You know what a customer costs to acquire. Do you know what they cost to talk to? Most small businesses track sales closely and never put a number on support, which is odd, because that number quietly decides whether growth makes you money or just makes you busy. Cost per conversation is how you find the figure.
It's a simple metric that carries a lot of weight. Once you can see what each support interaction costs, decisions about hiring, tooling, and automation stop being guesses.
What cost per conversation actually means
Cost per conversation is exactly what it sounds like: the total cost of running support over a period, divided by the number of conversations you handled in that period. If support cost you $4,000 last month and you handled 1,000 conversations, each one cost you $4.
The word "conversation" matters. A conversation isn't a single message; it's a whole back-and-forth to resolve one customer's need, which might span several questions. That's why the metric fits modern support better than counting individual messages, where a chatty resolution looks artificially expensive.
The number is useful because it's a rate, not a total. Your total support bill going up isn't automatically bad; if you handled twice the volume, cost per conversation may have held steady or fallen. The rate tells you whether your support is getting more or less efficient as you grow.
How to calculate it
You don't need a fancy dashboard to start. Pick a month and work through this:
- Add up your support costs for the period. Include the obvious things (salaries or hourly pay for anyone doing support) and the easy-to-forget ones (help desk software, chat tools, and a share of overhead if you want to be thorough).
- Count the conversations handled in the same period. Pull this from your inbox, help desk, or chat tool. Be consistent about what counts as one conversation.
- Divide costs by conversations. That's your cost per conversation.
- Write it down and repeat monthly. One number means little. A trend line means everything.
Keep the definition stable. If you include software costs one month and drop them the next, the trend becomes noise. Rough and consistent beats precise and erratic.
Why it beats counting tickets
You'll also hear "cost per ticket," and the two are close cousins. The difference is that "ticket" comes from an email-and-queue world where each message spawns a case. "Conversation" fits live chat and messaging, where one flowing session might resolve three questions that would have been three separate tickets over email.
Neither is holy. Use whichever matches how your customers actually reach you, and track it the same way every month. If most of your support happens in chat, cost per conversation will describe reality more honestly. What matters is picking one and staying with it, so the trend stays comparable to itself.
What a chatbot does to the number
Here's where the metric gets interesting. A human conversation has a fairly fixed cost, because it takes a person's time. An automated conversation has a near-zero marginal cost, because software handles it whether it's the first that day or the thousandth.
So when a chatbot resolves a share of your conversations on its own, your blended cost per conversation drops. The human conversations still cost what they cost, but they're now a smaller slice of a larger total, and the automated slice barely adds to the bill.
There's a catch worth stating plainly. This only works if the automated conversations actually resolve the issue. A bot that "handles" a conversation by frustrating the customer into giving up hasn't lowered your cost. It's deferred the cost into a lost sale or a bad review, which never shows up in this metric. A falling cost per conversation is only good news if satisfaction holds steady beside it.
A worked example
Take Verano Goods, a fictional online homeware store. Here's a simplified before-and-after once they added a chatbot to handle order-status and shipping questions:
| Before | After | |
|---|---|---|
| Monthly support cost | $5,000 | $5,300 |
| Conversations handled | 1,250 | 3,000 |
| Handled by bot | 0 | 1,900 |
| Handled by humans | 1,250 | 1,100 |
| Cost per conversation | $4.00 | $1.77 |
Look closely, because the total cost actually went up by $300 (the chatbot tool plus setup time). Yet cost per conversation more than halved, because the bot absorbed a wave of routine questions at almost no marginal cost, and the humans handled slightly fewer, higher-value ones. Verano didn't spend less overall. They got far more support out of nearly the same budget, which is usually the real goal for a growing business.
Reading the metric honestly
A single number can mislead if you stare at it alone. A few habits keep it honest:
- Pair it with satisfaction. Track cost per conversation and CSAT together. Falling cost with steady or rising satisfaction is a genuine win. Falling cost with falling satisfaction is a warning.
- Don't chase it to zero. There's a floor. Some conversations should be expensive, because a human spending real time on a high-stakes issue is exactly what you want.
- Segment when you can. Automated versus human, or sales versus support. A blended average can hide a problem in one bucket.
- Mind seasonality. A holiday spike changes both cost and volume. Compare like months, not a quiet March against a frantic December.
Turning the number into a decision
A metric earns its place only when it changes a decision. Cost per conversation is good at informing a few specific ones.
The first is hiring. If your cost per conversation is creeping up as you grow, it's a sign your current setup is straining, and it's usually cheaper to fix the process, often with automation, before you add headcount. The number tells you whether the next hire is genuinely needed or whether a chunk of that volume could be handled without one.
The second is where to automate. Break the metric down by conversation type if you can. If order-status questions are eating expensive human time, that's a flashing sign to hand them to a bot. The most repetitive, lowest-judgment conversations are both the priciest in aggregate and the easiest to automate, which makes them the obvious first target.
The third is pricing and margins. Once you know what a customer costs to support, you can spot the accounts or plans that quietly lose money on support alone, and price or structure them accordingly. That's a conversation most small businesses never have, because they never had the number.
Used well, cost per conversation turns a vague worry ("support feels expensive") into something you can act on. It tells you when it's time to automate, whether that automation is paying off, and how much room you have to grow before support becomes the thing holding you back. Start by calculating this month's number. Next month's will tell you a lot more.