What ecommerce support actually costs at $5M, $15M, and $50M

Ask an owner what customer support costs and you’ll usually get one number: an hourly wage, or a monthly invoice from whoever they’ve outsourced to. It’s the number in their head because it’s the number on the bill they actually see. It’s also wrong, often by a wide margin — not because anyone’s being careless, but because the rest of the cost is scattered across people, tools, and hours that nobody labels “support.”

None of the following shows up next to the wage line, and all of it is real:

• A founder catching the escalations nobody else on the team can handle.
• A manager who used to close tickets and now spends real time scheduling instead.
• Refunds granted because the answer arrived a day late.

The three figures in this piece — $5M, $15M, and $50M — aren’t measured revenue brackets. They’re markers for where the structure of a support operation tends to change, drawn from how these operations actually behave at each size. Your business might hit the second one at $11M or $19M. The point isn’t the number on the sign; it’s recognizing which one you just passed.

Building the real bill

Start with the wage, because it’s the only line most owners have already priced. Then add what actually sits on top of it, in the order you’d notice them rather than the order a spreadsheet would:

• Employer load. Payroll tax, benefits, insurance — whatever your structure adds on top of the wage itself. It’s real money and it’s not small, but the exact rate depends on your entity, your state, and decisions your bookkeeper or accountant has already made for you. This piece isn’t the place to tell you what that number is — that’s a conversation with whoever handles your payroll, not a blog post.
• The person who stopped answering tickets. Once you have more than a couple of people on support, someone has to schedule them, review their work, and take the calls that are too hard to hand to anyone else. That person’s time used to close tickets. Now it manages the people who do. Their full cost belongs to support, even though nothing about their job title says so.
• Quality checking. Somebody has to read a sample of what went out and decide whether it was actually good, not just fast. Skip this and you won’t notice the decline until a customer tells you about it in a one-star review.
• Recruiting and ramp. The job posting, the interviews, and — the part that’s easy to underweight — the weeks after someone starts, when they’re slower and make more mistakes than the person they replaced. That period isn’t free. It’s paid for in slower first responses and a few tickets that go sideways while someone new finds their footing.
• Coverage you don’t have. The order placed at 11pm, the account locked over a weekend, the shipping problem that surfaces on a holiday. If nobody’s there, that’s not a zero-cost gap — it’s a cost you’re paying in a different currency, which is the next line.
• Tooling that scales with headcount. Pricing models vary by help desk — some scale with seats, some with ticket volume — but either way, the bill moves as the operation grows. It’s not the flat fee you priced once and forgot.
• Your own hours. The founder or ops lead who gets pulled in when a ticket is genuinely hard — an angry customer, a policy exception, a decision nobody else is authorized to make. That time is worth something. Usually it’s worth more than anyone else’s on the team, and it’s the line most owners forget to count at all.

Add those up against whatever number you started with, using your own headcount and your own tools, and the gap between “what I thought this cost” and “what it actually costs” tends to be the single most useful number in this whole exercise.

The other side of the ledger

None of that is the whole bill, either. There’s a second column: what it costs when support doesn’t resolve something well or fast enough.

• A refund issued not because the claim was good but because the answer took three days and the customer stopped believing you’d get there.
• A discount offered to make an escalation go away.
• A customer who doesn’t say anything, doesn’t complain, and just doesn’t order again. This is the most expensive line and the hardest to see, because it never generates a ticket at all — it just shows up later as a reorder rate that’s a little lower than it should be.

We’ve written before about why support that nobody owns ends up costing more than support that does — the short version is that this leakage is a symptom of missing ownership, not of any single agent underperforming. Why Customer Service Without Ownership Costs More walks through that mechanism in more detail. This piece is about putting an actual line on it: whatever you estimate for refunds, discounts, and quiet churn tied to slow or poor resolutions, that number belongs in the total, next to the wage and everything itemized above it. Once you’ve added this second column to the first, divide the combined total by however many tickets you handle in a month, and you have your real cost-per-ticket number — the one “The staircase” uses next.

The staircase

Here’s the part that catches owners off guard: cost per ticket doesn’t fall smoothly as you grow. It steps.

• Around $5M, support is still small enough that the owner is answering tickets alongside whoever else is on it, and the cost per ticket looks great on paper. It looks great because the most expensive person in the business — the owner — is working the queue for free, and that time never gets counted against the support line.
• Around $15M, the first real break hits. Volume outgrows what one shift can handle, and once there are enough people on tickets that no single person can track all of it by feel, somebody has to stop closing tickets and start running the people who do — scheduling, quality, the hard calls, hiring the next person. That role is a genuine cost, and it produces zero tickets. This is where cost per ticket goes up, and it’s where a lot of businesses reading this are standing right now. The instinct is to read a rising number as a staffing shortage and add another agent. That usually makes it worse, not better — you’ve added headcount to a system nobody’s running, which is closer to what What Makes a Great Customer Service Hire in eCommerce? is really about: the environment the hire lands in matters more than the hire.
• Around $50M, it’s an operation. Coverage is a decision with a price tag attached rather than a default. QA is a function, not a favor someone does on the side. Recruiting is continuous, because turnover at that headcount is a standing line, not an occasional event. Cost per ticket can come back down here — but only if someone actually owns the system. Without that, you’ve bought the cost structure of a real operation without the economics of one.

Why the curve bends this way

At each of those steps, you stop buying hours and start buying a system: standards, supervision, quality, someone accountable for the outcome instead of just the queue. Owners who keep pricing support at the wage keep buying hours, one at a time, and keep being surprised when the number doesn’t behave the way they expected.

The staircase isn’t a flaw in how support scales. It’s what it looks like when a person stops being enough and a system hasn’t been built yet. A single agent, however good, has a ceiling on how much volume they can carry well — and once you’re past it, adding a second, third, or fourth agent doesn’t raise that ceiling. It just adds more people who also need scheduling, review, and someone to answer the questions they can’t. Nothing about the arithmetic changes until somebody is explicitly responsible for the operation as a whole, not just for their own queue.

What to do with this

The useful next step isn’t a pitch — it’s arithmetic. Build your own fully-loaded monthly number:

• Wage, load, supervision, QA, recruiting and ramp
• The coverage gaps and the tooling
• Your own hours
• The refunds and quiet churn on the other side

Divide it by your ticket volume and you have your real cost per ticket. Then ask which step on that staircase you’re standing on, and whether anyone actually owns the system, or whether you’re the one still holding it together.

That’s a harder question to answer alone than it sounds, mostly because it’s hard to see your own operation from inside it. If you want an outside read on which step you’re on and what’s actually driving the number — not a sales pitch — this is what that conversation actually looks like, aimed at telling the difference between a volume spike and a structural break.

Book a 15-Min CS Audit