You have a customer who pays $99/month. They submit 40 tickets per month (your average customer submits 2). They call your agent "incompetent" regularly. They've demanded to speak to a manager 11 times this quarter. Three agents have asked not to handle their tickets anymore.

At $10 per ticket in loaded agent cost, this customer costs you $400/month in support labor. They pay $99. You're losing $301/month, and that doesn't include the morale damage to your team.

It's time to fire this customer.

When Firing Is the Right Call

Firing a customer sounds extreme. It should be rare. But there are clear signals that a relationship has become net-negative.

The math doesn't work. When the cost to serve a customer exceeds their revenue, and there's no path to fixing that, you're subsidizing them with money that could go toward serving good customers better. Do the math explicitly: their monthly revenue vs their monthly support cost (tickets × average cost per ticket, plus any special handling, escalations, or manager time).

They abuse your team. Rudeness, hostility, personal attacks, profanity directed at agents, discriminatory language. One bad interaction is a bad day. A pattern of abuse is a management problem you're choosing not to solve.

They monopolize resources. Some customers file tickets about everything. They want hand-holding for features that are documented. They file bugs that aren't bugs. They demand responses within minutes for non-urgent issues. This customer consumes 10x the resources of an average customer, which means they're taking time away from 10 other customers.

They refuse reasonable solutions. You've offered refunds, credits, workarounds, personal attention. Nothing satisfies them. Every resolution generates a new complaint. The goalposts move every time.

The Decision Framework

Before you fire a customer, ask four questions.

Have we done everything reasonable? Did we try to resolve their underlying issue? Did we offer alternatives? Did we give them feedback about their behavior? Some customers don't know they're being difficult. A direct conversation ("I want to continue helping you, and I need us to communicate respectfully to do that") fixes it sometimes.

Is this a pattern or a moment? A customer having a terrible week because your product lost their data deserves patience, even if they're angry. A customer who's been hostile for 6 months isn't having a bad week.

What's the financial impact? Losing a $99/month customer is nothing. Losing a $10,000/month enterprise customer requires a different calculation, even if they're difficult. The threshold for firing should scale with the revenue at risk.

Is there a legal risk? Customers in protected classes, customers with active disputes, customers who might claim discrimination. Consult legal before terminating relationships that could create liability.

The Process

Step 1: Document the pattern. Compile the data: ticket count, agent complaints, cost to serve, specific incidents of abuse. This documentation protects you if the customer disputes the termination publicly.

Step 2: Give a final warning (if the issue is behavior). "We value your business, and we want to continue serving you. To do that, we need interactions to remain respectful. Specifically, [describe the behavior]. If this continues, we may need to end our service agreement." Put it in writing. Email, not chat.

Step 3: Make the decision. If the behavior continues, or if the math is the issue (not behavior), proceed.

Step 4: Send the termination notice. Be clear, direct, and professional. Not cold. Not punitive. Factual.

"After careful consideration, we've decided to discontinue your account effective [date, 30 days from now]. We'll ensure a smooth transition: your data will be available for export until [date], and any remaining balance will be refunded to your original payment method. We wish you well."

Give 30 days notice. This is respectful and covers you legally in most jurisdictions. Don't terminate immediately unless there's a safety concern (threats, illegal activity).

Step 5: Refund proactively. Refund any unused portion of their subscription. Don't make them ask. This removes a potential grievance and chargeback risk.

Step 6: Help them transition. Provide data export instructions. If they ask for recommendations of alternative products, give them (seriously). The goal is a clean ending, not scorched earth.

The Team Impact

Here's what most people don't realize: your team already knows which customers should be fired. They've been dealing with this person for months. They've vented in Slack about it. They've dreaded seeing the name in their queue.

When you fire a toxic customer, your team's morale improves immediately. The message it sends: "We value our people more than a single customer's revenue." That message is worth more than the $99/month you're giving up.

The opposite message (tolerating abuse to keep revenue) tells your team that their wellbeing is less important than the P&L. Your best agents will leave for a company that doesn't make that trade.

What About the Blowback?

The fired customer might leave a negative review. They might post on social media. They might tell their friends.

Usually, nothing happens. Most fired customers go quietly. They know they were difficult. They find another provider and the cycle may repeat, but that's no longer your problem.

If they do go public: don't engage in a public argument. A brief, professional response: "We wish [customer] well. We made the difficult decision to end our relationship and provided a full refund and transition period." Don't explain the details. Don't defend. The people reading the review will draw their own conclusions.

The Automation Angle

AI can flag candidates for termination automatically. Supp's analytics track ticket volume per customer, sentiment trends, and agent interactions. A customer with 40 tickets/month, consistently negative sentiment, and multiple agent escalations shows up in the data before a manager notices.

Set up alerts: any customer exceeding 10x your average ticket rate, or any customer with more than 5 agent complaint flags in a quarter. These alerts start the documentation process automatically.

You don't automate the firing. That always needs a human decision and a human communication. But the data that supports the decision should surface automatically, so you're making the call based on patterns, not a single bad interaction.