Reranking is only available for enterprise customers, email info@mixpeek.com for a demo.

Ranking Types

Feedback Signals

  • Click interactions
  • View duration
  • Skip actions
  • Custom events

Popularity Metrics

  • Historical engagement
  • Recent interactions
  • Time decay
  • Confidence scores

Collaborative Filtering

  • User similarities
  • Item-based patterns
  • Personalized scores
  • Interaction history

Configuration

Interaction Recording

Common Use Cases

{
  "query": "search term",
  "reranking": {
    "enable": true,
    "weights": {
      "feedback": 0.4,
      "popularity": 0.3,
      "collaborative": 0.3
    }
  }
}

Implementation Flow

Best Practices

1

Choose Ranking Method

Start with basic feedback-based ranking before enabling collaborative filtering

2

Configure Weights

Balance between immediate feedback and collaborative patterns

3

Handle Cold Start

Fallback to popularity metrics for new users/items

4

Monitor Performance

Track relevance metrics and adjust weights accordingly

Limitations

Consider these constraints:

  • Collaborative filtering requires sufficient interaction data
  • Cold-start challenges for new users/items
  • Processing overhead for large interaction datasets
  • Memory requirements for similarity matrices

For detailed implementation examples, see the Interactions API Reference.