Skip to content
MarginGraph

Your best customer might be your worst

Two customers, identical revenue, identical gross margin, and a fifteen thousand euro gap in what they leave behind. Cost to serve explains it, and it never appears in your accounts.

10 min readMarginGraph

Ask an owner to name their best customer and they will name the biggest one. Revenue is the only customer-level number most businesses actually have, so it becomes the ranking by default.

The costs that separate a good customer from a bad one sit below the gross margin line: order handling, delivery frequency, returns, technical support, rebates, and how long they take to pay. None of those are recorded per customer in a standard accounting system. They are real, they are large, and they are invisible.

Two customers, same revenue, same margin

Both buy €120,000 a year at a 34% gross margin. Both produce €40,800 of gross profit. Here is what happens after that.

Bakker Retail. Twelve orders a year, averaging eight lines each. Consolidated deliveries. Four returns. Four hours of technical support. Pays on average twelve days after terms.

Vos Group. Ninety-six orders a year, averaging three lines each, of which twenty-two were flagged urgent. Delivery per order. Sixty returns. Forty hours of technical support. Negotiated a 4% annual volume rebate. Pays on average sixty-eight days after terms.

Using an illustrative order-handling time equation of 5 minutes plus 3 minutes per line, plus 10 minutes for an expedited order, at a departmental rate of €0.80 per minute, and an 8% cost of financing receivables. Substitute your own times and rate:

Bakker RetailVos Group
Revenue120,000120,000
Gross profit at 34%40,80040,800
Order handling2781,251
Delivery5404,320
Volume rebate04,800
Returns handling1402,100
Technical support2202,200
Financing of late payment3161,788
Total cost to serve1,49416,460
Net contribution39,30624,340
As % of revenue32.8%20.3%

Same revenue, same headline margin, €14,966 of difference. On your sales report these two customers are identical. On your bank statement they are not.

The shape this takes across a customer base

Robert Kaplan's account of an insurance company's customer base, published by Harvard Business School in 2005, is the cleanest freely available example. After allocating cost to serve: "The most profitable 40 percent of customers generate 130 percent of annual profits; the middle 55 percent of customers break even, and the least profitable 5 percent of customers incur losses equal to 30 percent of annual profits."

Plot cumulative profit against customers ranked from most to least profitable and you get a curve that rises above 100%, flattens, then falls back. It is usually called the whale curve.

Kaplan and Narayanan describe the typical pattern as roughly 20% of customers generating 150 to 300% of profits, about 70% breaking even, and about 10% destroying 50 to 200%. Treat those as a shape practitioners report rather than a measurement, because that is what they are: no population is specified.

For an actual measured case, Guerreiro, Bio and Merschmann published a study of a Brazilian food manufacturer with around 350 customers in the International Journal of Logistics Management in 2008. Ranked by manufacturing contribution, 31% of customers generated 80% of it. After allocating cost to serve, 6% of customers generated 80% of the margin. And 82 customers, 23% of the total, together accounted for less than 0.05% of contribution.

The two rankings did not agree. That disagreement is the whole reason to do the exercise.

What actually drives the difference

The canonical checklist comes from Martin Christopher, reproduced by the Open University: cost of sales, commissions, sales calls, key account management time, order processing, promotional costs, non-standard packaging, dedicated inventory holding, dedicated warehousing, material handling, transport, documentation, returns and refusals, and credit taken.

Kaplan's shorter version of what makes a customer expensive: product customisation, small order quantities, special packaging, expedited delivery, substantial pre-sales support, extra post-sales support, and liberal payment terms.

Payment behaviour deserves separate attention because the numbers are large and public. Across the EU in 2024, business-to-business invoices were paid in an average of 60.3 days against average agreed terms of 43.0 days. Government took 69.8 days. Fifty-two percent of EU companies reported facing problems from late payment, up from 47% in 2023 and 42% in 2021.

The Late Payment Directive gives you tools most businesses never use: statutory interest of at least 8 percentage points above the ECB reference rate, and a minimum of €40 compensation for recovery costs, per invoice, automatically. The European Commission's impact assessment for the proposed replacement regulation reports that almost 81% of the SMEs responding to its consultation said that interest and compensation fees are never paid. That is a self-selected sample rather than a representative one, but the direction is not in doubt. The entitlement exists. Almost nobody invoices it.

Making this practical without building a costing department

Classic activity-based costing collapsed under its own maintenance cost. Kaplan and Anderson documented the failure modes in Harvard Business Review in November 2004: one bank's brokerage operation needed 70,000 employees across more than 100 facilities to file monthly reports, and 14 full-time people just to run the data. One fabricator took three days to calculate costs across 40 departments, 150 activities and 10,000 orders. Scale 150 activities across 600,000 cost objects monthly for two years and you are storing more than two billion items.

Their replacement, time-driven activity-based costing, needs two numbers per department: the cost per minute of supplying capacity, and the minutes each activity consumes. That is the approach used in the table above. It is tractable for a small business in an afternoon with a spreadsheet.

There is one important technical detail. When you ask staff to allocate their time, the percentages always add to 100, which silently assumes zero idle time and overstates your cost rates. Kaplan and Anderson use 80 to 85% of theoretical capacity as practical capacity for people. Use that, and the unused capacity shows up as unused capacity rather than being smeared across your customers.

What to do about a customer who is costing you money

Not fire them. That is last, and the literature is fairly clear on why.

Mittal, Sarkees and Murshed set out a five-step sequence in Harvard Business Review in April 2008: reassess the relationship, educate the customer, renegotiate the value proposition, migrate the customer to a cheaper channel, and divest as a last resort. Fidelity Investments identified low-margin customers making frequent service calls and taught them to use automated channels rather than dropping them. A commercial dye supplier renegotiated to charge separately for on-site service. Their warning about divestment is worth quoting: profitable customers you keep "may wonder if they're next in line and defect to friendlier providers."

The Institute of Management Accountants makes the same point from the arithmetic side: "it is generally less expensive to turn an L customer into a P or B customer than it is to obtain a new customer."

There is systematic evidence pointing the same way. Feng, Morgan and Rego, in the Journal of the Academy of Marketing Science in 2020, studied stock market reactions to public disclosures of unprofitable-customer-management decisions and found an average abnormal return of −0.53%, worsening to −0.61% for direct customer divestment. Investors respond more favourably to indirect strategies than to firing customers outright.

The best documented repricing outcome comes from Kaplan and Anderson's 2004 article. Kemps, a dairy, was losing money serving a chain of specialty shops due to low volume and high product variety. Rather than dropping them, Kemps proposed a price increase and a minimum order size, offered its standard branded product in place of the customer's private-label ice cream, and required full rather than partial truckload orders. The customer accepted a 13% price increase and agreed to eliminate two low-volume products. Annual benefit: $150,000, and a profitable customer where there had been an unprofitable one.

Applied to Vos Group above, the levers are visible in the table. The break-even arithmetic on a price rise tells you how much of this customer you can afford to lose while doing it. A minimum order value would remove most of the €4,320 of delivery cost. A charge for expedited handling would price the twenty-two rush orders. Enforcing terms, or invoicing the statutory interest, would recover most of the €1,788. The 4% rebate is a negotiated giveaway that could be tied to order consolidation rather than volume. None of that requires losing the customer.

Concentration, briefly

If the exercise reveals that a small number of customers carry your profit, that is worth knowing on its own terms. The evidence on customer concentration is genuinely two-sided and comes from listed companies rather than small firms, so read it as direction rather than magnitude.

Dhaliwal and colleagues found in the Journal of Accounting and Economics in 2016 that customer concentration is associated with a higher cost of equity and a higher cost of debt. Campello and Gao, in the Journal of Financial Economics in 2017, found it increases loan interest spreads, adds restrictive covenants and shortens loan maturities. But Patatoukas, in The Accounting Review in 2012, found concentration positively associated with accounting rates of return, through lower operating expenses per euro of sales and better asset utilisation.

Concentration buys operating efficiency and sells financing flexibility. It is a trade, not a fault.

What this does not tell you

Cost to serve is a snapshot of a relationship at one point in time. A customer who is expensive to serve in year one because you are learning their requirements may be cheap in year three. Ranking on a single year will penalise every new relationship.

The financing cost of late payment is an opportunity cost, not a cash cost, unless you are actually borrowing to cover it. If you have surplus cash, the €1,788 is theoretical. If you do not, it belongs in your cash schedule.

And the whole method assumes your cost rates are roughly right. They are averages over distributions. They tell you which customers differ by a lot, which is what you need. They do not tell you a customer's cost to the euro, and you should not present them as though they do.

Customer concentration is a valuation input

Buyers price the risk of your customer base, not just its size. The report states the concentration, the assumption behind it, and what it does to the range.

Business Valuation report — €9

Sources

A note on a statistic we did not use

The claim that acquiring a customer costs five to twenty-five times more than retaining one appears in Harvard Business Review without a citation, and every attempt to trace it leads back to another secondary source. We have used Reichheld and Sasser's actual 1990 findings instead, which are specific and attributable: reducing defections by 5% generated 85% more profit in one bank's branch system, 50% in an insurance brokerage and 30% in an auto-service chain. Frederick F. Reichheld and W. Earl Sasser Jr., "Zero Defections: Quality Comes to Services," Harvard Business Review, September–October 1990. https://hbr.org/1990/09/zero-defections-quality-comes-to-services

Frequently asked

Keep reading

Guide8 min

Gross margin lies: which of your products actually make money

The line that looks like your worst performer is often the one paying for the overhead. How allocation creates fake losses, and which number to use when you decide what to drop.

Guide11 min

Profitable and broke: the gap between your P&L and your bank balance

Your accountant says you made sixty thousand. Your bank balance fell by sixty-seven. Both are correct. Here are the seven bridges between them, and which ones you can move.

Answer2 min

Does customer concentration lower my valuation?

The direction is well evidenced and the size of the effect is not. Concentration raises a buyer's cost of capital and reduces the number of bidders, which is where the money goes.