Website Chat Funnel: Loads, Opens, Conversations and Messages, What a Normal Rate Looks Like
Four events tell you whether your website chat works: widget loaded, opened, conversation started and message sent. How to read each drop in the funnel.
In this article
Most teams judge their website chat by one number: how many chats came in this month. That number hides the real story. A website chat funnel has four stages before a single agent types a reply, and each stage leaks visitors for a different reason. If nobody opens the widget, the problem is the bubble. If people open it and never write, the problem is what they see inside. If they write once and vanish, the problem is your reply time.
This post is about reading that funnel, from widget loaded to message sent, and then downstream in the CRM when a chat becomes a contact and a deal. It avoids "industry average" figures, because a B2B pricing page in Austin and a holiday-let site in Cornwall will never behave the same way. The useful benchmark is you, last week.
The four events every website chat should record
Omnifox records four events for every widget, saves them as a funnel in the dashboard, and pushes them to your site's dataLayer and gtag if they exist:
| Event | What it means | Stage of the funnel |
|---|---|---|
widget_loaded |
The script ran and the chat button is on screen | Reach |
widget_opened |
The visitor clicked the button or a bubble opened the panel | Interest |
conversation_started |
A conversation record now exists in the shared inbox | Commitment |
message_sent |
The visitor typed and sent at least one message | Engagement |
Loads measure reach, not intent. Opens mean the button or a bubble caught attention. A conversation only starts once someone gets past the pre-chat form or the department selector, and a message means they actually said something. The gap between each pair of numbers is a diagnosis waiting to be read.
How to read each drop between stages
Loads to opens: a visibility problem
If loads are high and opens are tiny, visitors are not noticing or not caring about the chat. Common causes:
- The button sits where your cookie banner or a sticky "Book now" bar covers it on mobile.
- No proactive bubble on the pages where questions actually happen (pricing, checkout, product detail).
- The widget loads on every page, including the blog, where nobody has a buying question. That inflates loads and drags the open rate down without anything being wrong.
Filter the funnel by page before you panic. A low open rate across the whole site can hide a healthy one on the two pages that matter.
Opens to conversations: a friction problem
People clicked, then left before a conversation existed. Almost always this is what greeted them:
- A pre-chat form asking for name, email, phone, company and "how did you hear about us" before they can type a word.
- A mandatory department selector with five options that mean nothing to an outsider.
- A generic welcome message ("Hi! How can we help?") that gives no reason to write.
Cut the form to what you truly need at that moment, make departments optional or reduce them to two or three, and write a welcome line that names the page: "Comparing plans? Ask me which one fits a team of your size."
Conversations to messages: a hesitation problem
A conversation exists but no message came: the visitor got through the form and stalled at an empty text box. Quick-reply buttons from an automation ("Pricing", "Delivery times", "Talk to a person") or an AI contextual greeting that reads the page and opens with a relevant sentence remove the blank-page problem.
Message sent, then silence: a speed problem
The visitor wrote once and never again. Check the reply time on those conversations. If the first human reply came after they had closed the tab, the fix is routing (first available or least load), an AI agent set to answer the first message, or honest business hours with an after-hours name-and-email form so the lead still lands in the inbox.
A diagnostic table you can work through
| Symptom in the funnel | Most likely cause | First thing to try |
|---|---|---|
| Many loads, very few opens | Button hidden, no bubble, wrong pages | Add one proactive trigger on the pricing or product page, 20 seconds on page, once per session |
| Opens fine, conversations low | Long pre-chat form or forced department selector | Make the form skippable, cut fields to name + email |
| Conversations fine, messages low | Empty box, generic greeting | Quick-reply buttons or AI contextual greeting |
| One message, then nothing | Slow first reply | AI agent for the first message, or least-load routing |
| Good on desktop, poor on mobile | Button overlapped by other sticky elements | Change position or button size; add a device-only trigger for mobile |
| Good everywhere except one landing page | Traffic quality from that campaign | Check the UTM on those conversations before changing the widget |
Work through it one page and one week at a time. Changing three things at once tells you nothing.
What "normal" looks like without inventing numbers
Blog posts quoting precise open and engagement rates come from different industries, page types and traffic sources. What you can rely on:
- Your own trend. Compare this week's open rate to last week's on the same page. A move after a single change is a signal; a move after redesigning everything is noise.
- Page-level ratios. Checkout and pricing pages should convert opens to messages far better than the homepage. If they don't, that page's greeting is the problem.
- Downstream ratios. Of the conversations that reached the inbox, how many became a contact with an email or phone, and how many became a deal? That is the number your finance team cares about.
- Reply time on first message. Log it weekly. Every stage above it improves when this improves.
For what to watch beyond the widget, the post on live chat metrics that actually matter covers response time, CSAT and resolution.
From conversation to contact to deal
The widget funnel ends at message_sent, but the business funnel keeps going. In Omnifox, every website chat conversation stores the origin page, the referrer and the UTM parameters, and that data lands on the contact record in the CRM. That lets you ask which campaign produced chats that became deals, not just chats, and which landing page generates conversations that never leave a phone or email.
A weekly review looks like this, straight from the dashboard:
Week 37 · pricing page
widget_loaded 1,240
widget_opened 186
conversation_started 74
message_sent 61
→ contacts with email 38
→ deals created 12
Those are illustrative counts to show the shape, not a target. The chain shows the biggest leak is between loads and opens, and that once someone writes, the team converts well. That tells you where to spend the next hour.
Cross-checking with GA4 and UTM attribution
Because the same four events reach gtag automatically, you can mark conversation_started as a key event in GA4 and see it next to your other conversions by source and medium. GA4 confirms which campaigns drive chats at the top; the CRM shows which of those chats turned into revenue at the bottom. The setup is in GA4 and your website chat: tracking opens, conversations and conversions with dataLayer. One warning: if you built your own Google Tag Manager tags for the same events, you will double-count. Use the automatic events or your own, not both.
Frequently asked questions
What is a good open rate for website chat?
There is no universal figure; the open rate depends on the page, the traffic source and whether a proactive bubble is running. Measure the same page week over week and treat a consistent upward trend as success.
Why does my chat show many conversations but few messages?
A conversation is created as soon as the visitor gets past the pre-chat form or department selector, so a large gap to message_sent means people stall at the empty text box. Add quick-reply buttons or an AI contextual greeting that gives them something to respond to.
Should I count AI-answered chats in the funnel?
Yes. The four widget events fire regardless of who replies, and the AI agent's conversations sit in the same shared inbox. Track separately how many ended in a human handoff so you can tune the knowledge base.
Can I see the funnel per page or per campaign?
Yes. Each conversation stores the origin page, referrer and UTM parameters, so you can filter the funnel by landing page or campaign and compare them side by side.
Start measuring before you start tweaking
Install the widget, leave it running for a full week with no changes, and only then read the funnel. Every stage has a specific fix, and the leak is rarely where you assumed. Omnifox includes website chat in every plan with the four-event funnel, UTM attribution and CRM continuity built in. Start a 7-day free trial and compare your own numbers next week against this one.
Ready to try it in your business?
Bring WhatsApp, Instagram, Messenger, email and calls into one inbox, and let AI handle the first reply. Set it up today.
7 days free. No credit card required.
Read more
-
The Metrics Dashboard Every Business Messaging Program Needs
-

What Is Click-to-Call and How to Implement It in Your Business
-
Connecting Your CRM to a Shared Inbox: What Good Integration Looks Like
-

AI to Translate Conversations in Real Time
-
RCS, iMessage and the US Mobile Messaging Landscape
-

WhatsApp messaging limits and how to increase them
-
Sales Automation for Canadian Sales Teams: A Practical Guide
-

First Contact Resolution (FCR): A Complete Guide
Comments (0)
No comments yet. Be the first to share your thoughts.
Leave a comment
Your email is never published. Comments are moderated before appearing.