When to fire a customer

The best founders fire customers deliberately and periodically. The customers who are most expensive are not the ones generating the least revenue — they are the ones whose requirements are shaping the product away from its most valuable use case. Here is the diagnostic for identifying them and the process for exiting without destroying the relationship.

The customer who is costing you more than they pay.

There is a category of customer that most early-stage companies acquire and very few ever exit. They are usually not the smallest customer by revenue — often they are among the largest. They are demanding, detailed, and specific about what they need. They push back on pricing. They escalate when things break. They submit feature requests that are custom to their use case rather than generalisable to the product. They require more support hours per dollar of revenue than any other account. And they have been in the product long enough that the engineering team has built for them, sometimes without realising it.

These customers are expensive in ways that do not appear in revenue accounting. The direct cost is the service overhead: the engineering hours on custom work, the support tickets, the account management calls. That cost is real and can be calculated. But the strategic cost is larger and rarely measured: the product decisions that were made because this customer needed them, the features that are now in the product not because they serve your ideal user but because they satisfied a specific enterprise requirement, the roadmap conversations that are anchored to this customer's needs rather than to the use case that would produce the most compounding value at scale.

The strategic cost compounds. Every month you keep building for the wrong customer profile, your product drifts further from the version of itself that your best-fit customers would love. And the further it drifts, the harder it is to find and close the customers who would generate better unit economics, higher NPS, and organic word of mouth. The customer you cannot afford to lose may be the customer who is preventing you from finding the customer you cannot afford to miss.

The diagnostic: four questions.

The test for whether a customer should be exited is not a revenue threshold — it is a set of four questions that measure the relationship between this customer's requirements and the direction of the product.

Question one: Does this customer's use case represent the most valuable version of the problem we are solving? This is the foundational question. If you won ten more customers exactly like this one, would you have the company you want to build? If the answer is "I am not sure" or "probably not," you have a customer that is pulling your product in a direction that your business model cannot support.

Question two: Are we building features for this customer that we would not build without them? Audit the last six months of your roadmap. For every feature shipped, ask: would we have built this if this customer did not exist? Features built for a specific customer without clear generalisability are a tax on the product. They add complexity, require maintenance, and create surface area that must be tested and supported for a use case that only one customer has. If more than twenty percent of your shipped features over the last two quarters were driven by a single customer, that customer owns part of your product. That is not a partnership — it is a dependency.

Question three: What is the fully loaded cost of this customer? Calculate the direct service costs: support hours billed at your internal cost rate, custom engineering work, professional services delivered outside the contract. Then calculate the indirect costs: the engineering opportunity cost of the custom features, the product management time spent managing this account's expectations, the cost of the sales time required to renew. Divide the total cost into the annual contract value. The ratio will often surprise you.

Question four: Does this customer refer or expand? The customers who are worth keeping at high cost are the ones who generate return on relationship — they refer adjacent customers, they expand their own usage, they become reference accounts that reduce your CAC in the market segment they represent. A high-cost customer who also expands and refers may be worth the overhead. A high-cost customer who neither expands nor refers is a pure cost centre that happens to pay a retainer.

The exit process.

Firing a customer is not a single conversation — it is a managed process that, done well, ends the commercial relationship while preserving the professional one. There are four steps.

Step one: Stop building for them. Before you exit the revenue relationship, exit the roadmap relationship. Stop accepting feature requests from this customer as inputs to your planning process. This is not a communication to make to the customer — it is an internal decision. The roadmap influence ends now. Existing commitments are honoured. New commitments are declined.

Step two: Price the service accurately. At the next renewal conversation, present pricing that reflects the actual cost of serving this customer. This often resolves the problem without requiring an exit conversation: the customer decides the product is not worth the price, or agrees to a contract structure that makes them economically reasonable to serve. Either outcome is acceptable. What you are doing is removing the implicit subsidy that made the relationship expensive to maintain.

Step three: If repricing does not resolve it, have the direct conversation. Tell the customer that the product is moving in a direction that will not serve their needs as well as it currently does. Be specific about what is changing in the product direction. Give them a timeline — typically ninety days — and offer to help them identify alternative solutions. Do not blame the customer. The mismatch is a product-market fit issue on your side, not a failure on theirs.

Step four: Provide a genuine handoff. If there is a competitive product or adjacent service that would serve this customer better, name it and offer to make an introduction. This is counterintuitive but strategically sound: a customer who leaves cleanly and is well-served by your recommendation is a customer who will say positive things about you. A customer who leaves feeling cut off will not. Your reputation in a market segment is built as much by how you exit customers as by how you serve them.

The ICP as a filter, not a description.

The upstream fix for the bad-fit customer problem is an ideal customer profile that is used as a filter in the sales process before a customer is acquired, not a description derived from customers you already have. Most companies write ICPs backwards — they look at their existing customer base and extract a profile. The result is a profile that describes who you have, not who you want. It inherits the mistakes of the past.

A genuine ICP is written from first principles: given the problem we are solving, the product we are building, and the unit economics we need to make this company work, who is the customer who would get the most value from what we do, would be the cheapest to acquire, would churn the least, and would refer the most? Write that profile. Then use it as a go/no-go filter in your sales qualification process. The first time a salesperson brings you an account that does not fit the profile and argues for an exception because "the revenue is good," you are at the decision point that determines whether the problem reoccurs.

• 01
  Audit your roadmap quarterly for customer concentration. If more than twenty percent of features shipped in a quarter trace back to a single customer's requirements, you have a roadmap dependency that is also a strategic risk. The audit should happen before planning, not during it. You cannot plan toward your ideal customer profile if you have not first identified the accounts that are pulling you away from it.
• 02
  Calculate fully loaded customer cost at least twice a year. Include direct service costs, indirect engineering costs, and the opportunity cost of the custom work. The revenue line of the P&L tells you one thing about a customer. The fully loaded cost tells you something entirely different. You cannot make a rational exit decision without both numbers.
• 03
  Exit the roadmap relationship before you exit the revenue relationship. The first step in managing a bad-fit customer out is stopping the product influence. This is an internal decision, not a conversation with the customer. It happens before repricing, before the exit conversation, before anything else. The product needs to stop being built for this customer before the commercial relationship ends, or you will carry the legacy of their requirements indefinitely.
• 04
  Price custom service accurately at renewal. The single most effective mechanism for exiting a bad-fit customer without a difficult conversation is pricing their service accurately. Most high-maintenance customers are subsidised — the cost of serving them is not fully reflected in the contract. When you present a renewal price that reflects actual cost, the customer either self-selects out or agrees to terms that make the relationship economically rational. Both outcomes serve you.
• 05
  Maintain the professional relationship after the commercial one ends. A customer who leaves cleanly is not a burnt bridge — they are a reference point in a market you are still trying to serve. Offer a genuine handoff to an alternative. Be transparent about your product direction. The relationship ends; the reputation does not.

Frequently asked questions.