Velaro includes six built-in analytics reports. This guide explains what each metric means, what good looks like, and what actions to take when numbers are off.

---

Value Report

The Value Report shows ROI from chat automation over a selected time window (30, 60, or 90 days).

Key metrics:

• Bot Deflection Rate — percentage of conversations fully handled by the AI bot without agent involvement. 50–70% is good for most accounts. Below 40% means the bot needs more training or workflows are routing too aggressively to humans.
• Hours Saved — estimated agent hours saved by bot deflections. Calculated as: deflections × avg handle time ÷ 60.
• Estimated Savings — hours saved × configured agent cost per hour. Adjust the cost input to match your actual agent cost.
• AI Skill Executions — how many times the bot used an integration skill (CRM lookup, order status, etc.). High skill usage = high automation value.
• Resolution Rate — conversations where the customer's issue was resolved without escalation. Above 80% is healthy.
• Avg First Response Time — how quickly a human or bot first responded. Under 60 seconds is excellent. Over 3 minutes risks abandonment.

SLA Compliance — percentage of conversations where the first human response was within the configured SLA target (default 5 minutes). Below 90% usually means understaffing or routing issues during peak hours.

Actions when deflection rate is low:

• Check which intents are falling through to agents in Chat Intelligence
• Add or improve bot workflows for your top 5 missed topics
• Review Transfer step conditions — overly broad conditions hand off too early

Actions when response time is high:

• Check team routing — are conversations queuing in the right team?
• Review agent availability during peak hours
• Enable bot fallback responses for after-hours gaps

---

Agent Performance

AI-scored conversation quality for individual agents, based on a rubric applied to recent chats.

Key metrics:

• Avg Quality Score — 0–100 AI score across dimensions: resolution, empathy, professionalism, accuracy. Above 80 is good.
• Sentiment Breakdown — positive / neutral / negative customer sentiment. More than 15% negative warrants investigation.
• Per-agent scores — click any agent to see their individual conversation samples and where they lost points.

Actions when scores are low:

• Review low-scoring conversations with the agent directly
• Agents scoring below 65 consistently should be coached on resolution and empathy dimensions
• Agents with high volume but low scores may be rushing — check handle time vs quality tradeoff

---

CSAT Results

Customer satisfaction survey scores collected after chat conversations end.

Key metrics:

• Satisfaction Rate — percentage of surveys rated 4 or 5 out of 5. Industry benchmark is 85%+. Below 75% is a signal something is wrong.
• Top Teams — which teams are getting the highest scores. Large gaps between teams point to coaching or process differences.
• Score distribution — 1–2 star ratings need individual review. These are customers who had a bad enough experience to say so.

Actions when CSAT is low:

• Pull the 1–2 star conversations and look for patterns (long wait, wrong answer, transfers, bot failures)
• Check if low scores correlate with a specific channel, team, or time of day
• Configure surveys in Settings → CSAT if response rate is too low to draw conclusions

---

Chat Intelligence

AI-analyzed conversation themes — what customers are actually asking about and how well each topic is being handled.

Key metrics:

• Top Themes — the most common topics customers raise. These should align with your bot's trained workflows. If a top theme has no bot workflow, that's a deflection opportunity.
• Missed Conversations — conversations where the customer's issue was NOT resolved. Grouped by theme to show where you're losing customers.
• AI Resolved vs Agent Resolved vs Deflected — the three outcome buckets. More AI-resolved = more efficient. More missed = more churn risk.
• Missed by Page — which pages on your site generate the most unresolved conversations. Signals where to add proactive chat triggers or improve content.

Actions when a theme has high missed rate:

• Build or improve a bot workflow for that topic
• Add a knowledge base article and enable KB search for the bot
• Check if the bot is transferring these to the right team

---

Leaderboard

Ranked agent performance across configurable metrics, with optional team filter and competition mode.

Key metrics (selectable):

• Total Chats — raw conversation volume. Useful for workload planning.
• CSAT — satisfaction score per agent. Most meaningful for quality-focused teams.
• Resolution Rate — how often the agent resolves without escalation.
• Avg Handle Time — faster is not always better — pair with CSAT to distinguish efficiency from rushing.
• First Response Time — how quickly agents pick up conversations.

Competitions — create time-bounded competitions with prizes for top performers. Visible to agents on their own dashboard.

Actions:

• Use leaderboard monthly in team meetings — public recognition drives engagement
• Look at agents in the bottom quartile for two consecutive periods — that's a coaching conversation
• Filter by team to compare within peer groups rather than across different-volume teams

---

KB Performance

How your knowledge base articles are performing on help.velaro.com.

Key metrics:

• Views — how often each article is read. High-view articles are your most important content — keep them accurate.
• Search queries — what customers search for. Queries with no results = article gaps.
• Bot deflections via KB — how often the AI bot cited a KB article to answer a question.

Actions when search has no-result queries:

• Write articles for the top no-result queries
• Add synonyms and tags to existing articles that cover those topics

---

General Guidance

Trend direction matters more than absolute numbers. A 60% deflection rate that is rising is better than a 70% rate that is falling.

Compare periods, not snapshots. Always use the period selectors to compare this month vs last month before drawing conclusions.

Bot + Agent = one system. Low agent CSAT often traces back to the bot failing first — a bad handoff, a wrong answer, or a long queue wait created by the bot not deflecting enough.