What is cost to serve, and why does gross margin miss it?

Cost to serve is everything you spend to deliver on a customer relationship after the product itself is made: order handling, picking and packing, delivery drops, returns and credits, account management time, technical support, and the financing cost of the payment terms you granted. Gross margin stops at the cost of the product, so it measures how profitable a thing is, not how profitable a customer is. The two diverge whenever customers buy the same goods in very different ways. A customer ordering 480 times a year consumes far more handling and delivery than one ordering 26 times for the same revenue, and none of that difference appears above the gross margin line.

Why is my largest customer often my least profitable?

Because size buys leverage, and leverage gets spent on price and on service. Large accounts negotiate the lowest prices, the most frequent deliveries, the smallest order minimums, the most generous returns, and the longest payment terms. Every one of those concessions sits below the gross margin line, so the account still reports an acceptable margin percentage while consuming a disproportionate share of the operation. In the worked example here, a national account at 1,200,000 dollars of revenue and a 25 percent gross margin returns 10,000 dollars of contribution after cost to serve, while a regional account at 340,000 dollars returns 84,000 dollars.

Should I raise the price or change how I serve the customer?

Usually the service model, because the numbers are larger and the conversation is easier. Raising price is a single hard negotiation that transfers value away from the customer. Changing the service model removes cost that the customer often does not value in the first place. In the worked example, lifting the national account to a 10 percent contribution margin needs a 9.2 percent price increase, which few buyers accept. Consolidating 480 deliveries a year into 160 saves 128,000 dollars and moves the same account to an 11.5 percent contribution margin, without touching the price. Order minimums, delivery frequency, and returns policy are usually where the recoverable money sits.

Should I fire an unprofitable customer?

Rarely as a first move, and never on the analysis alone. Cost to serve identifies where value is leaking; it does not tell you what happens to the capacity afterwards. Much of what you save is only avoidable if you actually remove it. If you exit a customer and the trucks, the warehouse space, and the support headcount all remain, you have given up the contribution and kept the cost, which makes you worse off than before. The sequence that works is to reprice or re-serve first, agree in advance what capacity gets redeployed or removed if the customer walks, and treat exit as the outcome you are willing to accept rather than the outcome you are seeking.

When does cost-to-serve analysis mislead?

In four situations. When the cost drivers are guessed rather than measured, the allocation simply moves opinion around with more decimal places. When a customer carries strategic value the analysis cannot see, such as reference status, market access, or a volume commitment that keeps a plant economic. When fixed costs are treated as though they disappear on exit, which overstates the gain from losing an account. And when a single period is read as a trajectory, since a growing customer absorbing setup cost and a declining one consuming service look identical in a snapshot. The analysis is a lens on one year, not a verdict on a relationship.

The Revenue You Should Want to Lose

Core Argument

Two customers. One buys 1,200,000 dollars a year at a 25% gross margin. The other buys 340,000 dollars at 30%. Every account review in the business treats the first one as the priority. After cost to serve, the large account returns 10,000 dollars and the small one returns 84,000. The company is not being paid for its biggest relationship. It is subsidising it, and nothing in the standard P&L makes that visible.

The core insight

Ask a finance team which customers are profitable and you will usually get an answer built from gross margin. It is the number the system produces, it reconciles to the P&L, and it feels like customer profitability. It is not. Gross margin measures how profitable a product is. It says nothing about how expensive a customer is to keep.

Everything that separates one customer from another sits below the gross margin line. How often they order. How small each order is. How many separate places you deliver to. How much they send back. How much account management and technical support they absorb. How long they take to pay, and what that costs you in working capital. Two customers can buy identical goods at identical prices and consume wildly different amounts of the business.

That cost is real, it is large, and in most reporting it is pooled into overhead and spread across everyone as a percentage of revenue. Averaging is precisely the wrong treatment, because it charges the average to the customer whose costs are furthest above it. The demanding account looks fine. The easy account quietly pays for it.

"Gross margin tells you whether the product was worth making. Cost to serve tells you whether the customer was worth keeping."

The framework: contribution after cost to serve

The method is one line, and the work is entirely in populating it honestly.

Customer Contribution

Contribution = Gross Margin − Cost to Serve

Cost to serve covers order handling, delivery, returns and credits, account and support time, and the financing cost of payment terms.

You do not need a full activity-based costing programme to start. You need five drivers and a defensible rate for each: cost per order processed, cost per delivery drop, credits and returns, servicing hours at a loaded rate, and the carrying cost of receivables. Rates good to plus or minus 20% are enough, because the differences between customers are usually measured in multiples, not percentages.

Work through it with two real customers

Both buy from the same catalogue. The national account negotiated a lower price and a weekly delivery schedule into 4 regional depots. The regional account buys in full pallets, twice a month, to one address, and pays in 30 days.

Line	National account	Regional account
Revenue	$1,200,000	$340,000
Cost of goods	$900,000	$238,000
Gross margin	$300,000	$102,000
Gross margin %	25.0%	30.0%
Orders placed per year	480	26
Order handling at $180	$86,400	$4,680
Delivery at $220 per drop	$105,600	$5,720
Returns and credits	$42,000	$2,400
Account and support time	$38,000	$3,500
Financing of payment terms	$18,000	$1,700
Cost to serve	$290,000	$18,000
Cost to serve, % of revenue	24.2%	5.3%
Contribution after cost to serve	$10,000	$84,000
Contribution %	0.8%	24.7%

Illustrative figures. Rates are the kind most distributors and manufacturers can derive from existing systems in an afternoon.

The national account is 3.5 times the revenue and returns less than one eighth of the contribution. Its gross margin percentage was 5 points lower, which anyone could see. Its cost to serve was 19 points higher, which nobody was measuring.

The driver is order size, not price. The national account averages 2,500 dollars per order across 480 orders. The regional account averages 13,077 dollars across 26. The price concession cost 60,000 dollars a year. The ordering behaviour cost 192,000.

What actually fixes it

Once the number exists, there are only 3 levers, and they are not equally powerful.

Price. Lifting the national account from 0.8% to a 10% contribution margin requires 110,000 dollars more, which is a 9.2% price increase. Very few buyers of that size accept 9%.
Service model. Consolidating 480 deliveries into 160, by moving depots to a 3 week cycle with an order minimum, saves 57,600 in handling and 70,400 in delivery. That is 128,000 dollars, which takes the same account to an 11.5% contribution margin without touching price. It also costs the customer almost nothing they value.
Exit. Real, occasionally correct, and almost always the wrong opening move. See the limits below before you reach for it.

This is the reframe worth carrying: the recoverable money is usually in how you serve, not in what you charge. Repricing is a negotiation you might lose. Removing 320 unnecessary deliveries is an operational change the customer may not even notice.

Why experienced managers still get this wrong

These are structural failures, not careless ones, and they persist in well run finance teams.

1

They stop at gross margin because that is where the system stops.
The ERP traces product cost to the invoice and everything else to a cost centre. Customer profitability is not hidden, it is simply never assembled, and no report will ever surface a number nobody has built.
2

They allocate cost to serve as a flat percentage of revenue.
This is worse than not allocating at all, because it launders the problem. Spreading service cost evenly guarantees the heaviest user is charged the average, which is exactly the account you are trying to detect. Allocate on the driver, meaning orders, drops, and hours, or do not allocate.
3

They read revenue concentration as security.
The largest account is treated as the one that cannot be risked. Often it is the one contributing least per dollar of revenue, and the fear of losing it is what funded every concession that made it unprofitable in the first place.
4

They assume the fix has to be price.
Price is the most visible lever and the most confrontational. In most portfolios the service model holds more recoverable value and meets far less resistance, because you are removing cost the customer was never willing to pay for.
5

They never test the revenue hole they are afraid of.
Losing 1,200,000 dollars of revenue sounds catastrophic and is the reason nobody raises the subject. Losing 10,000 dollars of contribution, while releasing 320 deliveries of capacity, is a different sentence describing the same event. The analysis exists to let you say the second one out loud.

Putting it to work tomorrow

You do not need a costing system. You need one spreadsheet and a willingness to be approximately right.

Build it for the top 20 customers only. They will carry most of the revenue and nearly all of the distortion. A full portfolio rollout is how this work dies in committee.
Use 5 drivers and defensible rates. Orders, drops, returns, service hours, and receivable days. Rates within 20% are sufficient, because the answers differ by multiples. If you want the arithmetic done for you, the cost-to-serve calculator takes those 5 drivers and returns the contribution, with both levers priced.
Rank by contribution percentage, not by revenue. Then look at the gap between the two rankings. That gap is the agenda for the next commercial meeting, and it is usually the first time anyone has seen it.
Take the service lever before the price lever. For each loss-making account, cost the operational change first. Reprice only where the service model genuinely cannot move.

Know the limits of this analysis

Cost to serve is a lens, not a verdict. It misleads in 4 specific ways. If the driver rates are guessed rather than measured, you have moved opinion around and added decimal places. If a customer carries strategic value the model cannot see, such as reference status, market access, or a volume commitment that keeps a plant economic, the number is one input among several. If you treat fixed costs as though they vanish when a customer leaves, you will overstate the gain from an exit and end up with the contribution gone and the cost still there. And a single period is a snapshot, not a trajectory: a growing account absorbing setup cost looks identical to a declining one consuming service. Use it to find where value leaks, then decide with judgment.

The principle worth keeping

Revenue is not an achievement. It is a claim on your capacity, and some claims cost more than they pay.

"Before you fight to keep an account, work out what it contributes after everything it consumes. Some revenue is worth defending. Some is worth repricing. And some you should quietly want to lose."

Free Excel Model: Margin Bridge & Price/Volume/Mix

Once you know which accounts leak value, this separates an actual margin movement into price, volume, mix, and cost, so you can show how much of the period a single customer explains.

Download Free ↓