How to calculate cost to serve, step by step.

A worked example

Take one wholesale customer billing 240,000 euro a year, on goods that cost 168,000, so the gross margin is 72,000 euro, a healthy 30 percent. Now cost the way they actually buy. Their 320 orders, deliveries, support calls and returns consume, across warehouse, transport and service, about 30,000 minutes of capacity in the year. At a fully loaded 0.80 euro per minute, that is roughly 24,000 euro of cost to serve.

Illustrative. Cost to serve quietly eats a third of this customer's gross margin. Give the same revenue to an account that splits it into many small urgent orders with frequent returns, and those 30,000 minutes can double, halving the net margin from 20 percent to 10. It takes about three times those minutes before the account stops paying for itself, and no price has to change for that to happen.

What does cost to serve typically run?

Honestly: nobody can tell you yet. We have looked for a public benchmark of cost to serve as a share of revenue, by sector, and it does not exist. Consultancies quote it in private decks, ERP vendors imply it in case studies, and no one publishes the distribution. So we are building it.

The Cost-to-Serve Index is collecting, in defined ranges rather than exact figures, how much of revenue goes to serving customers, which activities drive it, and how confident finance teams are in their own number. Every response sharpens the picture. Once the sample per sector is large enough to be meaningful, the bands will be published here, on this page, with the method open for inspection. Until then, cost to serve by industry sets out what good looks like, sector by sector.

The same revenue, and the point where it stops paying

Take the customer above and give an identical 240,000 euro of revenue to a second account, on the same goods, so the same 72,000 euro of gross margin. The only thing that changes is how they buy: instead of 320 planned orders, 1,150 small urgent ones, with returns on roughly one delivery in twelve. Their orders consume about 60,000 minutes of warehouse, transport and service capacity instead of 30,000.

The useful number, though, is the one further along. At 0.80 euro a minute, this account absorbs its entire gross margin at 90,000 minutes, because 90,000 × 0.80 = 72,000 euro, exactly the margin it brings. Below that it still pays for itself. Above it, every extra order is served at a loss, and nothing in the revenue report says so. That threshold, not the average, is the number worth putting in front of a sales director, because it is the one they can act on: it turns an argument about which customers are bad into a question about how many minutes an account is allowed to consume before the terms have to change.

Illustrative figures, chosen to show the arithmetic. The break-even in minutes is specific to each resource pool and rate, but the shape holds: a customer stops paying at the point where minutes consumed multiplied by cost per minute reaches gross margin.

Where the spreadsheet stops working

A spreadsheet can show you the number once. It cannot show it to you every month, for ten thousand customers, in a way a board will trust.

The method above is honest work, and a capable finance team can build a first version in Excel. The limit is not the maths, it is the maintenance. The moment the model has to be rerun monthly, survive a data refresh, and be explained line by line to people who will act on it, manual spreadsheets break. That is the point where we build a model that updates itself and that your team owns afterwards.

Common mistakes when calculating cost to serve

None of these are silly errors, and I have made most of them myself on a first pass. They are the shortcuts a finance team reaches for when the data is not quite there, and each one moves the answer in a predictable direction, which is what makes them worth naming.

• Spreading cost by revenue. Allocating operational cost in proportion to what a customer bills assumes every customer consumes resources the same way. That is the exact assumption cost to serve exists to test, so an allocation built on it can only ever confirm what you already believed.
• Dividing by theoretical capacity. Using 100 percent of paid time as the denominator produces a rate that looks precise and is quietly too low, and it buries idle capacity inside the cost of every order. Practical capacity, usually somewhere around 80 to 85 percent, is the honest denominator.
• Leaving out the fixed minutes per order. Without a per-order term in the time equation, a stream of small frequent orders looks cheap. It is normally the opposite: the fixed handling time is paid again on every one of them.
• Forgetting returns and exceptions. Reverse logistics, reprocessing, expedited shipments, long payment terms and negotiated deductions are all cost to serve. They tend to concentrate in a small number of accounts, which is exactly why leaving them out flattens the ranking.
• Chasing precision too early. Five to eight activities and sensible estimates already show the pattern, and the pattern is what the decision needs. Refining a time equation to the second before anyone has seen the ranking is effort spent on the wrong end of the problem.

If you want a quick test of whether your current numbers are affected: pull the ten accounts with the most orders and the ten with the largest average order, and see whether your reporting shows any cost difference between them at all. If it does not, one of the five above is in play.

Common questions

What data do I need to calculate cost to serve?

Less than people expect. You need order and line counts per customer, a view of the activities that fulfil an order, the total cost of each resource pool (warehouse, transport, customer service) and a sensible estimate of how long each activity takes. Most of this already exists in your ERP exports and operational records.

Can I calculate cost to serve in Excel?

For a single snapshot, yes, and it is a good way to learn. A spreadsheet stops working when the model has to be rerun every month, cope with thousands of customers, or be trusted by a board. At that point the manual links break and no one can explain a number. That is when a maintainable operating model earns its keep.

How long does it take to build a cost-to-serve model?

A focused diagnostic runs in a few weeks. A full operating model typically lands in six to ten weeks, built alongside your finance team so they own it and can update it afterwards.

What is the cost-to-serve formula?

For each customer, cost to serve is the sum across all activities of the minutes that customer's orders consumed multiplied by the cost per minute of each activity. Net contribution is then revenue minus cost of goods minus that cost to serve.