Cost-to-serve: why your best customers may be losing you money

The customers a board talks about most are usually the biggest by revenue. They get the senior relationship, the bespoke service and the sharpest prices. Sometimes they are also the most profitable. Often nobody knows, because the costs they create are spread across the business as averages. These are the questions we get asked most when we raise it.

What do you mean by cost-to-serve?

Technically it includes every penny spent in serving a customer with the demand they place. This should include every cost incurred between a confirmed order and cash in the bank that varies by customer, including order handling, picking, packing, delivery, returns, credit notes, rebates, promotional support, account management, special labelling, compliance charges and the financing cost of payment terms.

Most management accounts allocate these by revenue share. But this mechanism assumes a customer buying £1m in full pallets to one distribution centre costs the same to serve as one buying £1m in mixed cases to 40 stores; which is never true!

How can a large customer lose money?

The table below is illustrative. It is built to show the mechanism, not taken from any business.

Customer A is nearly seven times larger by revenue but slightly loss-making after cost-to-serve. Customer B, which rarely features in board discussions, makes a 20% contribution. Neither result is visible if you look only at gross margin.

This pattern is common enough across organisations and is known as the whale curve. The curve plots cumulative profit against customers ranked from most to least profitable. A frequently cited pattern is that the top fifth of customers generate somewhere between 150% and 300% of net profit, with the tail giving much of it back. Your own curve will differ, but if you have never drawn it, the shape is likely to surprise you.

Why does this matter more now than five years ago?

Because the costs that sit inside cost-to-serve are rising faster than most prices. Logistics UK's July 2026 cost guide reported vehicle operating costs up more than 12% in the year to April 2026, driver employment costs up almost 8% and insurance up 7%. Prime warehouse rents for large units reached £13.38 per square foot, up 13% in a year.

Warehouse and customer service labour has also moved. The National Living Wage rose to £12.71 an hour in April 2026, and employer National Insurance is 15% above a £5,000 annual threshold. Those costs land on the activities that high-service customers consume most: small picks, split deliveries, returns processing and admin.

Price concessions compound the problem. McKinsey's pocket price analysis showed how on-invoice discounts and off-invoice items such as rebates and cash discounts can take a large share of list price before anyone notices. When that leakage is concentrated in the same accounts that consume the most service, the effect on contribution is doubled.

How do we measure it without a two-year project?

Use the data you already have. You need order lines, deliveries, pallets or cases, returns and credit notes by customer, plus the cost pools for warehouse, transport, customer service and finance. Allocate each pool by the activity that drives it: cost per drop, per line, per return, per credit note.

Aim for directionally right within six weeks, not perfect within a year. An allocation that is 80% accurate and finished will change more decisions than a perfect model nobody completes. Finance should own the model; operations should own the driver data and sign off that it reflects how the warehouse really works.

What do we do with the answer?

While losing loss-making customers to reduce losses is a logical answer, losing a large customer rarely removes its fixed costs, as the overheads it was carrying land on everyone else. Moreover, there could be some loss-making customers which are crucial to your strategic long-term goals. The goal is to change the way the supply chain operates, so that each customer pays, one way or another, for what they consume.

The usual levers are minimum order values or drop sizes, delivery charges or tiered service levels, rebates linked to behaviour that lowers your costs (full pallets, forecast accuracy, fewer returns) rather than to volume alone, and payment terms priced into the deal. The commercial director should lead these conversations with the numbers in hand. A menu of service options gives the buyer something to choose between, which tends to land better than a flat increase.

Some accounts will still remain marginal for strategic reasons. That is a legitimate choice, as long as it is a choice, and the board knows the cost of making it.

Our free Cost-to-Serve Quick Check gives a first view of where your own whale curve bends.