A practical guide to cost-to-serve

Most businesses know their gross margin by product and by customer. Far fewer know what it costs to actually serve each customer: the picking, the delivery, the minimum-order exceptions, the returns, the account management, the credit control. Those costs usually sit in overheads and get spread across customers in proportion to revenue. That is the problem. A customer who orders once a month on a full pallet and one who orders three cartons twice a week can bring in the same revenue and cost you very different amounts to serve.

Gartner put this plainly in April 2025, recommending that supply chain leaders build a cost-to-serve model because many already suspect some customers and products are unprofitable but lack a structured way to show it. You don't need software to start. A spreadsheet and a few days of a finance analyst's time will get you a long way.

Six steps to a first model

1. Pick a scope you can map effectively: one region, one channel or your top 50 customers. Gartner's guidance also suggests starting with a pilot rather than the whole business.
2. List the costs that sit below gross margin and vary by customer: typical ones are warehouse picking and packing, outbound delivery, returns handling, customer service, sales visits, credit control and any rebates or settlement discounts.
3. Find the driver for each cost: picking is driven by order lines, not revenue. Delivery is driven by drops, distance and weight. Customer service is driven by calls or tickets. Rebates are driven by the contract.
4. Work out a rate per driver from last year's actual costs: if your warehouse cost £600,000 and picked 300,000 lines, each line cost about £2.
5. Apply the rates to each customer's actual activity over the same year.
6. Sort customers by profit after cost-to-serve and look hard at the bottom of the list.

The key is not to chase precision. A model that is 80% right and used by the sales team is worth far more than a perfect one that sits in finance.

A worked example

All figures in this example are illustrative. They describe two imaginary trade customers of a distributor, not real businesses.

Both customers buy £200,000 a year at a 30% gross margin, so each brings in £60,000 of gross profit. Here is what serving them costs, using simple illustrative rates.

On a gross margin report, A and B look identical. After cost-to-serve, Customer A earns twice as much. Customer B also takes longer to pay, which ties up cash on top.

This doesn't mean Customer B should be dropped. It means the terms should change. Options include a minimum order value, a delivery charge below a threshold, consolidating deliveries to set days, tying the rebate to order size rather than annual spend, or moving small top-up orders to a parcel service. Each of those changes behaviour without a confrontation about price.

Where cost-to-serve goes wrong

The most common mistake is allocating costs by revenue in the end anyway, because the driver data is hard to find. If you do that, the model will tell you nothing you didn't already know. Spend the time to get order lines and drops by customer; most ERPs and WMSs hold them.

The second mistake is leaving out the commercial costs. Rebates, settlement discounts, free delivery and extended payment terms are often the biggest items, and they sit with sales rather than operations. McKinsey's work on the pocket price waterfall makes the point well: the gap between list price and what the business actually keeps after every discount and allowance is often where the margin goes.

The third is building the model once and filing it. Costs move. Royal Mail raised its business parcel prices from October 2025, with its Royal Mail 24 service going up by roughly 8% to 10% depending on weight, and the National Living Wage rose to £12.71 an hour in April 2026. Your rates per drop and per line need refreshing at least once a year, and whenever a major carrier or labour cost changes.

Some years ago, we were brought in to help improve margin for a food and beverage business. The sales director's first reaction to our proposed ranking of worst margin eroding customers was that the numbers must be wrong. But after walking through two customers line by line, he agreed. The change that made the biggest difference was a minimum order value with a delivery charge below it, which most customers accepted without complaint.

Who should own it

Finance builds the model. Sales has to use it. Operations has to trust the driver data. If any one of the three opts out, it fails. A practical arrangement is that finance owns the model and refreshes it, the commercial director reviews the top 20 focal customers every quarter, and new quotes above a certain size are checked against cost-to-serve before they go out.

The free Cost-to-Serve Quick Check gives you a first estimate in minutes, and Waterfront Advisory builds fuller models when you need one.