Conversations

Analyze conversation volume, multi-turn depth, topic clustering, resolution rates, and sentiment scores.

Overview

The Conversations Analytics module provides quantitative and qualitative intelligence on every dialogue conducted between visitors and your AI assistant.

Conversation Metrics Breakdown

Metric	Formula / Definition	Target Benchmark
Total Conversations	Count of distinct dialogue sessions with >= 1 user message	Monitored daily
Average Turns per Dialogue	`Total Messages / Total Conversations`	`3.2 – 5.8 turns`
Resolution Rate	Conversations concluded without escalation to support	`> 82%`
Average Dialogue Duration	Time between first message and last interaction	`1m 45s`
Unanswered Query Rate	Queries resulting in "I don't have verified info"	`< 4.0%`

Automated Topic Clustering

Asiyst uses background embedding clustering to group incoming visitor questions into thematic clusters:

Billing & Pricing: Inquiries regarding subscription tiers, annual discounts, refund policies.

Technical Integration: Questions on SDK setup, API keys, Webhooks, Next.js App Router.

Product Features: Inquiries regarding avatar customization, knowledge sync, rate limits.

Troubleshooting: Inquiries regarding CORS errors, WebGL canvas issues, domain verification.

Sentiment & Satisfaction Scoring

Every completed conversation receives an automated sentiment score (-1.0 negative to +1.0 positive) based on visitor language. In addition, explicit visitor feedback thumbs-up and thumbs-down ratings are aggregated into a Customer Satisfaction (CSAT) score:

$$\text{CSAT} = \left(\frac{\text{Positive Ratings}}{\text{Total Ratings}}\right) \times 100$$

Exporting Conversation Transcripts

Export full sanitized transcripts for compliance auditing or team review:

# Export monthly transcripts to JSON Lines format curl "https://asiyst.com/api/v1/projects/proj_9f8a2b3c4d/analytics/conversations/export?format=jsonl" \

-H "Authorization: Bearer $ASIYST_API_TOKEN" > transcripts.jsonl

Docs/Analytics

Analytics