A small-business owner working out support cost per customer with a calculator at a desk

Support cost per customer is one of those figures nearly every small business skips, right up until a month where the inbox eats the week. The formula is easy. Everything support costs you, divided by your customers. The useful part is deciding what goes on the top and what goes on the bottom, and almost every guide on the subject answers both questions for a company with a support department. Here is the version for a business with one or two people in it.

The short version

What goes into support cost per customer

Keep the list short. A big company allocates office space and a slice of the HR budget into this calculation. You do not need to, and the precision you would gain is smaller than the error in your own time estimate.

Four lines cover it:

That third bullet about fully loaded cost catches people out. Wages are not what an employee costs. According to federal compensation data, benefits made up 30.1 percent of what private employers paid per hour worked in March 2026. So somebody on a $22 wage costs you closer to $31 an hour once payroll taxes and benefits land. Use the bigger number.

Leave out anything you would pay for anyway. Your website hosting is not a support cost. Your CRM probably is not either, unless you bought it because of support.

Cost to serve a customer: which denominator to use

This is where the calculation earns its keep. Divide by every customer and you get a soothing, nearly useless figure, because most of your customers never contact you in a given month. Divide by the customers who did get in touch and you get the real cost to serve a customer who needs you.

Both are a support cost per customer. They just answer different questions, so work out both. The gap between them tells you how self-sufficient your product is. A wide gap means most people manage on their own and a small group does not. A narrow gap means everyone needs help, which is usually a product or onboarding problem rather than a support problem.

Use a month you can remember clearly. Do not average a year on your first attempt, because you will spend an hour reconstructing numbers and lose interest before you get to the useful part.

A worked example for a two person shop

Say you run a small business with one employee. Neither of you does support full time, you both just answer customers as part of the day.

A month is roughly 4.33 weeks, so your hours come to 21.65 at $60, which is $1,299. Your employee lands at 34.64 hours at $31, which is $1,074. Add the $39 tool and the $150 in refunds and the month cost you about $2,562.

You have 400 active customers, so the headline figure is $6.40 each. Comfortable. But only 70 of them contacted you that month, which puts the real support cost per customer at $36.60 for anyone who needed you.

The average is the number you report. The spread is the number you act on.

Now the part that changes decisions. Think back over those 70 and you will usually find about a dozen that took half the total time. If twelve customers consumed half of $2,562, they cost roughly $107 each that month, while the remaining 58 averaged about $22.

If your average customer pays you $40 a month, you have just found twelve relationships that are underwater, using a calculator and thirty days of inbox history.

How to find your expensive customers without analytics

You do not need reporting software for the first pass. You need one honest hour.

  1. Pick a recent month and sort your support inbox by sender.
  2. Write down every customer who started more than two threads.
  3. Next to each one, estimate minutes rather than counting messages. A single thread that ran fourteen replies costs far more than four quick ones.
  4. Add the minutes, multiply by your hourly figure, and you have a per customer cost for the top of the list.

Minutes, not tickets, is the whole trick. This is where cost per ticket and cost per customer part company. Ticket counts flatter the customer who sends one enormous thread a week and punish the one who asks three quick questions and gets on with their day.

Three shapes turn up again and again. There is the customer asking the same question in different words, which means your answer never got written down. There is the customer with a long thread every time, which usually means the first reply is not landing. And there is the customer who bought the wrong thing, who needs a different plan or a refund rather than more patience.

What to do about a customer who costs too much

Work through the fixes cheapest first. Most businesses skip to the last one and regret it.

Write the answer down once. If four customers asked the same thing, that is a help article you have not written. Put it where they can find it before they email. This is the single biggest lever a small team has, and we have written a full guide on how to cut repeat questions if you want the detail.

Fix the thing causing the question. Some tickets are a symptom. A confusing checkout step, a part that ships without instructions, a setting buried three menus deep. One afternoon on the cause beats a year of answering the effect.

Then, and only then, talk about the relationship. Sometimes the honest answer is a higher plan that includes more of your time, and sometimes it is a polite exit. Both are fine. Just make sure you have tried the other two first, because a customer who costs $107 a month usually costs that much because of something you have not fixed.

When a high number should not change anything

A high support cost per customer is not automatically a problem, and three cases are worth protecting.

New customers cost more, and they are supposed to. The first ninety days carry setup questions that never come back. If your expensive list is mostly people who joined last month, you are looking at onboarding, not waste.

Some customers are worth the hours. A customer who costs $107 a month in support and stays five years on a good plan is a fine customer. Check the figure against their customer lifetime value before you act on it, not against one month of revenue.

And if you have twenty customers, this number is noise. One bad week moves it thirty percent. Track it quarterly at that size and use it as a direction, not a decision.

The line you can control most easily

Labour is the only line that really moves, so anything that removes a repeated question is worth more than anything that shaves your software bill. The software bill is still worth getting right, though, mainly so it stops being a variable you have to think about.

That is the reasoning behind our own flat monthly pricing. Starter is $15 a month for one seat, Growth is $39 for up to three, and Enterprise is $79 for up to ten. Every feature is on every plan, including the shared inbox, the knowledge base, the live chat widget and the support bot, so the tools line in your calculation is a fixed number you can write down in advance. There are no per resolution fees, which matters here, because a pricing model that charges by the answer bills you most for exactly the expensive customers you are trying to understand.

Once you know which dozen customers cost you the most, the work is ordinary. Write the four answers you keep retyping, put them where people can find them, and watch that group get cheaper next month. If you want somewhere to put those answers today, you can start a free trial and have the first few articles live this afternoon.

Ready to make customer support simple?

Start a free 14-day trial of SupportifyGPT. Every feature included, no setup fee, live in minutes.

Start your free trial