How to See Which Website Buttons Start Chatbot Conversations
Why you cannot tell which button works
A website usually has several ways into the chat: a header link, a product page button, a footer link and the bubble itself. When conversations arrive in one list, nobody can say which of them brings the visitors who leave their details and which only bring questions.
Without that, decisions about where to put a button are guesses. A button that sounds important may start few conversations, while a quiet footer link may bring the leads. The only way to know is to count by button.
AI Chatbot keeps the reason a button sent with every conversation it starts. The dashboard turns that into a short table, for every chatbot, so you can compare buttons with each other and with chats opened the normal way.
Why this happens
- every button opens the chat the same way, so conversations carry no trace of where they came from.
- buttons are added and moved without ever checking whether they start conversations.
- leads are counted for the whole chatbot, never per button.
- a button gets a greeting that no one has looked at since it was written.
- nobody compares the buttons with chats opened from the bubble.
Biznsbook addresses this through Opened from page buttons, conversations, leads and live chats per reason, the Opened from filter and CSV columns in AI Chatbot, an AI chatbot for your website or app that is sold on its own and needs no other Biznsbook app.
Unanswered website questions rarely show up in a report. They show up as visitors who left and bought elsewhere. Answering from your own content, and passing the visitor to a person when needed, closes that gap.
Step-by-step: See Which Website Buttons Start Chatbot Conversations
Written for owners, marketers and support teams in small and mid-sized businesses. The steps follow AI Chatbot in Biznsbook.
- Give each button a reason. Add data-cw-open and a data-cw-intent reason to each button or link, and an optional data-cw-source label for where it sits, such as header or footer. Different buttons with the same purpose can share a reason and differ by location.
- Let conversations come in. The report counts conversations where the visitor wrote something. A button that opens the chat but is never followed by a message does not appear.
- Open the chatbot Dashboard. Open AI Chatbot, choose the chatbot, and pick 7, 14, 30 or 90 days at the top. Scroll to Opened from page buttons.
- Read each row. A row is one reason at one button location. It shows the conversations started, the average number of messages (questions and answers), the leads and the share of conversations that became a lead, and the live chats visitors asked for.
- Compare with the normal way. The row Opened the normal way counts chats not opened from a button. Compare each button's lead share with it.
- Read the conversations behind a number. Press a reason to open Conversations filtered to it. You can also use the Opened from filter yourself, and the CSV export has Opened from and Button location columns.
- Fix what you find. A no rule label means no launch rule exists for that reason, so the chat opened with the normal welcome message. Add a rule in Settings, Launch rules, or rewrite a greeting that visitors do not answer.
Review results after the first full weekly cycle. Adjust roles, mappings, or approvals where the same exception repeats.
Inside Biznsbook, open AI Chatbot: the chatbot Dashboard shows a short checklist until the chatbot is live. This article explains the process behind those screens.
Common mistakes to avoid
- Mistake 1: Every button opens the chat the same way, so conversations carry no trace of where they came from. Repeating this each month usually shows up first in visitors who asked and left without an answer.
- Mistake 2: Buttons are added and moved without ever checking whether they start conversations. Repeating this each month usually shows up first in visitors who asked and left without an answer.
- Mistake 3: Leads are counted for the whole chatbot, never per button. Repeating this each month usually shows up first in visitors who asked and left without an answer.
- Mistake 4: A button gets a greeting that no one has looked at since it was written. Repeating this each month usually shows up first in visitors who asked and left without an answer.
- Mistake 5: Nobody compares the buttons with chats opened from the bubble. Repeating this each month usually shows up first in visitors who asked and left without an answer.
Each of these shows up in Conversations or Leads. Fix it in the chatbot's knowledge or settings, publish, and check the next week's conversations.
Best practices that hold up as you scale
- Give each button a reason — Add data-cw-open and a data-cw-intent reason to each button or link, and an optional data-cw-source label for where it sits, such as header or footer.
- Let conversations come in — The report counts conversations where the visitor wrote something.
- Open the chatbot Dashboard — Open AI Chatbot, choose the chatbot, and pick 7, 14, 30 or 90 days at the top.
- Read each row — A row is one reason at one button location.
- Compare with the normal way — The row Opened the normal way counts chats not opened from a button.
Give one person the job of reading new conversations, unanswered questions and leads every day, and publish the fixes the same week.
How Biznsbook supports this workflow
Opened from page buttons is part of Biznsbook AI Chatbot. It works from the knowledge you publish, so you decide what the chatbot says, and every conversation, lead and live chat stays linked in one place.
conversations, leads and live chats per reason is part of Biznsbook AI Chatbot. It works from the knowledge you publish, so you decide what the chatbot says, and every conversation, lead and live chat stays linked in one place.
the Opened from filter and CSV columns is part of Biznsbook AI Chatbot. It works from the knowledge you publish, so you decide what the chatbot says, and every conversation, lead and live chat stays linked in one place.
AI Chatbot runs on its own, on the same plans as every Biznsbook app: Starter $19 a month (1 chatbot), Pro $39 (2 chatbots, with CRM and Bookings) and Business $89 (5 chatbots, your own branding). Access is permission-based: AI Chatbot Manager covers everything in AI Chatbot, AI Chatbot Agent covers live chat, conversations and leads, and Company Admin keeps full access.
Suggested implementation timeline
- Day 1: Choose a plan, create the chatbot and set up Opened from page buttons.
- Week 1: Publish your main knowledge sources and test with the questions customers ask most.
- Week 2: Put the chatbot live on your website and read every conversation for the first days.
- Month 1: Add answers for the questions it could not answer, and tune intents, buttons and the lead form.
- Ongoing: A monthly review of your buttons.
Reading the table without over-reading it
Each number answers a different question. Conversations tell you how many visitors wrote after using the button. Average messages count questions and answers together, so they show how far a conversation went. Leads show how many left a way to reach them, and the share is that number divided by the conversations for the same row. Live chats show how many asked for a person.
Small numbers move a lot. A button with three conversations and one lead has a high share that tells you little, so wait for more conversations or a longer range before changing anything. Use the 30 or 90 day range for buttons that are used rarely.
Comparing buttons fairly
Compare buttons with the same purpose and visitors with the same intent. A footer link and a header link that both send get_quote are separate rows, so you can see which location brings more conversations and leads. A button for a different purpose, such as support, will behave differently because those visitors want something else.
The row Opened the normal way is your baseline. If a button brings fewer leads than visitors who simply opened the bubble, look at its greeting, its buttons and whether a launch rule exists for it.
What to change when a button underperforms
Start with the conversations. Press the reason and read how they begin and where they stop. If visitors do not answer the greeting, rewrite it so it answers what the button promised. If they ask for something the chatbot does not know, add it to Knowledge and publish.
If a row says no rule, add a launch rule. If the buttons under the greeting do not match what visitors do next, change them. Then check the same row again after a week or two. The report counts only what happened, so give a change time before you judge it.
Metrics to track monthly
- Conversations and AI answers this month, against your plan
- Questions it could not answer, and how many now have an answer
- New leads, and leads still to follow up
- Live chats accepted, and visitors asked for details instead
- Buttons visitors pressed, by button
Start with three metrics; trend direction matters more than a single point-in-time snapshot.
Contact forms and guesswork vs AI Chatbot
Compare a typical website without a chatbot, or with a generic one, with Biznsbook AI Chatbot and Opened from page buttons.
| Capability | Manual / Spreadsheet | Biznsbook |
|---|---|---|
| Where conversations came from | ❌ One mixed list | ✅ Opened from on each conversation, by reason and button location |
| Comparing buttons | ❌ Guesswork | ✅ One row per reason and location with conversations, leads and live chats |
| A baseline | ❌ None | ✅ The row Opened the normal way for chats opened from the bubble |
| Reading the chats | ❌ Search through everything | ✅ Press a reason to filter Conversations to it |
| Exporting | ❌ A list without the source | ✅ CSV with Opened from and Button location columns |
| Spotting missing setup | ❌ You find out by accident | ✅ A no rule label on any reason without a launch rule |
A monthly review of your buttons
Once a month, open Opened from page buttons for 30 days, compare each reason with the normal opens, read a few conversations behind the weakest rows, and decide whether to rewrite a greeting, add a rule or remove a button.
Write the routine down so whoever covers holidays knows who reads leads and conversations, and how quickly they reply.
Frequently asked questions
Does the report work for every company or only for Biznsbook?
It works for every chatbot. Each company sees the reasons and button locations of its own chatbot.
What counts as a conversation here?
A conversation where the visitor wrote something. A chat that was opened by a button but never written in does not appear.
What is a lead in this table?
A visitor who left an email, phone number or social profile in the conversation. The table shows the leads and the share of that reason's conversations that became a lead.
Which reason does a conversation count for?
The one it had most recently. If a visitor opens the chat from one button and later from another, it counts for the later reason.
How do I read the conversations behind a row?
Press the reason to open Conversations filtered by it, or use the Opened from filter. The CSV export also has Opened from and Button location columns.
What does the no rule label mean?
There is no launch rule for that reason, so the chat opened with the normal welcome message. You can add a rule under Settings, Launch rules.
How this differs by industry
Retail
A shop compares its "Help me choose" buttons on category pages with the chat bubble and sees which pages start conversations that end in a lead.
Wholesale & distribution
A distributor compares a "Request trade prices" button in the header with one on catalogue pages, reading the conversations behind each before moving either.
Manufacturing
A manufacturer compares "Ask an engineer" buttons on data sheet pages by reason and reads the live chats they bring.