Progress, timing, and performance expectations

How progress works: The progress bar tracks four stages per test:

1. Preparing — converting the ticket into a test scenario
2. Simulating — GPT-4 generates a multi-turn customer/agent conversation
3. Evaluating — GPT-4 scores the conversation across 4 metrics
4. Complete — results are ready to view

Timing expectations:

• Single ticket test: 15–30 seconds
• Batch of 10 tickets: 2–4 minutes
• Batch of 25 tickets: 5–10 minutes
• Batch of 50+ tickets: 10–20 minutes

The main bottleneck is OpenAI API latency — each ticket requires two API calls (simulation + evaluation). Tickets are processed sequentially in the current version.

If tests are slow:

• Check your OpenAI API key has sufficient credit/quota
• Reduce batch size
• Check your internet connection
• Larger batches with async processing are planned for v1.2