Cost to serve: what it really costs to serve each customer

What cost-to-serve analysis shows.

A cost-to-serve analysis takes the customer book your revenue report describes and re-ranks it by what each relationship actually leaves behind. The output is not a theory; it is three concrete artefacts your team can open, sort and argue with.

The first is a net contribution P&L for every customer: revenue, cost of goods, then the operational cost that customer's orders actually consumed, activity by activity. The second is the ranked view of the whole book, the whale curve: a profitable core, a flat middle, and a tail of accounts that quietly gives margin back; the same ranking can be read by product, channel and region, and the tails are rarely in the same place. The third is a break-even line per account: the point where the minutes an account consumes, priced at their capacity cost, absorb its entire gross margin. The four decisions further down this page all flow from those three artefacts, because each one turns an argument about who is a bad customer into a number about how they buy.

The arithmetic behind all of it is one line, minutes consumed multiplied by cost per minute. The five steps that produce it, with a worked example and a downloadable calculation template, are on how to calculate cost to serve. If you would rather try it on a customer than read the method, the cost-to-serve calculator takes revenue, product cost and service behaviour and shows the net margin activity by activity. This page is about what the analysis shows and what to do with it.

Your accounts tell you the company made money. They don't tell you which work made it.

Most operationally complex businesses run on a P&L that aggregates everything above the gross-margin line, then aggregates everything below it. It tells you whether last quarter was good. It doesn't tell you which customers were quietly subsidising others, or which orders cost more to fulfil than they earned.

A cost-to-serve model attributes operational cost down to the individual customer and order. The result is a ranked, defensible view of who actually contributes.

Whale curve. A few customers carry the profit; the tail erodes it. The peak is the profit you would have without the loss-makers.

What is included in cost to serve?

Cost to serve covers everything a customer consumes after the product exists and before the cash arrives, and nothing that belongs to the product itself. The cost of the goods sits above it, on its own line. Unused capacity sits beside it, on its own line, so it is never smuggled into a rate. What is left is the work: taking the order and its amendments, picking and packing it, dispatching it with whatever paperwork the destination demands, delivering it and waiting at the door, taking it back when it is returned, answering the calls it generates, invoicing it in the format the customer requires, and chasing the payment. Each of those has a driver that moves its minutes, and in almost every case the driver is already a field in a system you own.

Two customers, the same revenue, very different margins.

Take two customers with the same annual revenue of 100,000 euros. Customer A orders on a regular pattern, pays on time and rarely needs support. Customer B places last-minute orders, asks for urgent deliveries, wants constant customisations, returns product often and expects a monthly meeting. In the accounts they look equally profitable. Once cost to serve is attributed, Customer B can cost three to five times more to serve than Customer A.

That is the pattern, not the exception. The customers that consume the most resource are rarely the ones that look expensive, and the largest accounts by revenue are rarely the most profitable. Small changes in how a customer is served move the margin more than most pricing rounds do.

Knowing the number is only useful if it changes something. Four moves follow from it:

• Reprice the expensive services: urgent deliveries, customisations, small-order handling and returns, so the price reflects the work.
• Introduce tiered service levels, set by the margin the account actually earns.
• Renegotiate terms with the high-cost accounts, order size, delivery frequency and lead time, before touching the price.
• Exit, in the extreme cases, the relationships that destroy value consistently and cannot be repaired.

Which of the four to reach for first, and in what order, is the question our cost management consulting work exists to answer, with the model in hand rather than a savings target.

Cost to serve, cost-to-serve analysis and total cost to serve.

The three phrases get used interchangeably and they are not the same thing. Cost to serve is a number: the operational cost of serving one customer, one order, one channel or one region, over a period. Cost-to-serve analysis is the exercise that produces those numbers and ranks them, which is where the whale curve comes from. Total cost to serve is the sum for the whole book, and it is the one to be careful with, because some writers use it to mean cost to serve plus the cost of the goods, which is a materially different figure. Before comparing two cost-to-serve percentages, check which of the two the other side counted.

The spelling varies as well and carries no meaning: cost to serve, cost-to-serve and CTS are the same thing. What is not the same thing is landed cost, which stops at the point the goods reach your warehouse, or cost of goods sold, which never leaves the product. Cost to serve starts where both of those end.

Cost-to-serve software: spreadsheet, BI or a costing engine.

The tool question comes up in every engagement, and the honest answer has three tiers. A spreadsheet is the right place to learn: one snapshot, a few thousand rows, one analyst who understands every link. A BI layer on top of your data warehouse is the right place to publish: it shows the ranked customer book beautifully, but it does not compute it; someone still has to build the allocation logic underneath. A costing engine is the right place to run the model monthly: activities, time equations and capacity cost rates live as first-class objects, the model reruns when the data refreshes, and every number stays explainable line by line.

We build client models on the CostCtrl platform because maintenance is where spreadsheet models die, but the method is software-agnostic and the model you are handed is yours whichever tier you run it on. If you are evaluating tools, the buyer's guide to cost-to-serve software sets out the questions to ask any vendor, including us.