Skip to main content

Analytics

The Analytics tab gives you a clear picture of how Lira is performing as your support layer — how many conversations it's handling autonomously, how customers are rating their experience, and where escalations are happening.

Navigate to Support → Analytics.


Overview tab

The Overview tab shows your aggregate performance metrics.

Key metrics

Four headline numbers appear at the top:

MetricWhat it tracks
Total ConversationsAll conversations received in the selected period
AutonomousHow many were resolved by Lira without human intervention
EscalatedHow many were flagged for human review
Avg CSATAverage customer satisfaction score (out of 5) across all rated conversations

A high Autonomous count and a high Avg CSAT together indicate a healthy support operation — Lira is resolving conversations independently and customers are satisfied with the quality.


Resolution breakdown

A horizontal bar chart showing how conversations resolved:

CategoryMeaning
Autonomous (green)Resolved by Lira without human intervention
Human (blue)Resolved by your team after manual review
Escalated (red)Currently escalated or closed after escalation
Open (amber)Still active — not yet resolved

A high autonomous rate means your Knowledge Base is well-populated and your confidence threshold is calibrated well. A high escalation rate may indicate gaps in your Knowledge Base or a threshold that's set too conservatively.


Average first response time

The average time between a customer's first message and Lira's first reply. Because Lira responds programmatically, this is typically sub-5 seconds for chat and widget conversations, and slightly longer for email (due to email delivery latency).

This metric matters for customer experience and SLA compliance. If you've set an SLA Target in escalation settings, tickets that breach it appear highlighted in your Inbox.


Monthly usage

A simple count of:

  • Conversations this month — total conversations started in the current calendar month
  • AI replies this month — total AI-generated replies sent

Compare these against your plan limits (visible in Settings → Behaviour → Volume & Limits) to track your usage trajectory.


Weekly Report tab

The Weekly Report gives you a snapshot of the most recent 7-day period — useful for Monday reviews and team reporting.

Summary card

Shows the date range covered and four metrics side-by-side:

  • Conversations
  • Autonomous resolutions
  • Escalations
  • Average CSAT

Top intents

A ranked list of the most common intents detected in conversations during the week. This reveals:

  • What your customers ask about most
  • Whether your Knowledge Base has adequate coverage for those topics
  • Opportunities to create proactive triggers for high-volume, low-complexity intents

If billing or refund appears often, for example, consider adding a detailed billing FAQ to your Knowledge Base.

Knowledge improvement

Shows how many Knowledge Base drafts were created and approved during the week. When Lira encounters a question it can't answer well, it creates a Knowledge Base draft suggesting what content should be added. Reviewing and approving these drafts (in Support → Knowledge or Knowledge Base → Documents) directly improves future response quality.


Using analytics to improve performance

If autonomous rate is low:

  • Your Knowledge Base may be sparse — add more documents and connected sources
  • Your confidence threshold may be too high — try lowering it from Settings → Behaviour
  • Check top intents to find which topics Lira is struggling with

If CSAT is low:

  • Read the conversations where customers rated 1–2 stars
  • Identify what Lira said that didn't help
  • Add or update the relevant content in your Knowledge Base
  • Check if force-escalate intents should catch those types of questions earlier

If escalations are high:

  • Same diagnosis as low autonomous rate — usually a Knowledge Base gap
  • Check if any specific intent is driving escalations disproportionately (visible in Top Intents)

If first response time is slow:

  • For email: this is usually latency in email delivery from the customer's provider, not Lira's processing time
  • For chat/widget: if response times are above a few seconds, contact support