Analytics

Analytics provides detailed, real-time visibility into how your team uses the platform. Analytics data is always scoped to the currently selected project and is refreshed from the API with results cached for up to 5 minutes. To access Analytics settings, select Settings > Analytics. Analytics includes these elements:

• Page header displays the name of the current project, along with Refresh and Export to Excel buttons on the right.
• Refresh button manually refreshes analytics data, updating the currently active tab with the latest data from the API.
• Export to Excel button downloads a comprehensive Excel workbook containing all analytics data (Overview, Costs, Tokens, Agents & Pipelines, Tools, Users, and Health) for the selected date range. The export includes metadata about the project, date range, and timezone.
• Date filter bar contains the four quick-preset buttons (Last 24h, Last 7d, Last 30d, Last 90d) on the left and the From / To datetime pickers on the right.
• Tabs: Eight tabs to switch between different analytics views (Overview, Costs, Tokens, Agents & Pipelines, Tools, Users, Health, and Guide).
• Content area: The main scrollable panel that renders KPI cards, charts, and tables for the active tab.

In Analytics, you can see eight tabs:

​

Date Range Controls

You can control the time window applied to all tabs using the filter bar at the top of the page. Quick Presets Use four preset buttons to set the date range with a single click: Custom Date/Time Pickers Use the From and To datetime pickers for precise time windows:

• From sets the start of the analysis period. Cannot be set later than To.
• To sets the end of the analysis period. Cannot be set earlier than From.

Both pickers support entering a specific date and time (24-hour format). The pickers include Clear and Ok action buttons. After adjusting the custom pickers, the active preset button is automatically deselected.

​

Overview

Overview displays project-wide KPI cards and summary charts for the selected date range.

​

KPI Cards

KPI cards show the most important metrics for the selected date range.

​

Daily Activity

A multi-series area chart showing AI usage trends over time:

• LLM Calls (purple): AI model invocations per day.
• Tool Runs (orange): Tool executions per day.
• Agent & Pipeline Runs (green): Agent and pipeline interactions per day.
• Active Users (blue, Team Projects only): unique active users per day.

Use this chart to spot usage spikes, identify quiet periods, or track adoption trends over time. For Personal Projects, the Active Users series is not displayed.

​

Top 5 AI Adopters

A leaderboard table that shows five users with the most combined AI events (LLM + Tool + Agent). Each row displays:

• Rank and color-coded avatar (gold/silver/bronze for top 3).
• User email.
• Per-type breakdown: N LLM · N Tool · N Agent.
• Total AI events score.

Clicking a row navigates you directly to the user’s detail view in Users.

​

Costs

Costs shows estimated LLM spend for the selected date range.

​

KPI Cards

Five KPI cards appear at the top of the Costs tab:

​

Daily Cost Trend

Shows estimated daily cost as a bar chart.

• The X axis shows the date.
• The Y axis shows estimated cost.

If no daily cost data is available for the selected date range, ELITEA shows No data.

​

Cost by User

Ranks users by estimated total cost. If no user cost data is available for the selected date range, ELITEA shows an empty-state message.

​

Cost by Model

Ranks models by estimated total cost. If no model cost data is available for the selected date range, ELITEA shows an empty-state message.

​

Cost by Agent & Pipeline

Ranks agents and pipelines by estimated total cost. If no agent or pipeline cost data is available for the selected date range, ELITEA shows an empty-state message.

​

Tokens

Tokens provides detailed token usage analytics for the selected date range, including cache token tracking.

​

KPI Cards

Five KPI cards appear at the top of the Tokens tab:

​

Daily Token Usage

A multi-bar chart showing daily total, input, and output token usage for LLM requests in the selected date range:

• Total Tokens (primary bar): combined token usage per day
• Input Tokens (second bar): prompt tokens per day
• Output Tokens (third bar): completion tokens per day
• Cache Read Tokens (fourth bar): cached tokens read per day
• Cache Write Tokens (fifth bar): tokens written to cache per day

If no token data is available for the selected date range, ELITEA shows No token usage is available for the selected date range.

​

Token Usage by User

Shows token usage attributed to each project user during the selected date range. If no user token data is available for the selected date range, ELITEA shows an empty-state message.

​

Token Usage by Model

Shows token usage for each AI model used in the selected date range. If no model token data is available for the selected date range, ELITEA shows an empty-state message.

​

Token Usage by Agent & Pipeline

Shows token usage attributed to Agents and Pipelines during the selected date range. If no agent or pipeline token data is available for the selected date range, ELITEA shows an empty-state message.

​

Agents & Pipelines

Agents & Pipelines show how individual agents (applications) are used within the project.

​

Most Active Agents

A bar chart showing the top 20 agents ranked by total event count. Each bar is color-coded and labeled with the agent name.

​

Chat Messages

An area chart showing the number of user messages (SIO chat_predict events) sent per day. Useful for tracking chat engagement trends independent of agent-specific metrics.

​

Agent Activity

A paginated, searchable table listing all agents and pipelines in the project: Use the search box (top-right of the table) to filter agents and pipelines by name. Page size can be set to 10, 20, or 50 rows per page.

​

Agent Details

Click any agent row to open a drill-down view showing:

• Header: Agent name with a back arrow to return to the full list.
• KPI Cards: Total Events, Unique Users, Avg Latency, Errors, Error Rate, Total Tokens, Input Tokens, Output Tokens, Total Cost, and Avg Cost / Call.
• Runs by Day: Area chart showing events and errors per day for this agent.
• Users: Lists each user who interacted with the agent or pipeline, with per-user events, average latency, and errors.
• Tools: Lists each tool called by this agent or pipeline, with call count.

To return to the Agents list, click the back arrow in the header.

​

Tools

Tools show usage patterns for individual tools executed by agents or users.

​

Most Popular Tools

A bar chart showing the top 20 tools ranked by number of calls. Tool names on the X axis, call count on the Y axis.

​

Tool Details

A paginated, searchable table listing all tools used in the project: Use the search box to filter tools by name. Page size options: 10, 20, or 50.

​

Tool Details View

Click any tool row to open a drill-down view:

• KPI Cards: Total Calls, Unique Users, Avg Latency, Errors, Error Rate.
• Daily Usage Chart: Area chart showing calls and errors per day for this tool.
• Users table: Users who called this tool, with per-user calls, average latency, and errors.
• Agents table: Agents that used this tool, with call count (resolved by correlating trace IDs).

To return to the Tools list, click the back arrow.

​

Users

Users provides per-person analysis of platform activity.

​

User Activity

A paginated, searchable table listing all users active in the project during the selected period: Use the search box to filter by email address. Page size options: 10, 20, or 50.

​

User Details

Click any user row to open a drill-down view:

• Header: User email with a back arrow.
• KPI Cards: Active Days, LLM Calls, Tool Calls, Agent & Pipeline Runs, Chat Msg, Errors, Total Tokens, Input Tokens, Output Tokens, and Total Cost.
• Daily Activity Chart: Area chart with four series — LLM, Tool, Chat Msg, and Agent — showing how this user’s activity is distributed over time.
• Models Used list: AI models this user queried, with call counts.
• Tools Used list: Tools this user triggered, with call counts.
• Agents & Pipelines Used list: Agents and pipelines this user interacted with, with run counts.

​

Health

Health provides system reliability metrics, helping you identify error patterns and latency issues.

​

Requests vs Errors

A dual-series area chart showing:

• Total Requests (blue): all events per day.
• Errors (red): events that resulted in an error per day.

Spikes in the red area indicate periods with elevated error rates.

​

Health by Event Type

A breakdown table showing reliability per event type: Event types tracked:

​

Guide

Guide is a built-in metric glossary embedded directly in Analytics. It explains every KPI, chart, and table column with:

• Description: What the metric represents in plain language.
• Calculation: The exact formula used to compute the value.
• Data source: Which event types or platform actions contribute to the metric.

Guide covers all seven tabs in Analytics: Study all metric definitions in Guide before drawing conclusions from the data.

​

Limitations

​

Practical Examples

Measuring AI Adoption Across Your Team

1. Open Settings → Analytics.
2. Select the Last 30d preset.
3. On the Overview tab, read the TEAM and AI ACTIVE KPI cards.
4. The Adoption Rate badge (e.g., ↑ 72%) tells you what proportion of your registered team members actively used AI features in the past 30 days.
5. Scroll to the Top 5 AI Adopters leaderboard to identify your most active members — click any name to drill into their individual activity in the Users tab.

Identifying Underused or Overloaded Agents

1. Open the Agents tab.
2. Use the Most Active Agents bar chart to quickly see which agents receive the most traffic.
3. In the Agent Activity table, sort mentally by the Events column (highest first by default) to find heavily used agents.
4. Click any agent to open its detail view and check Avg Latency and Error Rate. A high error rate (> 5%) highlights agents that may need debugging or configuration review.
5. Agents with zero events in the selected period may be deprecated or not yet discovered by your team.

Auditing Tool Usage and Reliability

1. Open the Tools tab.
2. The Most Popular Tools chart gives an at-a-glance view of which integrations are relied upon most.
3. In the Tool Details table, look for tools with a non-zero Errors count (displayed in red).
4. Click a tool with errors to open its detail view: the Daily Usage chart reveals when errors spiked, and the Users sub-table shows which users encountered them.
5. Cross-reference with the Health tab to check the system-wide error rate for the tool event type during the same period.

Reviewing an Individual User's Activity

1. Open the Users tab.
2. Use the search box to filter by the user’s email address.
3. Click the user’s row to open their detail view.
4. Review the Daily Activity area chart to see on which days they were most active and which event types dominate.
5. The Models Used, Tools Used, and Agents Used lists show exactly which platform resources this user engaged with — useful for onboarding support or license reviews.

Investigating an Error Spike

1. Open the Health tab.
2. In the Requests vs Errors chart, identify the date range of the error spike.
3. In the Health by Event Type table, find the row with the highest Error Rate (values > 5% are highlighted in red).
4. Note the event type (e.g., llm or tool) and navigate to the corresponding tab (Agents or Tools) to identify which specific agent or tool was responsible.
5. Narrow the date range using the From/To pickers to focus on the spike period, then re-examine the relevant tab.

Monitoring Token Usage and Optimizing Costs

1. Open the Tokens tab to get a complete view of token consumption.
2. Review the Daily Token Usage chart to identify days with unusually high token consumption.
3. Check the Token Usage by Model table to see which AI models are consuming the most tokens. Models with high output tokens may indicate verbose responses that could be optimized.
4. In the Token Usage by Agent & Pipeline table, identify agents with disproportionately high token usage relative to their value.
5. Look at CACHE READ and CACHE WRITE KPI cards — high cache read percentages indicate good prompt caching efficiency, which reduces costs for supported models (e.g., Claude).
6. Cross-reference with the Costs tab to understand the financial impact of token usage patterns.

Evaluating Prompt Caching Effectiveness

1. Open the Tokens tab and note the CACHE READ and CACHE WRITE values.
2. Calculate cache hit ratio: divide CACHE READ tokens by (CACHE READ + INPUT tokens) × 100%.
3. Review Token Usage by Agent & Pipeline to identify which agents benefit most from caching.
4. Agents with high cache read tokens are efficiently reusing prompts, reducing costs by up to 90% for cached portions.
5. Agents with low or zero cache tokens may benefit from prompt restructuring to enable caching — place static instructions at the beginning of prompts.
6. Use Daily Token Usage to track how caching efficiency changes over time after prompt optimizations.

​

Best Practices

Start with a Meaningful Date Range

Before reading any metrics, set the date range that matches your analysis goal. Use Last 7d for recent activity reviews, Last 30d for monthly reporting, and custom pickers for audit periods tied to specific events or releases.

Use the Guide Tab Before Drawing Conclusions

Open the Guide tab to review exact metric definitions and calculation formulas. Metrics like TEAM (active users vs. total ever seen) and AI ACTIVE (LLM/Tool/Agent users only) have specific scopes that affect interpretation.

Combine Overview and Users Tabs for Adoption Analysis

Use the Overview tab to get the aggregate adoption picture, then drill down via the Top 5 AI Adopters leaderboard or the Users tab to identify specific individuals for coaching or recognition.

Monitor Health Regularly

Review the Health tab after deployments or integrations changes. A sudden increase in llm or tool error rates often signals a misconfigured AI model or a broken external service credential.

Cross-Reference Agents and Tools Tabs

When an agent shows high errors in the Agents tab, open its detail view and check the Tools sub-table. Then visit the Tools tab, click the problematic tool, and check which other agents also rely on it — helping you assess the blast radius of a broken integration.

Leverage Cross-Tab Navigation

You can navigate directly from the Overview leaderboard to a specific user’s detail view without going through the Users tab manually. Similarly, using the back arrow in any detail view returns you smoothly to the list without losing your sort or search state.

Use Tokens Tab to Optimize AI Costs

The Tokens tab is your primary tool for cost optimization. Start by identifying the biggest token consumers in the Token Usage by Model and Token Usage by Agent & Pipeline tables. Focus optimization efforts on the top 20% of consumers, which typically account for 80% of total usage. For models supporting prompt caching (like Claude), monitor cache read tokens — a healthy cache hit rate can reduce costs by 50-90% on repeated prompts. If cache tokens are low, restructure prompts to place static instructions at the beginning, enabling the provider’s caching system to work effectively.

Track Token Trends After Prompt Changes

Whenever you modify agent prompts or system instructions, use the Tokens tab to measure the impact. Set a custom date range spanning one week before and one week after the change. Compare Daily Token Usage patterns to see if input tokens decreased (more concise prompts) or if output tokens changed (different response length). Review Token Usage by Agent & Pipeline to isolate the impact to specific agents. This data-driven approach helps validate that prompt engineering efforts are actually reducing token consumption and costs.

Monitor Per-User Token Consumption for Fair Usage Policies

If your organization has fair usage policies or user quotas, use the Token Usage by User table in the Tokens tab to monitor individual consumption. Sort by TOTAL TOKENS to identify outliers. Click through to the Users tab for detailed daily activity patterns. High token users may be running inefficient workflows, testing extensively, or legitimately processing large volumes. Combine token data with the Costs tab to understand the financial impact per user, helping inform capacity planning and budget allocation decisions.