Skip to main content

User Cost Metrics

This section shows where your expenses are divided among your users and shapes usage habits. You can see who is utilizing your AI the most and what it is costing you.

user-cost-metrics

  • Active Users: Number of users who made LLM calls within the time interval selected
  • Avg Cost/User: Average cost per active user for all activity
  • Cost P95: 95th percentile cost, showing what high-usage users typically pay
  • Top User: Most cost-causing user and how much they spend

Table Top Cost Users​

The breakdown by user has per-user detailed information:

Table Columns​

  • Rank: Ranking of users by overall cost contribution (1st, 2nd, 3rd, etc.)
  • User ID: Each user's unique ID (e.g., researcher-789, user-123)
  • Total Cost: Total spent per user with input/output cost split shown
  • Calls: Discrete LLM calls made by individual user
  • Avg/Call: Average spent per discrete call by that user
  • Tokens: Total tokens processed by that specific user
  • Top Model: Most used model by each user and usage ratio

User Analysis​

The table leads you to understand:

  • Cost Distribution: Notice how the expenditure is spread over your users
  • Patterns of Use: Heavy users vs light users
  • Model of Choice: What models individual users like
  • Efficiency of Call: Compare cost per call between different users

Summary of User Metrics​

More context added:

  • Total Users: Total number of users who had some LLM usage
  • Average Cost: Average expenditure across all active users
  • Ranking of Users: Clear ranking system to determine highest spenders

Perfect for understanding user behavior, identifying heavy users, planning user-based billing or limits, and optimizing resource allocation based on actual usage patterns.

Key Benefits​

User Cost Metrics Overview helps you:

  • User accountability: Understand which users or teams are driving costs
  • Usage insights: See how different users interact with your AI systems
  • Billing preparation: Get data needed for user-based billing models
  • Resource planning: Allocate resources based on actual user demand
  • Cost optimization: Run the workflows of heavy users in conjunction
  • Fair usage policies: Apply limits according to real usage patterns

Table of Contents