How to measure whether website chat is paying for itself
Website chat pays for itself when the sales and bookings that started in chat are worth more than the subscription plus the staff time spent answering. You can check that every month with five numbers: chats received, time to first reply, chats that turned into a sale or booking, cost per chat, and questions you could not answer. None of them needs a data team; a sheet and twenty minutes at the end of the month are enough.
The five numbers
| # | Number | How to get it | What it tells you |
|---|---|---|---|
| 1 | Chats received | Count new conversations in the month | Whether people use the chat at all |
| 2 | Time to first useful reply | Sample 20 chats, note the wait | Whether you answer while they are still deciding |
| 3 | Chats won | Bookings, orders or sales that started in a chat | Whether chat brings money |
| 4 | Cost per chat | (Subscription + staff time) ÷ chats | What each conversation costs you |
| 5 | Unanswered questions | Questions with no ready answer | Where you are losing people |
1. Chats received
Count new website conversations for the month. Compare with your website's visitor numbers. A very low number usually means the widget is on the wrong pages or the welcome message gives no reason to write, not that nobody has questions.
2. Time to first useful reply
Open twenty conversations spread across the month and note how long each customer waited for the first reply that actually helped. Write down the median, not the average: one chat that waited all weekend should not hide nineteen fast ones. How fast to answer website chat suggests targets and a rota.
3. Chats won
This is the number that pays the bill, and the one most teams skip. The simplest honest method: when a chat ends in a booking, an order or a sale, the person who handled it tags the conversation "won". At the end of the month, count the tags and, if you can, add up their value. Asking for contact details well raises this number; see collecting a lead without being pushy.
Be strict about what counts. A customer who says "I paid, here is the screenshot" has not paid until your bank or payment system shows it. StayReplied treats it the same way: a payment screenshot moves a lead to "awaiting verification", not "won", and confirmation comes from your payment system or a person.
4. Cost per chat
Add your monthly subscription and an estimate of the staff time spent on chat, then divide by chats received.
For the subscription part, StayReplied's regular prices on the pricing page are AED 70 a month for Starter (2 seats), AED 210 for Team (5 seats) and AED 420 for Business (10 seats), plus 5% VAT shown before you pay. AI is optional and no plan includes it; if you add your own Anthropic key, Anthropic bills you directly for that usage, so add that bill too. Staff time is usually the larger part: hours on chat per week, times four, times the hourly cost of the person answering.
Compare cost per chat with the value of one won chat. If one booking is worth more than ten chats' cost and at least one chat in ten ends in a booking, chat is paying.
5. Unanswered questions
Count the questions where your team had no ready answer or had to ask someone else. Each one costs time now and customers later. If you use AI, StayReplied keeps a list of questions the assistant could not answer, with suggested facts a person can approve into the knowledge base; if you answer by hand, keep a simple list as you go and turn it into the 20 answers customers ask.
A monthly review sheet
- Chats received this month, and last month
- Median time to first useful reply (20-chat sample)
- Chats tagged won, and their total value
- Subscription + AI provider bill + staff hours × hourly cost
- Cost per chat, and value per won chat
- Top five unanswered questions, each with a person who will write the answer
Trends over three months tell you more than any single month.
Common mistakes
- Counting every chat as a lead. "Where are you?" from a regular customer is service, not a new sale. Both matter, but only one belongs in number 3.
- Measuring only speed. A fast reply that gets the price wrong costs more than a slow, correct one; a 30-minute quality review catches those.
- Forgetting staff time. A cheap tool answered slowly by an expensive manager is not cheap.
- Changing everything at once. Change one thing a month (the welcome message, the pages the widget sits on, the shift rota) so you can tell what worked.
Where the inbox helps
In the StayReplied unified inbox every website conversation is in one searchable list, so sampling twenty chats takes minutes. Tags are included on every plan, and so is a sales pipeline for enquiries that take more than one conversation to close. When number 5 points at a missing answer, write it as a knowledge document, review it and publish it; the docs describe how versions and rollback work.
When this is not for you
If your business gets a few chats a month, these numbers will be too small to mean much; read every chat instead and ask whether any of them became a customer. And if you cannot say what one booking or order is worth, start there: without that, no tool can tell you whether chat pays. Who StayReplied is for describes the teams it fits.
Questions and answers
How do I know if live chat is worth it?
Compare the value of bookings and sales that started in chat with the subscription plus the staff time spent answering. Track it monthly with five numbers.
What is a good way to count sales from chat?
Have the person who handled the chat tag it "won" when it ends in a confirmed booking, order or sale, and count the tags at month end.
Should a payment screenshot count as a sale?
No. Count it when your bank or payment system confirms it. StayReplied marks such a lead "awaiting verification", not "won".
What does StayReplied cost?
Regular prices are AED 70, AED 210 and AED 420 a month for Starter, Team and Business, plus 5% VAT. Your own AI provider bills you separately if you use AI.
Measure it on your own site for two weeks.
start your free trial, 14 days, no card needed.
