📊

Usage Tracking

Understanding fragments and how usage is calculated

Outharm uses a fragment-based pricing system for AI moderation to provide transparent and predictable billing. Learn how different content types are counted and billed.

Fragment System

AI moderation uses a fragment-based billing system where different content types consume different amounts of fragments. This approach ensures you only pay for the computational resources your content actually requires.

📝Text Moderation

1

fragment per text field

Each text field in your schema consumes exactly one fragment, regardless of the text length. Whether it's a short title or a long article, each field counts as 1 fragment.

🖼️Image Moderation

5

fragments per image

Each image field in your schema consumes 5 fragments due to the increased computational complexity of image analysis and AI model processing requirements.

Manual Moderation System

Manual moderation uses review units optimized for human review speed and expertise costs. Expert moderators provide nuanced analysis with different processing rates for text and images.

📝Manual Text Review

25

words per review unit

Smaller text chunks allow for careful human analysis of context, nuance, and cultural sensitivity that AI cannot match. Minimum 1 review unit per text field regardless of length.

🖼️Manual Image Review

1

image per review unit

Each image receives individual expert review for visual content analysis and contextual understanding.

Fragment Calculation Examples

Understanding how fragments are calculated is essential for predicting your moderation costs. Here are real-world examples showing different schema configurations for both AI and Manual services.

Example 1: Blog Post Moderation

Schema Configuration

{
  "title": ["Blog Post Title"],
  "content": ["Long blog post content..."],
  "tags": ["technology, AI, moderation"]
}
🤖 AI Moderation
  • • title: 1 text field = 1 fragment
  • • content: 1 text field = 1 fragment
  • • tags: 1 text field = 1 fragment
  • → Comma-separated: "technology, AI, moderation"
3 fragments
👥 Manual Moderation
  • • title: 4 words = 1 review unit
  • • content: 500 words = 20 review units
  • • tags: 3 words = 1 review unit
  • → Under 25 words = 1 review unit minimum
22 review units

Example 2: Social Media Post

Schema Configuration

{
  "caption": ["Check out this amazing view!"],
  "images": ["image1.jpg", "image2.jpg"]
}
🤖 AI Moderation
  • • caption: 1 text field = 1 fragment
  • • images: 2 images = 10 fragments
  • → Each image: 5 fragments
11 fragments
👥 Manual Moderation
  • • caption: 6 words = 1 review unit
  • • images: 2 images = 2 review units
  • → Each image: 1 review unit
3 review units

Example 3: E-commerce Product

Schema Configuration

{
  "product_name": ["Gaming Laptop"],
  "description": ["High-performance laptop for gaming..."],
  "review": ["Great product, highly recommend!"],
  "product_images": ["main.jpg", "side.jpg", "screen.jpg"]
}
🤖 AI Moderation
  • • product_name: 1 text field = 1 fragment
  • • description: 1 text field = 1 fragment
  • • review: 1 text field = 1 fragment
  • • product_images: 3 images = 15 fragments
  • → Each image: 5 fragments × 3
18 fragments
👥 Manual Moderation
  • • product_name: 2 words = 1 review unit
  • • description: 150 words = 6 review units
  • • review: 30 words = 2 review units
  • • product_images: 3 images = 3 review units
  • → 150÷25=6, 30÷25=2 (rounded up)
12 review units

Example 4: Complex Mixed Content

Schema Configuration

{
  "title": ["Community Event Announcement"],
  "description": ["Join us for an amazing community event..."],
  "location": ["Central Park, NYC"],
  "event_poster": ["poster.jpg"],
  "gallery": ["photo1.jpg", "photo2.jpg", "photo3.jpg", "photo4.jpg"],
  "contact_info": ["contact@event.com"],
  "hashtags": ["#community, #NYC, #event"]
}
🤖 AI Moderation
  • • title: 1 text field = 1 fragment
  • • description: 1 text field = 1 fragment
  • • location: 1 text field = 1 fragment
  • • contact_info: 1 text field = 1 fragment
  • • hashtags: 1 text field = 1 fragment
  • → Comma-separated: "#community, #NYC, #event"
  • • event_poster: 1 image = 5 fragments
  • • gallery: 4 images = 20 fragments
30 fragments
👥 Manual Moderation
  • • title: 4 words = 1 review unit
  • • description: 200 words = 8 review units
  • • location: 3 words = 1 review unit
  • • contact_info: 1 word = 1 review unit
  • • hashtags: 3 words = 1 review unit
  • → 200÷25=8, others under 25 words = min 1
  • • event_poster: 1 image = 1 review unit
  • • gallery: 4 images = 4 review units
17 review units

Usage Counting Rules

Usage Counting Rules

🤖AI Moderation Rules

  • • Text: 1 field = 1 fragment (any length)
  • • Images: 1 image = 5 fragments
  • • Arrays: Each item counted separately
  • • Tags/Lists: Combine as comma-separated text
  • • Empty fields: Ignored (0 fragments)

👥Manual Moderation Rules

  • • Text: 25 words = 1 review unit
  • • Images: 1 image = 1 review unit
  • • Minimum rule: Any text field = at least 1 review unit
  • • Word counting: Rounds up (30 words = 2 review units)
  • • Tags/Lists: Combine as comma-separated text
📏 Universal Rules
  • • File format, size, and compression don't affect pricing
  • • Content complexity doesn't change fragment count
  • • Both services process all content simultaneously
  • • Empty/null fields are ignored in both systems

Usage Optimization Tips

💡 Optimization Strategies

🤖AI Moderation Tips
  • • Combine related text fields to reduce fragments
  • • Use for high-volume, straightforward content
  • • Consider image-heavy content carefully (5x cost)
  • • Perfect for initial screening and filtering
  • • Test with free trial before scaling
👥Manual Moderation Tips
  • • Better value for text-heavy content
  • • Use for complex, nuanced decisions
  • • Ideal for appeals and escalations
  • • Images are much more cost-effective
  • • Perfect for sensitive content areas
🎯 Hybrid Strategy
  • • Use AI for initial screening, manual for escalations
  • • AI excels at volume, manual excels at accuracy
  • • Consider content type when choosing service
  • • Monitor both usage patterns in dashboard

Ready to Optimize Your Usage?

Understanding how fragment calculations work helps you predict and optimize your usage across both services. Each service uses a different fragment system based on their processing characteristics and computational requirements.

View Usage Dashboard

Track your fragment consumption in real-time

→

Schema Optimization

Learn to design cost-effective schemas

→

Related Documentation