Skip to content

Why 'engaged conversations only' changes chatbot economics

By Michael Berris · April 23, 2026 · 6 min read

Every chatbot platform has to solve the same problem: how do we price this?

The conventional answers are:

When we built NightMaven, we looked at all four and concluded none of them match how buyers actually experience value from a lead-capture chatbot. So we picked a fifth: we only count conversations that had two or more messages.

We call these "engaged conversations." A bot-open that never got a reply doesn't count. A visitor who opens the chat widget and closes it without typing doesn't count. A scraper that posts one spam message doesn't count. A real person who asks a real question and gets a real answer — that counts.

Here's why this matters.

The shape of chatbot traffic

A lot of chatbot traffic never turns into anything. Chat opens on a marketing site tend to fall into four groups:

Per-conversation pricing counts all four groups. You pay for widget opens that never became conversations. You pay for single-message tests.

Per-resolution pricing depends on how "resolution" is defined — usually it means "the bot answered something without needing human escalation." That includes a lot of single-message non-conversations where the bot responded to "hi" and the user left.

Neither aligns with how you, the buyer, actually measure value. You measure value in real conversations with real prospects that lead to real leads.

What "engaged" actually means

Our definition: a conversation with at least two messages from the visitor, or one message plus a qualification field collection. In practice, this means:

Below that threshold, we don't think you should be paying. And because we bill only on engaged conversations, our incentives are aligned with yours — we make money when your bot actually does its job.

What this looks like on a monthly bill

On NightMaven's Pro tier, you're paying $1,499/mo for 3,500 engaged conversations/month. That's 3,500 real back-and-forths. If your bot also handled 1,500 single-message opens and 4,500 widget opens that never became conversations — those don't count.

On per-resolution platforms, the same traffic can cost more, depending on whether the platform counts an interaction where the bot technically "responded" as a resolution. Ask for the exact definition.

On per-conversation platforms, you'd pay for every open.

At Starter ($49/mo), Growth ($499/mo), Pro ($1,499/mo), Business ($2,499/mo), and Business+ ($7,499/mo), the engaged-only definition means your billing metric aligns with your value metric. That's the whole point.

Why no one else does this

The usual reasons boil down to one of these:

  1. "It's how we've always billed." Per-resolution / per-conversation models predate the current generation of AI bots. They're inherited from the live-chat era.
  2. "Our infrastructure costs are per-call, not per-conversation." This is true — every widget open probably costs the platform a few fractions of a cent in compute and data transfer. But it's the platform's job to amortize that cost across paying customers, not to bill you per widget-open.
  3. "Customers would try to game the definition." This is fixable. We define "engaged" narrowly and consistently, publish the definition, and don't change it without notice.
  4. "Per-resolution sounds better in sales conversations because the unit cost is low." A small per-unit price sounds cheap. But when one ad campaign drives 10,000 chat opens, a small fee on each one adds up fast, even though few of them were real conversations.

What buyers should ask

When you're evaluating any chatbot platform, ask specifically:

  1. What counts as a billed unit? (resolution, conversation, user, token, call — get the exact definition)
  2. What doesn't count? (this is where vendors hedge — get a clear list)
  3. What happens on a traffic spike? (is your bill capped, or does it scale linearly with opens?)
  4. What happens with bot traffic or scrapers? (do those count against your quota?)
  5. Can I see the per-unit cost breakdown on my invoice? (if not, you can't audit)

If the answer to any of those is murky, you're probably going to be surprised by at least one month's bill per year.

Our side of the math

NightMaven's business works because we've decided that 3,500 engaged conversations/month on Pro is a volume we can serve sustainably for $1,499. Underneath, our costs are a mix of AI usage, data storage, servers, and the humans who build and support the product. We've sized the tiers so that they fit those costs with reasonable margin, and so that customers experience the bill they expected.

You don't have to take our word on whether the economics work. You can test them: your first 25 conversations are free on every tier. Once you've used them, you've either captured real engaged conversations (and add a card to keep going) or you haven't (and you walk away, never having entered a card). That's the cleanest possible alignment of incentive.

The bottom line

The billing model is part of the product. A platform that bills you for events you don't value is quietly misaligned with you forever. A platform that bills only for events where you and they both did something worth doing is aligned from day one.

We picked alignment. It's a smaller number of billable events per month, but it's the right number.

Pricing you can predict, billing that matches your value.

25 free conversations, no card needed. Only engaged conversations count, on every tier.