When to stop selling to a customer who costs you money

There is an account in your customer list that everyone in the company already knows about. It pays late, it escalates most weeks, and it renewed last time at a discount. Nobody has worked out what it costs to keep.

This is not the churn that happens to you. I have written about that one: what your churned customers had in common is about reading the leavers for a pattern, and your first renewal conversation is a sale, not admin is about earning the renewal you want. This piece is the other kind: churn you choose, and the arithmetic behind choosing it.

The question lands harder here than it would in a bigger market. MBIE's small business fact sheets, built on the Stats NZ Business Demography Survey, count 101,253 New Zealand businesses with 1 to 5 employees and another 43,068 with 6 to 19. At that size a difficult account is not a line in a portfolio report. It takes up a visible share of your week.

Work out the number before you argue about the feeling

The argument inside the company is always about how the customer behaves. The decision has to be about what they contribute. Those are two different conversations, and only one of them has an answer.

Take a customer paying $2,000 a month, so $24,000 a year. At a 70 percent gross margin that is $16,800 of contribution before anyone touches them. Now count the touching.

• Four hours a month of support, at a loaded cost of $90 an hour: $4,320 a year.
• Two escalations a quarter, four hours of senior time each, at $150 an hour: $4,800 a year.
• Paid 45 days beyond terms, so roughly $3,000 of cash sitting out at any time, which at a 10 percent cost of capital is about $300 a year.

That is $9,420 of cost to serve against $16,800 of contribution. You keep $7,380 on $24,000 of revenue, about 31 percent, on a book that should sit closer to 60. Then the renewal lands at a 15 percent discount, which takes $3,600 off the top and about $2,520 off contribution, and you are down to $4,860. Twenty percent, for your most demanding customer.

Which cost do founders miss most?

Not the support hours. Those get counted, because somebody logs them.

What gets missed is senior attention, because it never appears on an invoice. When your lead engineer spends a Thursday on an escalation, nothing in your accounts moves. What moves is the release that slipped, and the three other customers who were waiting on it.

The discount gets missed too, because it compounds. A price you conceded once becomes the base for the next negotiation, so the account that costs the most to serve is now the cheapest you sell to.

And an unhappy customer is not neutral in a market this size. They are a reference you have paid for and cannot use.

Three repairs to try before you exit

Most accounts like this were sold wrong rather than bought wrong, and that is fixable.

Reprice it. Put the real cost of serving them in front of them, hours included, and quote the number that makes it work. Some will pay it, and that is the cheapest outcome available to you. You will not know until you ask.

Rescope it. Usually the cost sits in one thing nobody ever agreed to: a report rebuilt by hand each month, a phone line that was never in the contract, or an integration you maintain for a single customer. Take that out of the scope rather than out of the relationship.

Write it down. Payment terms, one support channel with a stated response time, named contacts, and a cap on included hours. Behaviour follows the agreement far more reliably than it follows a conversation about behaviour.

Give the repair a date. One renewal cycle, or one quarter. Without a date it becomes the thing you keep meaning to raise.

Leaving without a public fight

If the repair does not hold, exit on your terms and keep it boring. Give real notice, in writing, with a reason that is commercial rather than personal: the scope has moved beyond what you can support at this price. Help them land somewhere, including an introduction to a partner or a competitor who fits them better. Export their data before they ask for it. Honour the term you are in rather than cutting it short.

New Zealand is small enough that how you leave gets repeated. A clean exit is forgotten within a month. A messy one gets told to the next three buyers who ring that customer for a reference on you.

Then look at what the account was blocking, because that is where most of the return sits. The margin you recover is the smaller half of it. The bigger half is the week your senior people get back, and the customer you hand it to instead.

So take your worst account and run the number. What is it actually paying you to keep?

Sources. Payment times: Xero Small Business Insights, New Zealand, September 2026 quarter, published 29 October 2026, which reports that New Zealand small businesses waited 25.4 days on average to be paid, 0.3 days longer than the June quarter, and were paid 5.2 days beyond terms on average, up 0.9 days. The 45 days beyond terms in the worked example is not that average. It is the kind of problem account this article is about, and it is my number rather than Xero's. Business counts: MBIE, "Small business and manufacturing fact sheets" (2026 release, small business fact sheet), built on the Stats NZ Business Demography Survey at February 2025, which counts 617,334 New Zealand businesses, of which 101,253 have 1 to 5 employees and 43,068 have 6 to 19, and states that 97.2 percent have fewer than 20 employees. I have used the MBIE table rather than mixing it with Stats NZ's own provisional release, because a count has to come from one source to mean anything. The worked example is arithmetic rather than research, and the rates are illustrative, so substitute your own: $2,000 a month is $24,000 a year; at a 70 percent gross margin that is $16,800 of contribution; 48 support hours at $90 is $4,320; 32 hours of senior time at $150 is $4,800; $3,000 of cash outstanding at a 10 percent cost of capital is $300; total cost to serve $9,420, leaving $7,380, which is 30.8 percent of $24,000; a 15 percent discount removes $3,600 of revenue and $2,520 of contribution, leaving $4,860, or 20.3 percent. No published survey measures what share of New Zealand B2B accounts are unprofitable to serve, and the often quoted claim that the best 20 percent of customers produce 150 to 300 percent of total profits traces only to secondary summaries, so there is no figure here for either. The three repairs, the exit sequence and the judgement about what gets repeated in a small market come from my own work with New Zealand B2B tech companies rather than from published data.

Nick Burns is a fractional CRO for New Zealand B2B tech companies, at home and expanding internationally. He co-founded Emendo and sold it to McKesson, then the 14th-largest company on the Fortune 500, and has since helped 60+ tech companies grow sales.