Interactive Chat Plugin: The Price Names the Behaviour

Most articles on this topic treat "interactive chat" as a single product. It is not. The label covers widgets that wait to be clicked, widgets that open themselves with a greeting, and widgets that answer from your content without routing to anyone. Each one bills differently, and each one produces a different conversation volume from the same traffic.

What "interactive" means in practice

The word is doing a lot of work in vendor copy. Three patterns show up, and they behave differently once the visitor types.

1. The widget that waits. A chat icon sits in the corner. The visitor has to find it and click it. The conversation only starts if the visitor asks for one. This is the lowest-volume pattern, and most "interactive" claims in vendor copy describe nothing more than this.
2. The widget that opens itself. A greeting, a popup, or a timed nudge slides in after a few seconds on the page. The conversation is initiated by the widget, not by the visitor. The open rate rises sharply. The useful-conversation rate usually does not, because most of those opens are people closing the nudge.
3. The widget that answers. The visitor types a question and receives an answer drawn from the site's content. No routing to a human happens unless the answer fails or the visitor asks for a person. This is what "interactive chat" tends to mean in product-led marketing, and it is the pattern that produces a useful conversation from the same traffic.

The price ladder of the product will tell you which behaviour the vendor is selling. A per-seat meter funds option one. A per-conversation meter funds option two, because every opened nudge counts. A per-message meter funds option three, because what matters is what the bot answered.

The five published price ladders

Each product below charges a different thing for the chat window in the corner. Read the meter column before the dollar figure, because the meter is what determines the bill.

The Asyntai row is ours, so read that as a disclosure. Asyntai charges by the message the visitor sends, not by the seat staffing the inbox or the conversation the widget opens.

Five meters appear in that table: workspace, conversation, branding, agent (or seat), and message. The same widget produces five different bills depending on which one you picked, and the meter tells you which behaviour the vendor is selling.

The behaviour each meter funds

A per-workspace meter (Crisp) is sold to a team. You pay for seats and profiles, which is the structure you use when a human reads every chat. The widget waits in the corner, and a person types back. This works for a team that has decided every conversation is a person-to-person conversation.

A per-conversation meter (Tidio) is sold to a site that wants the widget to open itself. Every time the widget initiates a chat, whether or not the visitor types anything, counts as a conversation. This is the meter that turns an open rate into revenue, and it is the meter that bills you for the nudge you showed and the visitor closed.

A per-branding meter (tawk.to) is sold as a free product with the vendor's name on it. The $19 figure removes the tawk.to label from the widget. This meter is unusual in that it has nothing to do with conversations or seats; it is a marketing fee paid to a vendor who can put their brand on your chat window.

A per-agent meter (LiveChat, Olark) is sold to a staffing plan. Every agent who logs in costs the price. Olark's higher tier at $400/mo adds an AI agent with the first seat included, which is a different shape from the others: the seat and the bot are bundled at that price.

A per-message meter (Asyntai) is sold to a bot. What counts is the message the visitor sends and the bot answers. A visitor who opens the widget and says nothing costs nothing. A visitor who asks a question costs one message. This is the meter that funds option three, the widget that answers.

One site, five monthly bills

Take a stated site: assume 2,000 chat messages a month, with 600 of them real conversations a human answers and the rest answered by the bot. Run the maths for each product at the tier that fits.

Crisp on Essentials is $95/mo for 10 seats and 100 profiles, which covers the human team. The bot answers are bundled into the workspace. The bill is $95/mo plus $10 for each extra agent past ten. For a 4-person team, that is $95/mo flat.

Tidio on Growth starts at $49.17/mo for 250 conversations and rises with conversation count. On 2,000 messages a month, the conversation count is what determines the tier, not the message count. Tidio prices its conversation tiers above this volume (a $300 tier and a $749 tier are published), so a site at this volume lands on Growth at the entry price.

tawk.to is $0 with branding, or $19/mo to remove it. The free product has no conversation cap and no seat fee. The bill is either nothing or $19/mo, and the difference is whether your visitors see "tawk.to" in the widget.

LiveChat on Team billed annually is $49/agent/mo. A 4-agent team is 4 × $49/mo. Billed monthly instead, the same Team tier is $59/agent/mo, which is the more common small-site arrangement.

Olark live chat is $29/agent/mo. The same 4-agent team is 4 × $29/mo. Olark's AI tier is a different product: $400/mo billed annually with the first seat included.

Asyntai on Starter is $39/mo for 2,500 messages, which fits this site exactly. A site that exceeds 2,500 messages moves to Standard at $139/mo for 15,000 messages.

Five products, five different bills, and the spread is wide. The cheapest tier for the stated site is tawk.to at $0 (or $19/mo branding-removed). The most expensive is Olark's AI Website Agent at $400/mo. The published prices are not on the same scale.

What changes the answer

Volume. A site at 200 messages a month has a very different bill from a site at 20,000. Per-seat products are flat with volume: you pay the same for one chat or one thousand. Per-conversation products scale with the number of times the widget opens. Per-message products scale with the number of times the visitor types.

Team size. Per-seat and per-agent products rise linearly with headcount. Per-message products do not, because the bot does most of the work. A 2-person team on LiveChat Team is 2 × $49 = $98/mo annually; a 10-person team is 10 × $49 = $490/mo. The same traffic on Asyntai Starter is $39/mo either way.

Page weight. A 30-second product page and a 30-minute reading session produce different conversation volumes. The longer the dwell, the more chance the widget opens itself, which is the behaviour that inflates a per-conversation bill without producing a single useful conversation.

Opening hours. A site that answers after-hours with a human pays for the after-hours seat. A site that answers after-hours with a bot does not, because the bot does not have a shift. The bill does not change when the office closes.

The mistake most people make

The mistake is reading "interactive chat" on a pricing page and assuming the widget waits in the corner for the visitor. The widget does not. Most products open themselves with a greeting after a few seconds, and on a per-conversation meter every opened nudge counts as a conversation. The open rate rises sharply. The useful-conversation rate stays flat, because most opened nudges are closed without a single message sent.

This is the behaviour that inflates a bill without producing a single useful conversation. The vendor's analytics show thousands of conversations. The team's actual inbox shows the same handful of real questions as the month before. The vendor is happy because they billed you for the opens. The team is unhappy because they paid for a queue that does not exist.

The fix is to read the meter. A per-conversation meter funds the widget that opens itself. A per-message meter funds the widget that answers. The same traffic produces a different bill and a different conversation on the two meters.

What to measure before you buy

Three numbers tell you which product fits and which one bills you for opens you did not want.

1. Messages per month. Count the chats a visitor actually types. This is the number that matters on a per-message meter.
2. Conversations per month. Count the chats the widget opens, including the ones the visitor closed. This is the number on a per-conversation meter.
3. Useful-conversation rate. Of those opened chats, how many produced a real answer. If the rate is under 20%, the widget is opening itself more than visitors are asking for help, and the meter is billing you for opens.

Pull those three numbers from your current site before you change products. If you do not have a chat widget, you can use Google Analytics events on a stub button to estimate. The point is to know what you are paying for before the bill arrives.

How to install a chat widget

All five products in this article install the same way: a script tag dropped into the site, or a code-injection block on platforms like Squarespace. The script loads the widget, the widget renders the button, and the button opens the chat window.

<script>
  (function() {
    var s = document.createElement('script');
    s.type = 'text/javascript';
    s.async = true;
    s.src = 'your-vendor-loader-url-here.js';
    var x = document.getElementsByTagName('script')[0];
    x.parentNode.insertBefore(s, x);
  })();
</script>

Two notes on that snippet. First, the loader URL is vendor-specific; replace it with the one from the product you pick. Second, the snippet is async, so the chat button does not block the page rendering. If the script is slow, the page is not slow.

What changes when the widget answers instead of routing

A widget that routes to a human produces a queue. A widget that answers produces a log. The queue is bounded by the team; the log is bounded by your content. On a per-seat meter, the queue is what you paid for; on a per-message meter, the log is what you paid for.

The two products behave differently when the visitor asks the same question twice. A human queue answers twice, the same way, and bills the seat twice. A bot answers twice, from the same source, and bills one message each time. Across 2,000 messages a month, the difference in bill is the difference between staffing and metering.

The two products also behave differently when the question has no answer on your site. A human queue escalates. A bot that cannot find an answer hands the visitor to the team, or routes straight to your inbox. Either way, the meter is unchanged, because the question was answered or handed off, not counted twice.

Picking the right meter

Three rules of thumb from the maths above:

• If your visitors rarely need to chat and your team answers everything personally, a per-seat or per-agent meter (LiveChat, Olark, Crisp) is the right shape. You pay for the people, not for the widget.
• If your visitors chat often and most questions are answerable from your site, a per-message meter (Asyntai) is the right shape. You pay for what the bot answered, and the team only handles the rest.
• If your site gets a lot of opens but few useful chats, a per-conversation meter (Tidio) will bill you for the opens. Read the meter before you read the price.

None of those rules of thumb override the published price. The published price is what you will pay. The meter is what the bill counts. Read both before you pick.