Simpplr uses a machine learning algorithm that adjusts content recommendations in Feed, Tile, and Popular Content Emails based on user behavior. This means that each user will see different content personalized to them.
- All content types are supported (page, album, blog, and event).
- There are up to five content recommendations for the daily email, 10 content recommendations for the weekly email, and 15 content recommendations in feed and tile.
What are recommendations based on?
Our model bases recommendations on:
- The content users have engaged with in the past.
- The content viewed by the users with similar profiles.
- Recency of content. (latest news are more relevant than old news)
- Do not recommend the content users have seen before since they are less likely to engage with them again.
- Sites and people users are following.
- If the user has engaged with less than 5 content (cold start), the model will provide 10 most popular and recent content to collect more signals.
- Must read content and content on home and site carousel will be prioritized higher. (customer curated content)
- Content older than a month is not recommended.
- The model doesn't recommend more than two pieces of content from the same category (diversity).
- The same content will not be recommended to the same users more than twice in daily/weekly email to avoid spamming (feedback loop).
- If there is no content to recommend to users, we won't send an email and we will hide "Recommended" tab in feed and tile.