Loading
Most businesses are drowning in web analytics data and starving for insight. Pageviews and bounce rate are the metrics everyone reports and almost nobody acts on. Here are the engagement metrics that actually connect to revenue — and what to do when they're telling you something is wrong.
What percentage of visitors open your chat widget? A rate below 2% typically signals that your widget placement is poor, your opening message is uninviting, or your visitors don't trust it. A rate above 8% — especially with proactive triggers — is a sign of a well-optimised engagement setup. Track this per page type: pricing pages should have the highest initiation rate; if they don't, something is broken in your most important conversion corridor.
The percentage of chat conversations fully resolved by AI without human escalation. A healthy deflection rate is 55–70% — high enough to reduce agent workload substantially, low enough to suggest your AI isn't dismissing complex queries it should be escalating. If deflection is above 90%, audit the chat histories: your AI may be giving overconfident or incorrect answers that visitors are accepting without pushback.
The time between a visitor sending their first message and receiving their first response — whether from AI or a human agent. AI should always be sub-second. Human-escalated responses should target under 30 seconds during business hours and under 2 minutes around the clock. Every additional minute of first response time correlates with measurable drops in satisfaction score and conversion probability. This is the single metric most predictive of chat ROI.
Of all visitors who initiate a chat conversation, what percentage provide a name and email address? This is your true engagement conversion metric. A well-designed conversation flow — friendly, helpful, not pushy — achieves 30–50% lead capture from initiated chats. Compare this to your contact form submission rate (typically 1–3% of all visitors) and the value of your chat strategy becomes immediately obvious.
A single-question rating at the end of every chat conversation: "How would you rate this experience?" Track CSAT separately for AI-only conversations and human-assisted conversations. This split reveals whether your AI training needs improvement or whether specific agent interactions are dragging scores down. Target a blended CSAT above 4.2 out of 5. Below 3.8, investigate immediately — dissatisfied chat visitors are unlikely to convert and likely to seek competitors.
Biz AI Last surfaces chat initiation rate, deflection rate, response time, lead capture, and CSAT automatically — so you always know exactly how your engagement is performing.
Explore the Analytics