlanes
Section titled “Recommendation Swimlanes”Add recommendation components to your app layouts in App Studio → Layout Editor:
| Component | Description |
|---|---|
| Recommendations | Personalized content grid based on user profile |
| Continue Watching | Resume in-progress videos |
| Because You Watched | Suggestions based on recent viewing |
| Trending Now | Popular content across all users |
| New Releases | Recently added content |
- Go to App Studio → Layout Editor
- Select the screen to edit (e.g., Home)
- Click Add Component
- Choose a recommendation component type
- Configure display options (title, max items)
- Save and preview
Content Signals
Section titled “Content Signals”The recommendation engine considers multiple signals to rank content:
User Signals
Section titled “User Signals”| Signal | Weight | Description |
|---|---|---|
| Watch history | High | Videos watched to completion |
| Favorites | High | Content explicitly liked |
| Watchlist | Medium | Saved for later viewing |
| Search queries | Medium | Topics of interest |
| Browse behavior | Low | Categories explored |
Content Signals
Section titled “Content Signals”| Signal | Description |
|---|---|
| Genres | Primary and secondary genres |
| Tags | Content tags and keywords |
| Cast & crew | Actors, directors, creators |
| Release date | Newer content gets a boost |
| Popularity | View count and engagement |
Cold Start Handling
Section titled “Cold Start Handling”For new users with no viewing history, Vidori handles recommendations gracefully:
- First Visit - Shows trending and popular content
- After First Watch - Immediately shows “More Like This” suggestions
- After 3-5 Videos - Personalized recommendations begin appearing
- Ongoing - Recommendations continuously improve with more data
Recommendation Settings
Section titled “Recommendation Settings”Configure recommendation behavior in Settings → Viewer Settings:
| Setting | Options | Default |
|---|---|---|
| Personalization | Enabled / Disabled | Enabled |
| Continue Watching threshold | 5-20% completion | 5% |
| Completed threshold | 85-98% completion | 95% |
| History retention | 30-365 days | 90 days |
Analytics
Section titled “Analytics”Track recommendation performance in Analytics → Engagement:
- Recommendation CTR - Click-through rate on recommended content
- Recommendation conversions - Videos watched from recommendations
- Personalization lift - Engagement increase vs. non-personalized
Best Practices
Section titled “Best Practices”Consistent Tagging
Use a consistent tagging taxonomy across your library. Avoid duplicate or similar tags (e.g., “Sci-Fi” vs “Science Fiction”).
Fresh Content
Regularly add new content to give the algorithm fresh material to recommend and keep users engaged.
Monitor Performance
Review recommendation analytics weekly to identify opportunities for metadata improvements.