When to Fire a Customer

Retention advice runs in one direction: churn down, always, as far as it will go. That breaks at the bottom of the range. You can have too little churn.

I have told a client this to their face. Terrific retention, relatively slow sales growth. My read was that they actually didn’t have enough churn, and that they should accelerate sales growth even if it raised the number. They could afford to buy more churn. Churn only becomes a problem when it makes it hard to grow. Below that point the goal isn’t the lowest number, it’s the balance of sales growth and churn that produces the most overall growth.

Zero churn is not a valid goal, because it really just means zero growth.

Too little churn says something specific about how you are behaving. It means you are not taking enough bets on customers, given how effectively your solution provides results. If your product reliably gets people to a result, then every customer you decline to sell is a result you decided not to produce. Near-zero churn says you are only betting where you already know you will win.

Zero churn is not a goal. It really just means zero growth. I have sat across from a client with terrific retention and told them they don’t have enough churn — and that they should accelerate sales even if it costs them some.

There are two reasons zero churn can’t exist. The first is structural: for any decay curve, the slope of the slant asymptote is the natural death rate, and that exogenous churn stays there no matter what you do. The second is operational: cutting churn requires limiting sales, and below some company-specific threshold, cutting it further makes growth impossible. Which means an optimal churn rate exists.

That optimum has a floor and a ceiling. “The optimal churn rate must be: 1. below the sustainable sales growth rate; 2. above the rate at which it slows sales.” The level that maximizes growth is the right one, not the level closest to zero.

In practice a very low churn number usually means one specific thing. Only about 3% of people and companies get extraordinary results — in SaaS, very low churn and rapid growth at the same time. For everyone else it runs binary: high growth with high churn, or low growth with low churn. When I see near-zero churn without the growth beside it, the safe reading is a company that has stopped selling past the customers it already knows how to keep.

This argument gets misused, so let me be exact about what I am not saying. I don’t believe in good churn. Bad-fit churn is arguably the worst churn of all, because those customers should never have been sold in the first place. Churn is always bad, and here is what that means precisely: every customer who leaves is one you did not successfully drive results for. That is what the number is counting. So the two halves of this argument sit at different levels — each customer who leaves is a failure, and a company with almost none is a company that stopped betting. Unavoidable is not the same as good.

None of this is an argument for losing more customers. It’s an argument for stopping one specific habit: treating the churn number as the objective, and avoiding the decision underneath it — who you are not going to serve.

Start with who. “There’s no such thing as your ideal customer. There’s just customers and those who leave.” You don’t have less-than-ideal customers. You have some who haven’t figured out yet that they really aren’t a customer.

That reads cold until you notice it’s a claim about your data, not about people. The wrong-fit customer is not a silhouette you invent in a workshop. ICP is bunk — it’s all about use case. Use case matters more than company type; customers with the right use case stay regardless of who they are. And the biggest mistake SaaS companies make is presuming they know their customers’ purpose, because they built the product and assume they know why anyone signed up. Your ideal customers are the ones who achieve the best results, and that isn’t a decision you make — it’s something you observe.

Once you can see the split, the cost of ignoring it shows up in the customers you are trying to keep.

Customer success is not for all customers — it is for the good ones. Knowing which are the good ones is one of the real methods of the job, and the group you can meaningfully serve tops out around 20%, commonly closer to 5%. The rest get fired or automated, not because they’re bad people, but because there isn’t enough of you to go around.

The mechanism is in the decay curve. “Interventions must be designed to specifically impact the customers above the asymptote; interventions for customers below the asymptote can have no positive impact on churn.” Effort spent below the line does nothing at all — those customers were always leaving. In one base, 60% of churn was not retainable and 40% was. Meanwhile, flooding the system with every candidate “diminishes economic performance significantly… by diluting efforts” on the customers who would have succeeded and told other people about it.

An hour spent on a customer who was never going to stay is an hour taken from one who would have. That is how the wrong customer costs you retention somewhere else in your base.

That is the retention argument for letting someone go, and almost nobody runs it. Keeping a customer your product can’t make successful doesn’t cost you nothing. It costs you the results a right-fit customer didn’t reach, because the person who would have driven those results spent the quarter somewhere else.

The rest of the cost is the ordinary kind. 80% of trouble comes from 20% of customers. Fire them. You lose money on them, they ruin your brand, and they ruin your people. Then dedicate yourself to the best 20%, profile them to find more, and automate the rest. I’ve noted the same pattern from outside business: the bottom quintile is killing education — maybe it’s the same for the bottom quintile of customers.

My test is a single question: is this customer damaging the machine? The machine is your economics, your product, your market, your strategy, your speed, and your people. If none of those are being damaged, the customer stays. If one of them is and you can name which, you have a decision to make and you should stop pretending you don’t.

The other half of that is procedural. There should be process and rules for escalating bad, malignant behavior and endless scope increases. Written down, in advance. A consequence only counts if it was real before it was needed, and a boundary you have never been willing to enforce is just stated tolerance.

Two guardrails, because this argument gets abused.

The first: a troubled account is not an unsave-able one. Never write a client off as bad — troubled accounts need success planning more, and have the potential to respond even better than comfortable ones. The filter is save-ability, not visible distress. Distress is often the signal of a customer doing real work.

The second: a customer who cannot get a result from your product is not being served by staying. Customers who achieve measurable results stay six times longer on average. The ones who never get there are paying you every month for an outcome they are not going to reach, and every renewal you protect is another year of that. Letting them go ends it.

Most of this decision, though, is never a firing at all. It happens years earlier. 90% of churn happens in the sale — selling something we don’t do, or selling to the wrong fit — and how we sell and who we sell to have a bigger impact on customer outcomes than anything else we do. Virtually all churn comes from selling to the wrong customers. When you scale sales and marketing up-market or down, you’re not buying revenue or margins. You’re buying a churn rate.

That reframes the sales function. The purpose of sales and marketing is not convincing but filtering — causing the customers who are interested enough to subscribe, but not interested enough to change their behavior, to choose not to subscribe. Craft the offering so it only works for the right clients, and price it so it disincentivizes the wrong use case. Don’t sell customers who are too big for you either — successful large customers often grow out of the solution, and you don’t have the higher-end functionality to hold them.

You don’t have to sell correctly to all your customers, just the ones you want to keep. And you don’t have to make all your customers successful, just the ones you want to keep.

Everyone agrees with that sentence and then staffs and sells as though it were false. The decision it describes is already being made inside your company — by whoever accepted the last bad-fit deal, and by whoever spent Tuesday on an account that was never coming back.

You can check all of it this week without a single new tool. Rank your customers by how long they actually stayed, and look at the top group — that’s your real profile, written in behavior rather than in a persona doc. Then look at the bottom group and at what your team spent on it last quarter. Then put your churn rate next to your growth rate and work out which one is holding the other back. A flat trajectory is death, so the answer is never to stop growing. It’s to grow well.