Analytics

Deep dive into your AI spending patterns and optimization opportunities.

Overview

Analytics gives you advanced insights into:

• Spending trends and forecasts
• Model efficiency comparisons
• Customer profitability
• Feature cost analysis
• Optimization opportunities

Spending Trends

Time Series Analysis

View cost trends over:

• Hourly - Last 24 hours
• Daily - Last 7, 30, or 90 days
• Weekly - Last 12 weeks
• Monthly - Last 12 months

Chart features:

• Zoom and pan
• Compare time periods
• Overlay different metrics
• Export as PNG/CSV

Forecast

AI-powered spending forecast:

• Predicts next 30 days
• Based on historical patterns
• Accounts for trends and seasonality
• Confidence intervals shown

Example:

Current month: $450 (18 days elapsed)
Forecast: $750 ± $50 (end of month)
Recommendation: On track for budget

Anomaly Detection

Automatically detects unusual patterns:

• Unexpected cost spikes
• Model switching
• Traffic patterns
• Error rate changes

Alert types:

• 🔴 Critical - Costs 3x normal
• 🟡 Warning - Costs 2x normal
• 🔵 Info - Unusual but not concerning

Model Performance

Cost per Model

Compare models by:

• Total cost
• Cost per 1,000 tokens
• Average latency
• Error rate
• Usage frequency

Example table:

Model Recommendations

AI suggests model optimizations:

Example:

💡 Optimization Opportunity

Feature: content-summarization
Current: gpt-4 ($234/month)
Suggested: claude-3-sonnet ($89/month)

Savings: $145/month (62% reduction)
Quality: Comparable (based on benchmarks)
Action: Run A/B test →

Token Efficiency

Analyze token usage:

• Average tokens per request
• Prompt vs completion ratio
• Token waste detection
• Optimization suggestions

Token waste indicators:

• Long system prompts (repeated unnecessarily)
• Excessive context (entire docs when summary needed)
• Redundant API calls (caching opportunities)

Customer Analytics

Customer Profitability

See which customers are profitable:

Profitability score:

Revenue - (AI costs + other costs) = Profit
Profit / Revenue = Margin %

Dashboard shows:

• Top 10 most profitable customers
• Top 10 least profitable customers
• Customer margin distribution
• Churn risk by profitability

Customer Segments

Segment customers by:

• Spending level (low/medium/high)
• Growth rate (growing/stable/declining)
• Model preference (GPT-4, Claude, mix)
• Feature usage (power users vs casual)

Use cases:

• Pricing tier placement
• Upsell opportunities
• Churn prevention
• Feature targeting

Cohort Analysis

Track customer cohorts over time:

• Signup month cohorts
• Spending evolution
• Retention by cost tier
• Lifetime value (LTV)

Feature Analytics

Feature Performance

Compare features by:

• Total cost
• Cost per user
• Usage frequency
• User satisfaction (if tracked)

Example:

Feature ROI

Calculate ROI per feature:

Formula:

Revenue from feature - Cost of feature = Profit
Profit / Cost = ROI %

Requires revenue attribution (manual input or integration).

Feature Optimization

AI suggests feature improvements:

Example:

📊 Feature Insight: email-summarization

Current usage:
- 450 calls/day
- $8.50/day cost
- gpt-4 model

Opportunity:
Switch to claude-3-haiku for emails <500 words
Estimated savings: $4.20/day ($126/month)
Quality impact: Minimal

Cost Optimization

Optimization Score

Overall score (0-100) based on:

• Model selection efficiency
• Token usage efficiency
• Caching opportunities
• Error rate
• Rate limit headroom

Example:

Your Optimization Score: 72/100

✅ Strengths:
- Low error rate (0.3%)
- Good model selection

⚠️ Opportunities:
- High token waste (18%)
- No caching implemented
- Inefficient prompts detected

Potential savings: $234/month

Caching Opportunities

Identify repeated requests:

• Same prompts sent multiple times
• Similar prompts (can be cached)
• Static context (system prompts)

Savings calculator:

Repeated calls detected: 1,240/week
Current cost: $45/week
With caching: $12/week
Savings: $33/week ($143/month)

Prompt Optimization

Analyze prompts for efficiency:

• Token count distribution
• Long prompts (>500 tokens)
• Repeated instructions
• Unnecessary context

Suggestions:

• Shorten system prompts
• Remove redundant context
• Use prompt templates
• Implement prompt compression

Reports

Automated Reports

Schedule reports:

• Daily - Cost summary, top models
• Weekly - Trends, anomalies, recommendations
• Monthly - Full analytics, forecasts, optimization

Delivery:

• Email (PDF attached)
• Slack (summary + link)
• Webhook (JSON data)

Custom Reports

Build custom reports:

1. Choose metrics (cost, tokens, latency, etc.)
2. Choose dimensions (model, customer, feature)
3. Choose time range
4. Add filters
5. Schedule or download

Export formats:

• PDF (with charts)
• CSV (raw data)
• JSON (via API)
• Excel (Pro+)

Report Templates

Pre-built templates:

• Executive Summary - High-level overview for leadership
• Engineering Metrics - Latency, errors, performance
• Finance Report - Costs, forecasts, budgets
• Customer Report - Per-customer analytics

Integrations

Data Export

Export analytics data to:

• Google Sheets - Auto-sync daily/weekly
• Excel - Download or email
• Data warehouse - BigQuery, Snowflake (Enterprise)
• BI tools - Tableau, Looker, Power BI (Enterprise)

API Access

Programmatic access to analytics:

// Get cost analytics
const analytics = await fetch('https://api.aispendtrack.com/v1/analytics/costs', {
  headers: {
    'Authorization': 'Bearer your_api_key',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    start_date: '2024-01-01',
    end_date: '2024-01-31',
    group_by: 'model'
  })
});

Full API documentation →

Best Practices

1. Review Weekly

Set aside time each week:

• Check spending trends
• Review top costs
• Implement one optimization
• Track savings

2. Set Baselines

Establish baselines for:

• Average daily cost
• Cost per customer
• Cost per feature
• Token efficiency

Track improvements over time.

3. A/B Test Optimizations

Before making changes:

• Implement change for 10% of traffic
• Monitor for 7 days
• Compare quality and cost
• Roll out if successful

4. Automate Reports

Set up automated reports:

• Daily summary for yourself
• Weekly report for team
• Monthly report for finance/leadership

5. Track ROI

Measure impact of optimizations:

• Cost before optimization
• Cost after optimization
• Savings per month
• Quality impact (if any)

Analytics Roadmap

Coming soon:

• 🔜 Real-time analytics (sub-second updates)
• 🔜 Custom dashboards
• 🔜 Advanced forecasting (ML-based)
• 🔜 Benchmark against similar companies
• 🔜 Carbon footprint tracking
• 🔜 Multi-account rollups (Enterprise)

Next Steps

• Set up automated reports
• Export data to Google Sheets
• Access analytics via API