AI for Performance Analysis¶

Stage 3 is where the AI assistant becomes your performance investigator. After a load test completes, the AI helps you analyze results, identify bottlenecks, interpret metrics trends, and turn raw numbers into recommendations your development team can act on.

This is a reference guide to AI capabilities for analysis. Detailed prompts and examples are embedded in each Analyzing Results section topic where you will actually use them.

What AI Helps With in Analysis¶

Bottleneck Identification¶

AI can: - Identify which pages or transactions are bottlenecks - Determine if bottlenecks are client-side, server-side, or network - Correlate slow response times with server resource usage - Explain why specific pages degrade under load - Suggest which metrics to investigate further

Example prompts: - "What are the bottlenecks in this load test?" - "Why is the checkout page 10x slower than other pages?" - "Is this a database bottleneck or application server bottleneck?" - "Which transactions should I optimize first?"

Where AI guidance appears: - Identifying Bottlenecks - Performance Analysis Workflow - Embedded Analytics Dashboard

Metrics Interpretation¶

AI can: - Explain what response time trends mean - Interpret throughput plateaus - Analyze error rate patterns - Compare performance across different user levels - Identify unusual metric patterns

Example prompts: - "Explain why response times doubled at 75 users" - "What does a throughput plateau at 500 req/sec mean?" - "Why are 95th percentile response times much worse than average?" - "Compare metrics at 25 users vs 100 users"

Where AI guidance appears: - Understanding Metrics - Results Overview - Embedded Analytics Dashboard

Performance Degradation Analysis¶

AI can: - Identify when performance started degrading - Determine if degradation is linear or sudden - Analyze degradation by user level - Correlate degradation with specific transactions - Explain whether degradation is expected or problematic

Example prompts: - "At what user level did performance start degrading?" - "Why did response times suddenly spike after 50 users?" - "Is this gradual degradation normal or a sign of a problem?" - "Show me the degradation trend for the login page"

Where AI guidance appears: - Performance Analysis Workflow - Embedded Analytics Dashboard - Identifying Bottlenecks

Server Resource Correlation¶

AI can: - Correlate application performance with server metrics - Identify which server resources are constraining performance - Explain CPU, memory, disk, or network bottlenecks - Determine if infrastructure is the limiting factor - Suggest server-side optimizations

Example prompts: - "Why are response times slow when CPU is only 40%?" - "Server memory is maxed out. Is this causing the performance issue?" - "What's causing disk I/O spikes during the test?" - "Are these slow response times due to the database or application server?"

Where AI guidance appears: - Identifying Bottlenecks - Server Performance Checklist - Server Monitoring Introduction

Capacity Planning¶

AI can: - Estimate maximum user capacity from test results - Project performance at higher user levels - Identify capacity constraints - Suggest infrastructure scaling strategies - Provide recommendations for handling target load

Example prompts: - "Based on this test, how many users can the system handle?" - "What will response times be at 500 users?" - "What needs to be optimized to support 1000 concurrent users?" - "Is the bottleneck in application servers or database?"

Where AI guidance appears: - Performance Analysis Workflow - Identifying Bottlenecks - Understanding Metrics

Report Generation Assistance¶

AI can: - Summarize key findings from load test - Highlight critical performance issues - Prioritize recommendations by impact - Explain results for non-technical stakeholders - Generate executive summaries

Example prompts: - "Summarize the key findings from this load test" - "What are the top 3 performance issues to fix?" - "Explain these results in non-technical terms for management" - "Prioritize recommendations by business impact"

Where AI guidance appears: - Reports - Performance Analysis Workflow - Embedded Analytics Dashboard

Common Analysis Scenarios¶

Scenario 1: Identifying the Primary Bottleneck¶

Your situation: Load test completed, need to identify what's limiting performance.

Ask the AI:

I ran a load test with 100 users. Response times degraded
significantly. What's the primary bottleneck?

AI will help you: - Analyze response time distribution across pages - Identify which transactions are slowest - Correlate with server resource usage - Determine if bottleneck is application, database, or infrastructure - Provide specific recommendations

Next: Identifying Bottlenecks

Scenario 2: Understanding Unexpected Degradation¶

Your situation: Performance degraded much worse than expected.

Ask the AI:

Response times started at 300ms with 10 users but jumped to
8000ms at 50 users. Why such a dramatic increase?

AI will help you: - Identify if degradation is linear or exponential - Check for resource exhaustion (memory leaks, connection pools) - Analyze which transactions degraded most - Determine if degradation indicates a critical issue - Suggest specific areas to investigate

Next: Performance Analysis Workflow

Scenario 3: Throughput Plateau Investigation¶

Your situation: Adding more users doesn't increase throughput.

Ask the AI:

Throughput plateaued at 500 requests/sec despite adding more
users. What's the limiting factor?

AI will help you: - Identify resource constraints (CPU, memory, threads, connections) - Analyze if application is saturated or infrastructure is maxed - Check for serialization bottlenecks or locking issues - Suggest whether scaling horizontally or vertically helps - Provide recommendations for increasing capacity

Next: Identifying Bottlenecks

Scenario 4: Comparing Different Load Levels¶

Your situation: Need to understand performance at different user counts.

Ask the AI:

Compare performance at 25 users, 50 users, and 100 users.
What changed as load increased?

AI will help you: - Show response time trends across user levels - Identify at what point degradation accelerated - Compare throughput, error rates, and resource usage - Determine optimal operating range - Suggest sustainable user capacity

Next: Understanding Metrics

Scenario 5: Server Resource Analysis¶

Your situation: Application is slow but server metrics are confusing.

Ask the AI:

CPU is at 30%, memory at 50%, but response times are terrible.
Why isn't the server working harder?

AI will help you: - Identify non-CPU bottlenecks (disk I/O, network, database locks) - Analyze if application is waiting on external dependencies - Check for thread starvation or connection pool exhaustion - Determine if bottleneck is outside the application server - Suggest specific monitoring to add

Next: Server Performance Checklist

Scenario 6: Preparing Executive Summary¶

Your situation: Need to explain results to non-technical stakeholders.

Ask the AI:

I need to present load test results to management. Summarize
the key findings and recommendations in business terms.

AI will help you: - Identify critical performance issues - Explain impact in business terms (user experience, capacity limits) - Prioritize recommendations by business value - Provide estimated costs of NOT fixing issues - Generate executive-friendly summary

Next: Reports

When to Use AI vs Manual Documentation¶

Use AI when:¶

✅ You need to interpret specific results - "Why did THIS test show degradation?"

✅ You want bottleneck identification - "What's the primary issue in MY results?"

✅ You need actionable recommendations - "What should I fix first?"

✅ You want trend analysis - "How did performance change across user levels?"

✅ You need context-specific insights - AI sees YOUR specific test results and metrics

Use the Manual when:¶

📖 Learning analysis concepts - What metrics matter? How to read charts?

📖 Understanding methodologies - Performance analysis workflow, bottleneck identification techniques

📖 Reference information - What does 95th percentile mean? How are metrics calculated?

📖 Planning your analysis - Understanding analysis strategy before diving into results

📖 Comprehensive background - Deep dive into performance engineering concepts

Effective Analysis Prompts¶

✅ Good Analysis Prompts¶

Specific and contextual: - "The checkout page response time jumped from 500ms to 5000ms at 75 users. What's causing this?" - "CPU is only at 40% but response times are 10x slower. What's the bottleneck?" - "Throughput plateaued at 400 req/sec. Which resource is constraining capacity?"

Include relevant details: - "At 25 users, average response time was 300ms. At 100 users, it's 2000ms. Is this linear degradation or exponential?" - "The database server shows 95% CPU, but application server is at 30%. Is the database the bottleneck?"

Ask for actionable recommendations: - "Based on these results, what should I optimize first?" - "What infrastructure changes would improve performance?" - "How many users can the system handle with current configuration?"

❌ Avoid Vague Prompts¶

Too general: - "Analyze my test" (What specifically should AI analyze?) - "The test is slow" (Which pages? At what user level?) - "What's wrong?" (Be specific about symptoms)

Missing context: - "Is this good performance?" (Compared to what? What are your goals?) - "Should I be concerned?" (About which metrics? What's the business impact?)

No clear question: - "Here are my results" (And what would you like to know about them?) - "Look at this" (The AI is looking. Tell it what to look for.)

AI Analysis Workflow¶

Here's how AI fits into the typical analysis workflow:

1. Load Test Completes¶

→ Automatic step

2. Open Results View¶

→ Manual step

3. Ask AI: "What are the key findings from this test?"¶

→ AI provides high-level summary

4. Review Metrics Charts¶

→ Manual step: Look at response times, throughput, errors

5. Ask AI: "What's causing the performance degradation?"¶

→ AI identifies bottlenecks

6. Drill Into Specific Transactions¶

→ Manual step: Click on slow pages

7. Ask AI: "Why is this page so slow?"¶

→ AI analyzes transaction-specific issues

8. Review Server Metrics¶

→ Manual step: Check CPU, memory, disk

9. Ask AI: "Correlate server metrics with application performance"¶

→ AI explains resource constraints

10. Generate Report¶

→ Manual step: Create executive summary

11. Ask AI: "Prioritize recommendations by impact"¶

→ AI ranks fixes by business value

Key point: The AI guides your analysis and surfaces patterns you might miss, but you make all decisions about what to investigate and what to fix.

Using the Embedded Analytics Dashboard¶

The v7.0 Embedded Analytics Dashboard integrates AI-powered analysis directly into the results view. This means you can:

✅ Ask questions while looking at metrics - AI sees the same data you're viewing

✅ Get instant explanations - Click on any chart and ask "Why did this spike?"

✅ Drill down with AI guidance - AI suggests which metrics to investigate next

✅ Generate insights automatically - Dashboard highlights anomalies AI detected

See: Embedded Analytics Dashboard for details on using AI-powered interactive analysis.

Next Steps¶

Learn AI for configuration - Debug test cases and fix correlation issues (Stage 1)
Learn AI for monitoring - Real-time guidance during active tests (Stage 2)
Understand AI limitations - What AI can and cannot do

Or: Return to the Analyzing Results section and look for AI guidance embedded in each topic: