Home · Blog · Online chat and requests from the website: what is changing and how to check it
Online chat and requests from the website: what is changing and how to check it
Chat on the site does not “increase conversion by 30%” in itself - it changes the composition of applications: those who would not fill out the form, and those who came at night, appear. Let's look at what settings of the CHBT widget affect this - the "Leave Contacts" button, pre-chat form, proactive messages, offline mode - and how to check the result by "Contacts", CRM and "Analytics", and not by feelings. We will not repeat the installation and configuration of the widget here: it is described step by step in <a href="/en/blog/onlajn-chat-dlya-sajta-vidzhet">widget instructions</a>, here - only about what each setting does to the flow of applications and how to measure it.
Three sources of applications that were not there before the chat
The form on the website collects one category of visitors - those who have already decided to leave a contact. The chat adds three more, and it is from them that the increase is made up.
- Question before lead. “Do you do measurements in the Moscow region?”, “Is there a size 44?”, “Do you work with individual entrepreneurs?” The person is not ready to leave the phone until he receives an answer. The form loses it, the chat responds and then asks for a contact.
- Application without form. Some people don’t like questionnaires in principle: five fields, captcha, “I agree to be processed.” In chat, a contact is asked in one message when the conversation is already underway.
- Night and weekend. The form is rarely filled out at night - it is not clear that anyone will answer. A chat with an honest offline message and a working script collects a contact and promises a response in the morning.
The downside: those who simply ask “how much is it” and leave also come to the chat. These are not lost leads, these are conversations that previously did not exist at all. Therefore, the effect must be measured by leads and transactions, and not by the number of conversations - more on this below.
It's helpful to understand in advance how each of the three sources will show up in the numbers. “Question before request” gives an increase in conversations with a moderate increase in contacts - people ask, some leave their phone number. “Request without form” is visible as contacts with the tag widget-lead with a constant or slightly decreased number of leads through the form, the chat attracted those who had previously filled out the form, and added those who had not. “Night and weekend” are reflected in the time of contact creation: if before the chat leads came from 10 to 19, and after - at 23:40, this is a new channel. If after two weeks you don’t see any of the three pictures, the problem is not with the chat as an idea, but with its setup - analysis in the last section.
“Leave contacts” button: lead in two clicks
The “Leave contacts” button in the chat window is the shortest path to the lead: the visitor clicks, enters a phone number or e-mail, and the contact with the transaction is already in the account (what exactly happens when clicked and how to enable the button - step 3 widget instructions). Two things are important for conversion: the path is reduced to two actions, and it can be offered at the right moment - for example, after answering a question about the price, the bot writes: “If you want, we’ll send you a calculation - click “Leave contacts” below.” Each such lead receives a tag widget-lead, and it is from this that we will further calculate the growth.
If the contact needs to be asked strictly at a certain point in the conversation, and not at the request of the visitor, the script has a “Data Collection” block with a “Telephone” check - how this works is discussed in the article about collecting applications via bot. For measurement, there is no difference: both methods add contacts into one “Contacts”, and the “Transaction in CRM” action after “Data Collection” creates a card in the same CRM.
Pre-chat form: the only setting that can take away leads
The pre-chat form asks for a contact before the first message, and this is the only widget setting that can reduce the number of requests: some visitors will close the window without filling out the fields. Mandatory fields are justified only where a question without contacting the manager is meaningless; in other cases, they cut off exactly the source of leads for which the chat was set up - “question before lead”.
A reasonable compromise is one optional "Name" field. Calling by name increases trust, the barrier is almost zero, and the phone number will come through a button or through “Data Collection” in the script. If in doubt, run the first two weeks without pre-chat, and turn it on in the third or fourth week and compare the number of conversations: if there are noticeably fewer of them, the form cuts off real people, and not “just ask.”
Proactive messages: which rules increase applications
The "Proactive messages" in the widget settings allows the chat to speak first: rule = when (after N seconds, when scrolling, inactivity or leaving the site) + where (all pages or a specific URL) + action (bubble next to a button or open chat with text) + how often. How to assemble a rule and what ready-made ones are available - in widget instructions. Here we talk about which of these actually works for leads.
- Rule by URL on the pricing page. “Choose an option to suit your volume?” - the person is already comparing, the question is appropriate. The most reliable rule; a bubble is enough for it; there is no need to force open the chat.
- Leaving the site on the product card or service page. “Leave a request for this option? We’ll call you back in 15 minutes” - here it’s appropriate to open a chat: the person leaves anyway. It works if the promise is feasible, and only on the desktop.
- By time on a long page with a delay of 40–60 seconds: during this time the person managed to read and probably got stuck on the question. The same rule with a 5 second delay on all pages is annoying and reduces conversion - visitors close the window reflexively and then do not open it themselves.
Two rules that have the greatest impact on the result. The first is the frequency: you set it yourself, “once per visit” or “once every N days”; For the rule about leaving the site, set it to “once every 7 days”, otherwise the regular visitor will see it every day. The second is a bubble instead of opening a chat wherever the visitor is still reading: the bubble does not overlap the page, and its text will still become the first message of the bot when the person clicks. The text of the rule itself is not written into correspondence - only the visitor’s answer will appear in “Conversations”, so formulate the rule as a question that you want to answer.
Night orders: how to separate them in statistics
The widget’s offline mode (the “online” status goes out when everyone has left the office and the visitor sees your offline text) can be set up in a minute - step 5 instructions. For applications, something else is important: the script works at night as during the day - it responds to buttons, asks for a contact, creates a deal - and the offline text should say what to do now: “Leave your question and phone number with the button below - we’ll call you back in the morning.”
To prevent night requests from disappearing among daytime requests, add a “Condition” block to the script based on the time of day and attach a tag to requests after closing night action "Add tag". In two weeks, a filter for this tag in “Contacts” will show how many applications came in when no one was working. For many companies, this is the most noticeable increase from chat: not because chat is more convincing than form, but because before there was no night channel at all. If tag night empty, but the site is live, check the offline text: the promise “we will respond during business hours” without asking to leave a contact gives conversations, but not leads.
What chat won't do: limitations that are best known in advance
To avoid disappointment in the second week:
- The widget will not increase traffic to the site. If 30 people visit the page a day, there is nowhere to wait for dozens of applications: only a part of the visitors will open the chat and write to it.
- CHBT does not have a counter for widget impressions and openings as a separate metric. There are conversations, contacts and transactions - we count on them.
- There is no A/B test of the widget itself (color, welcome text). You can test script branches using the “Random branch” action - how to do this is discussed in the article about How does a bot affect conversions?.
- On the Free plan - 30 conversations per month per bot. A conversation is considered to be an interlocutor who has written at least one message in a calendar month; Opening a window without a message is not included in the limit. For a site with live traffic, the limit may expire long before the end of the month, and the bot will stop responding to new visitors according to the script. Basic for 650 ₽ removes the limit and the signature “CHBT” in the window - plan comparison.
- The chat itself does not answer complex questions. The script closes the typical one, the rest must reach the person through the “Operator” block - otherwise unanswered questions will become negative reviews.
What is considered an lead from a chat: conversation, contact, deal
The main source of self-deception when evaluating chat is the confusion of the three units of accounting. In CHBT they are in different sections and mean different things.
- Conversation — the visitor wrote at least one message. This is a demand for communication, not an lead: this includes “how much does it cost”, “are you open on Saturday” and tests of competitors. How many such people per month is shown by the conversation counter at the bottom of the sidebar of the account and in the “Payment” section - it is common for all channels. The “By Channels” block in “Analytics” counts more broadly: the “Widget on the site” line also includes those who opened the chat window but did not write anything.
- Contact — a person has a phone number or e-mail in the “Contacts” card. This is an lead in the usual sense: you can call back. From the widget, contacts come with a tag
widget-lead, from the script - through “Data Collection”. - Transaction — a card in CRM, which the manager has brought to at least the “In progress” stage. This is an lead recognized as live: a person answered the call and is discussing the terms.
A chat report is honest when it contains three numbers in a row: conversations - contacts - deals in progress. A ratio like 60 - 18 - 6 says more about the channel than any one number: in this example, 30% of those asking left a contact, and a third of the contacts reached the conversation. If managers complain that “garbage comes out of the chat,” look at the third number, not the first: garbage is conversations without contact, and they are worth nothing.
Test method: two weeks before, two weeks after
You can honestly compare “was/has become” only under the same conditions, so fix the base before turning on the chat.
- 1-2 weeks before chat. Count leads from the site through your current channel - form, calls, letters. Write down the number and source of traffic (advertising included or not, whether there were promotions).
- Turning on. Install the widget with minimal settings: a “Leave contacts” button, an offline message, one proactive bubble rule on the pricing page. Do not change anything on the site on the same day.
- Week 1-2 with chat. Count three numbers separately: requests through the form (didn’t they fall - sometimes the chat “pulls” them over), contacts with the tag
widget-lead, transactions in CRM that have reached the “In Progress” stage. - Comparison. Calculate the growth by the amount of applications “form + chat”, and the quality - by the share of transactions transferred from “New” to further. If the chat added 20 contacts, of which 2 went into work, and the form gave 10 applications with 5 in progress, there is an increase, but the chat brings in a different audience, and you need to work with it differently (answer faster, ask qualifying questions in the script).
Two weeks - minimum: weekly seasonality (Monday vs. Saturday) is smoothed out only on full weeks. If the traffic is less than 100 visitors per day, take it monthly.
Two typical errors of the method. The first is to turn on the chat simultaneously with a new advertising campaign or discount: the increase will be attributed to the chat, and the decline will also be attributed to the chat. The second is to count “after” from the first day: in the first two or three days, you and your colleagues will open the widget a dozen times, and these test conversations will be included in the statistics. Tag test pins test and exclude them from the count or start counting “after” from the fourth day.
Where to look at the numbers inside CHBT
Three sections of the cabinet answer three different questions.
| Question | Section | How to count |
|---|---|---|
| How many people left contact via the widget | “Contacts” | Filter by tag widget-lead; export to CSV if you need to compare with the previous period |
| How many of them became deals and at what stage did they get stuck? | «CRM» | Cards “Lead from the site - …” by columns “New”, “In progress”, “Paid”, “Rejection” |
| How many conversations are there and how long are they waiting for the operator? | "Analytics" | Block “By Channels” - conversations for 30 days, line “Widget on the site”; tile “Waiting for a response”; schedule of incoming messages and responses for 14 days |
What “Analyst” doesn’t have is counters for script blocks: you can’t see how many people have reached a specific question. Tags replace this: place the “Action” → “Add tag” block after the key steps (asked the price, left-phone, called the operator) and read contacts by tags. Five tags per script is enough to see where people are leaving.
If requests are already sent to an external CRM through the “HTTP request” action, add a source variable to the script - for example, source = widget - and count on the CRM side.
Separately about proactive messages: their effect is visible not directly, but through quick responses. If the rule contains the “Choose plan” and “Have a question” buttons, clicking on them goes to the chat as a regular message - put the “Keyword” trigger on these phrases and immediately after it “Add tag” proactive. After two weeks, the number of contacts with this tag will show how many applications the rule brought, not the overall growth.
No more applications: what to check
If the numbers haven't moved after two weeks, the reason is almost always in one of five places.
- Chat does not respond immediately. The widget opens, and the first message is from the person who will come “someday.” Set the script to the “Open widget” trigger with a greeting and 3-4 buttons so that the response is immediate. Template “Support” - ready-made base.
- The contact is asked at the entrance. Required pre-chat fields on a store or service landing page. Remove the requirement or disable pre-chat altogether.
- Proactive rule on all pages. The chat opens in 5 seconds everywhere - visitors close it without reading it. Leave one URL rule on the pricing page and switch the action to bubble.
- Operator response within 24 hours. People write, no one answers, tag
widget-leadaccumulates without transactions. Enable Web Push in “Dialogs” (button 🔔) - a notification will come even when the tab is closed. How to distribute what a bot answers and what a person answers - in the article about transmission of conversation to the operator. - Nobody looks at the second and third numbers. There are many conversations, few contacts - which means the script does not ask for contact. After answering a frequently asked question, add a message asking you to leave your phone number and make sure the “Leave Contacts” button is turned on. There are many contacts, no transactions - the problem is on the side of managers or in the speed of response, and not in the chat.
- Wrong question about chat at all. Sometimes the task is not “chat or not chat”, but “chat or form”. Analysis in a separate article: chat or feedback form.
You can start checking with the Free plan: one bot and 30 conversations are enough to see how many questions actually come from the site in two weeks. Connect chat to the site for free →
Install the widget in 5 minutes“Support” templatePlansIntegrations: webhook and HTTP