
A cost to serve analysis tells you what it really costs to deliver your product to a specific customer, channel or order, beyond what it cost to make. You subtract that cost from the gross profit each account produces, and the accounts that look healthy on a P&L often turn out to be the ones quietly draining cash. This guide shows you how to build the model, what to put in it, how to read it and where it goes wrong.
TL;DR
- Cost to serve (CTS) captures the costs that occur after a unit is made: fulfillment, freight, customer support, returns and account management. It lets you see profit by customer, channel or product, not just in aggregate.
- The core formula is simple: CTS equals the sum of each cost driver rate multiplied by the activity volume a customer consumes.
- Start with one question and one unit of analysis, such as wholesale accounts or sales channels. A rough model on good data beats a perfect model that never ships.
- Results are only as good as the least granular input, so document every allocation choice as a business decision.
- The output is a ranked list of accounts. Reprice, redesign the service or exit the tail that loses money.
Table of Contents
- What cost to serve actually measures
- Cost to serve versus COGS and total cost
- What to include in your model
- How to calculate cost to serve, step by step
- A worked example
- Reading the results
- Where cost to serve models go wrong
- Our perspective
- Get a founder-side view of your numbers
- FAQ
- Sources
What cost to serve actually measures
Cost to serve is the total cost of serving a particular customer, segment, channel or service, not just the cost of the product itself. It is an input. The output is customer profitability, which is what remains once you net cost to serve against the revenue that customer brings in.
Why does this matter? Traditional accounting spreads overhead evenly or ignores it at the customer level. Gartner found that conventional practices neglect to allocate the cost of supporting customers and products because of the complexity those activities create. The result is that an account that demands custom packaging, split shipments and weekly emails looks identical on paper to one that places a clean pallet order twice a year.
CTS fixes that by assigning costs based on what each customer actually consumes. Two accounts with the same revenue and the same gross margin can have wildly different real profit.
Cost to serve versus COGS and total cost
Founders often ask whether cost to serve is just another name for COGS. It is not. COGS tells you what it costs to produce a unit, and it is applied roughly the same to every unit. Cost to serve tells you what it costs to get that unit to a specific customer and support the relationship around it. COGS sets the floor for your price. CTS tells you whether a given account earns its keep above that floor.
There is also a boundary question: where does CTS stop? APQC, the benchmarking body, defines total cost to serve as all costs incurred after the sale is made. That covers warehouse operations, pick, pack and ship labor and materials, outbound freight, order processing, order-related customer service and returns processing. It explicitly excludes COGS, sales and marketing, product development and general and administrative costs.
That is a clean, benchmarkable definition, and it is a good starting point. For a consumer brand, though, you will likely want a second view that adds customer acquisition cost, trade spend and account management, because those are the costs founders feel in cash. Keep the two views separate and label them, so you never compare your wider number against a narrow benchmark.

Speaking of benchmarks: APQC reports a median cost to serve of $56.00 per $1,000 of revenue across a sample of 199 companies. That is about 5.6% of revenue. It spans many industries, so treat it as a rough reference point and not a target for a food, beauty or apparel brand.
What to include in your model
A workable model has three layers of cost.
Direct customer costs. Anything you can tie to one account without allocation: outbound freight billed to you, pallet and carton charges, retailer-specific labeling, chargebacks and deductions, payment processing fees on a channel, marketplace fees.
Activity-based service costs. The work your team does because of that customer. Order processing, inventory handling and picking, customer support contacts, returns processing and reverse logistics, and account management. These are the costs that traditional P&Ls bury in operating expense.
Allocated shared costs. IT, finance and facilities that support all of the above. Allocate these last, and keep them visibly separate, because they are the easiest to argue about and the least likely to change a decision.
For a DTC and wholesale brand, the categories that most often change the answer are:
- Returns and reverse logistics, especially in apparel and footwear
- Retailer deductions, compliance fines and markdown allowances
- Split shipments and small-order handling
- Customer service time per order, which varies enormously by channel
- Payment, marketplace and platform fees
One warning from the logistics side of the industry: some models leave out returns handling, reverse freight, failed delivery reattempts, refunds processing and write-offs on returned goods. Leaving them out flatters exactly the customers most likely to be unprofitable. Put them in.
How to calculate cost to serve, step by step
1. Pick the unit of analysis
Decide what you are trying to rank: customers, sales channels, product categories, fulfillment methods or individual orders. Choose one. If you sell to ten retailers and through three DTC channels, start with the thirteen accounts or channels and nothing finer.
2. Define scope
Agree what is in and out before you touch a spreadsheet. Gartner’s six-step model for supply chain leaders starts with mapping cost components and agreeing scope, then linking each component to a cost driver, modeling an approximation of actual cost per activity, revealing profitability at the transaction level and using the findings to improve performance. Note the word approximation. You are modeling, not auditing.
3. Build cost pools
Group your post-sale costs into pools: warehouse labor, outbound freight, customer support, returns processing, account management, chargebacks and so on. Pull the totals from your general ledger and your 3PL invoices.
4. Choose a driver for each pool
A driver is the activity that causes the cost. Warehouse labor might follow order lines picked. Customer support follows tickets. Freight follows shipments or pounds. Account management follows hours or, if you lack time data, the number of purchase orders.
5. Calculate the rate
Divide the pool cost by the total driver volume. If customer support costs $120,000 a year and your team handled 24,000 tickets, the rate is $5 per ticket.
6. Apply the rates to actual consumption
Multiply each customer’s actual volume by each rate and add up the pools. Then subtract the total from that customer’s gross profit.
A worked example
The numbers below are hypothetical, built to show the mechanics for a brand with roughly $20M in revenue. Three wholesale accounts look similar at the gross margin line:
| Account | Revenue | Gross profit | Cost to serve | Profit after CTS | Margin after CTS |
|---|---|---|---|---|---|
| Account A | $1,200,000 | $540,000 | $150,000 | $390,000 | 32.5% |
| Account B | $800,000 | $360,000 | $190,000 | $170,000 | 21.3% |
| Account C | $500,000 | $225,000 | $240,000 | -$15,000 | -3.0% |
Every account earns a 45% gross margin. Account A consumes 12.5% of revenue in service cost. Account C consumes 48%, driven by small frequent orders, compliance fines, heavy deductions and a buyer who emails daily. On the standard P&L, C looks like a healthy half-million in sales. After cost to serve, it loses money, and the cash it absorbs could fund growth elsewhere.
Reading the results
Once you have a profit figure per account, rank them from most to least profitable after cost to serve and plot the cumulative profit. This is often called the whale curve. It typically reveals a profitable core of accounts that generate more than 100% of total profit, a roughly break-even middle, and a tail of loss-making accounts that pull the total back down. The technique appears in the cost accounting literature on time-driven activity-based costing, where the authors lay out how to act on unprofitable customers.
The curve is only a picture. The decision work comes next, and there are four moves, roughly in order of how gentle they are:
- Reprice. Add a minimum order, a small-order fee or a freight threshold. Most unprofitable accounts are fixable with price.
- Redesign the service. Consolidate shipments, move to a standard pallet configuration, tighten the returns policy, or route low-value orders to a cheaper fulfillment method.
- Invest in the process. If support contacts drive the cost, fix the cause: better product pages, clearer sizing, proactive shipping notices. If you train a team to cut handling time, calculating training ROI gives you a way to show finance whether the program pays back.
- Exit or deprioritize. If an account will not change and cannot reach a positive margin, stop feeding it growth capital.
The cash-flow link is direct. Every dollar of loss-making volume ties up inventory, working capital and team hours. Removing or repricing it releases all three. If you want to go deeper on the account level, our guide to customer profitability analysis covers the follow-through, and channel profitability does the same at the channel level.
Where cost to serve models go wrong
A model can be precise and still be wrong. Four failure modes come up again and again.
Data grain limits the answer. Your results can only be as detailed as your least granular input. If your 3PL bills you one blended handling fee, you cannot see which accounts cost more to pick. Different allocation choices can also change which orders look expensive. Make the cost drivers transparent and configurable, and write them down as business decisions, not facts of nature.
Over-engineering. Time-driven activity-based costing promises more accurate data at lower cost than older approaches, but an academic comparison of the method notes that data collection can become prohibitive when you must collect information on every subtask, and that accuracy depends on being able to trace resources to activities. For a founder-led brand, an aggregate model built on a handful of drivers is usually the right call.
Missing costs. As above, returns, reattempted deliveries and write-offs get dropped. So do chargebacks, which sit in a revenue contra account in many ledgers.

Treating it as a one-time project. Product mix, channel mix and volume move constantly. Rerun the model at least quarterly, and any time you sign a major account or change 3PL.
Our perspective
Our approach at Commerce Catalyst is to build the smallest model that can change a decision, then help you act on it. That means pricing conversations, shipping terms, channel priorities and, when needed, hands-on implementation support. A diagnostic is only worth the decisions it moves.
If you take one thing from this guide, make it this: do not wait for perfect data. Build a first version this month, find your tail, and fix the worst account first.
Get a founder-side view of your numbers
If you would rather not build the model from scratch, our DTC Operator Diagnostic is a $197 one-off review that helps you identify the operating constraints holding back your profit and cash flow. For a working session on your own accounts and channels, you can book a Founder Hour at $500 per hour.
If the analysis points to a larger restructuring, the 90-Day Profit Sprint for Consumer Brands is built for exactly that. Before an investor conversation, our guide on how to demonstrate a path to profitability shows how this work feeds the story.
FAQ
How do I calculate cost to serve?
Identify your post-sale cost pools, such as freight, warehouse labor, support and returns. Choose a driver for each, divide the pool cost by total driver volume to get a rate, then multiply each rate by the activity a customer consumes. Add the pools and subtract the total from that customer’s gross profit.
Is cost to serve the same as COGS?
No. COGS covers the cost of producing a unit and is applied evenly per unit. Cost to serve covers what it takes to deliver that unit to a specific customer and support the relationship, so it varies by account, channel and order pattern.
What should a cost to serve model include?
At minimum: warehouse and fulfillment, outbound freight, order processing, customer service related to orders, and returns processing. APQC’s standard definition excludes COGS, sales and marketing, product development and general and administrative costs. Many brands run a second, wider view that adds trade spend and account management.
How often should I rerun the analysis?
Quarterly is a sensible default for most consumer brands, plus whenever you add a major account, change fulfillment partners or shift your channel mix. The model reflects a moment in your operations, and that moment changes.
How is cost to serve different from activity-based costing?
It draws on the same logic but is lighter. Cost to serve tends to work from aggregate analysis with a blend of cost drivers, while full activity-based costing traces more detailed activities and demands more data. For most founder-led brands, the lighter approach is enough to change decisions.
Sources
- Gartner: supply chain leaders should implement a cost-to-serve model
- APQC: cost to serve per $1,000 revenue
- Transforming unprofitable customers, time-driven activity-based costing
- Academic comparison of TDABC and ABC
- Training ROI calculation
Recommended
