What is live chat? How it works, metrics and staffing

What is live chat?

Live chat is a way for website and app visitors to message a business in real time, through a small chat window on the page. A visitor types a question on the pricing page; a support or sales agent (or an AI assistant) answers within seconds or minutes, while the visitor is still there to act on the answer.

It sits between email (asynchronous, slow) and phone (synchronous, expensive): conversations are fast but text-based, agents can handle several at once, and the visitor never leaves the page.

How live chat works

1. A script loads the widget. A small JavaScript snippet adds the chat bubble to the page.
2. A persistent connection opens when the visitor starts chatting, usually a WebSocket, so messages flow both ways without page reloads.
3. The server routes the conversation to an agent's inbox: to whoever is available, by team, language or page, or first to an AI assistant.
4. The agent sees context: which page the visitor is on, where they came from, earlier conversations.
5. Replies return in real time; attachments, typing indicators and read receipts are optional extras.
6. Outside office hours, the widget switches to an offline form or an AI answer.

Live chat compared with other channels

"Live chat vs chatbot" is a false choice: most teams let AI answer common questions and hand over to a person when it cannot.

What live chat is used for

• Pre-sales: plans, shipping, compatibility, "will this work for me?"
• Checkout help: payment failures, discount codes, delivery dates.
• Support: account access, how-to questions, bug reports.
• Onboarding: helping new users reach first value.
• Proactive invites: offering help on high-intent pages after a visitor lingers.

Metrics to track

A worked staffing example

Sample data: 600 chats a week, average handle time 8 minutes, agents handle 2 chats at once on average, and you plan for 75% occupancy (agents are not chatting every minute they are online).

• Chat time: 600 × 8 min = 4,800 min = 80 hours
• With 2 concurrent chats: 80 ÷ 2 = 40 agent hours
• At 75% occupancy: 40 ÷ 0.75 ≈ 53 staffed hours a week

That is more than one full-time person, but demand is not flat. If 40% of chats arrive between 10:00 and 14:00 on weekdays, those 20 hours a week need about 21 staffed hours on their own (600 × 40% × 8 ÷ 60 ÷ 2 ÷ 0.75 ≈ 21). Plan staff by hour from your chat heatmap; an AI assistant that resolves the routine 30% reduces every number above proportionally.

Best practices

• Show real availability. Do not show "online" with nobody there; switch to an offline form or AI.
• Answer fast, then fully. Acknowledge within a minute; a canned response library helps with both speed and consistency.
• Use context. Knowing the page and source avoids "how can I help you?" ping-pong.
• Do not interrupt. Proactive invites on the right pages, once per visitor.
• Make it accessible. The widget must work by keyboard, announce new messages to screen readers and meet WCAG contrast requirements.
• Protect data. Mask card and ID numbers in transcripts.

Live chat in TapCub

On a site that already loads the TapCub analytics script, one switch shows the chat bubble; no extra code. See enable live chat.

• Inbox with visitor context: region, device, landing page, referrer and a 2-day browsing trail from analytics, because chat and analytics share one visitor ID.
• Canned replies typed with /, internal notes and keyboard shortcuts.
• AI replies from your knowledge base in three modes (off, only when all agents are offline, AI first), with handoff to a person.
• Two-way translation, proactive invites, office hours with an offline form, and sensitive data masking.
• Reports: conversations, resolution rate, first response (AI replies excluded), satisfaction, agent workload, and chat vs no-chat conversion.

Sources

• W3C, Web Content Accessibility Guidelines (WCAG) 2.2.
• Fette, I. and Melnikov, A. (2011). RFC 6455: The WebSocket Protocol.