Cost to Serve Analysis: How to Find Which Customers Cost You Money
Learn how to run a cost to serve analysis: which costs to include, how to calculate profit by customer, and how to cut cost to serve without losing good accounts.

Cost to Serve Analysis: How to Find Which Customers Cost You Money
A cost to serve analysis tells you what it really costs to serve each customer, product or channel, and what is left of the revenue once you subtract it. You take a customer’s net revenue, subtract the product cost, then subtract every activity that customer triggers: order handling, support calls, special deliveries, account management time, returns. The result is profit by customer, which is often very different from the gross margin your accounting system reports.
This guide shows how to do it with the data a small or mid-sized business already has, what to include, how to read the results, and how to act on them without damaging the accounts that matter.
RaremindFind the Gaps Behind Customer CostsRaremind’s Money Map helps identify lost revenue, excessive costs and process inefficiencies, without pushing unnecessary technology.Explore Raremind’s approach
TL;DR
Cost to serve analysis assigns selling, delivery, service and support costs to the customers that cause them, so you see profit by customer instead of average margin.
Start with 5 to 10 activities and a handful of cost drivers. A rough, defensible model you finish beats a perfect one you abandon.
Time-driven activity-based costing (TDABC) is the practical method: a cost per minute of capacity multiplied by the minutes each activity takes.
Act on the results by changing terms, minimums, service levels and processes for specific segments, not by dropping customers on impulse.
Fix the process before buying software. Many cost-to-serve problems are workflow problems that a new tool would only speed up.
Table of Contents
What cost to serve means and why gross margin misleads
Who needs it, and when it is worth the effort
What to include: costs, activities and drivers
How to calculate cost to serve, step by step
Reading the results
How to reduce cost to serve without losing good customers
Common mistakes
Our perspective: start with how the money is made
Want a second pair of eyes on your numbers?
A simple 30-day plan
Sources
What cost to serve means and why gross margin misleads
Cost to serve (CTS) is the cost of everything you do to win, deliver and keep a customer’s business, beyond making the product itself. Customer profitability analysis is the outcome: once CTS is attached to each customer, you can rank them by what they actually earn you.
The reason this matters is that most accounting systems were built around products. According to the Open University’s material on advanced management accounting, traditional systems focus on product costs rather than customer-related costs, which can lead a business to favor unprofitable customers and overlook profitable ones. The same source notes that selling, distribution, service and support costs can reach up to 60% of sales value, so ignoring them leaves a big part of the picture blank.
Two common shortcuts fail for the same reason. Measuring only gross margin (sales minus product cost) ignores service effort. Spreading functional costs as an average treats a customer who orders once a quarter in full pallets the same as one who sends 30 small orders a week and calls support after each. The Open University source describes both approaches as inadequate because they miss the large variability in service costs between customers.
The pattern shows up repeatedly. A Harvard Business Review article reports that companies often find a small share of customers, around 20%, generating a very large proportion of profits, in some cases 150%, while “profit drains” erode them and “profit deserts” consume resources for minimal return. Your numbers will differ, but the shape is worth testing in your own data.

Who needs it, and when it is worth the effort
You do not need a supply chain with 40 warehouses. The analysis pays off whenever customers differ in how much work they create. Typical signs:
Some customers negotiate custom terms, reports or delivery windows, and others take the standard offer.
Your team complains about a few accounts that eat their week.
Revenue is growing but profit is flat.
Prices were set years ago and never tied to the effort behind each account.
You sell through several channels (direct, distributors, online) with unclear differences in cost.
It works in service businesses too. A consultancy, agency or software company can calculate cost to serve from support tickets, onboarding hours, meetings and custom work per client.
What to include: costs, activities and drivers
Three terms do most of the work.
Activities are the things people do for customers: process an order, handle an invoice query, run a monthly review call, arrange a special shipment.
Resources are what those activities consume: staff time, vehicles, systems, space.
Cost drivers are the measurable counts that tell you how much of an activity each customer used: orders placed, invoices raised, tickets opened, deliveries made, complaints logged.
The Open University source suggests including attributable costs such as order processing, key-account time, transport and support, with drivers such as invoices raised and customer complaints. The logic behind it is activity-based cost management, which, as the Institute of Management Accountants explains, assigns resources to activities and activities to cost objects based on use, recognizing the causal link between drivers and activities.
A practical starting list for a typical small or mid-sized business:
Area | Example activities | Typical driver |
|---|---|---|
Sales | Quoting, negotiation, account reviews | Quotes issued, hours of account time |
Order handling | Entry, confirmation, changes | Orders, order lines, change requests |
Fulfillment | Picking, packing, delivery, installation | Shipments, drops, miles, delivery hours |
Billing and collections | Invoicing, disputes, chasing late payments | Invoices, disputes, days late |
Support | Questions, training, fixes, returns | Tickets, calls, returns |
Finance cost | Extended payment terms | Revenue multiplied by days outstanding |
Keep fixed overhead that no customer decision changes (the lease, the owner’s salary) out of the customer allocation, or put it in a separate layer so it does not drown the signal. Allocation without a real usage measure is risky: the IMA’s guidance on cost allocation warns that allocations lacking a direct measure of activity use can overstate or understate costs for products with different service requirements.
How to calculate cost to serve, step by step
Step 1: Choose the unit and the question
Decide what you are ranking: customers, segments, products or channels. For most small businesses, start with customers or customer groups. Write one sentence for the question, such as “Which 20 accounts earn us the least per hour of effort?” A clear question keeps the model small.
Step 2: Get true revenue per customer
Use the price actually paid, net of discounts, rebates, credits and returns, not list price. The Business Development Bank of Canada’s profitability guide frames it this way: profit is the true price paid minus the total cost to serve that customer, summed across every activity performed for them.
Step 3: Map activities and pick drivers
List the activities in the table above that apply to you. Talk to the people who do the work, not just finance. Ask what customers trigger most often and which exceptions waste the most time. Five to ten activities is plenty for a first pass.
Step 4: Put a cost on each activity
This is where most models stall. Traditional activity-based costing relied on detailed employee surveys, and Robert S. Kaplan has noted that such implementations are often abandoned because of high costs and employee irritation. His alternative, time-driven activity-based costing, uses managerial estimates of the cost per time unit of a resource and the time each activity needs, rather than costly, time-consuming surveys.
In practice:
Take the fully loaded monthly cost of a team (salaries, benefits, tools, space).
Divide by the practical minutes available (not the theoretical ones, since meetings, training and idle time are real).
That gives a cost per minute.
Estimate minutes per activity, then adjust with a few rules. Kaplan and Anderson’s time equations show how characteristics of an order or customer make processing time vary: a custom order takes longer than a standard one, and a new customer takes longer than an existing one.
The same minute-based thinking underlies time and materials billing, where labor hours and equipment use have to be captured accurately to be charged or analyzed. If you already track hours per job, you have much of the data for a cost-to-serve model.
Step 5: Assign costs to customers
Multiply each customer’s driver counts by the cost per driver. Add the product cost and any financing cost of slow payment. Subtract the total from net revenue. You now have profit by customer.

A worked example (illustrative numbers)
Imagine a wholesale distributor with two customers who each buy €100,000 a year at a 30% gross margin, so €30,000 of gross profit each.
Customer A | Customer B | |
|---|---|---|
Net revenue | €100,000 | €100,000 |
Gross profit | €30,000 | €30,000 |
Orders per year | 12 | 150 |
Order handling at €25 per order | €300 | €3,750 |
Deliveries at €60 each | €720 | €9,000 |
Support and queries at €40 per ticket | €400 | €2,800 |
Late-payment financing | €200 | €1,500 |
Total cost to serve | €1,620 | €17,050 |
Profit after cost to serve | €28,380 | €12,950 |
Both look identical on a gross margin report. After cost to serve, one is more than twice as valuable. These figures are made up to show the mechanics, and your cost per order will differ, but the logic holds for any business where order frequency and service effort vary.
Reading the results
Sort customers from most to least profitable and look at three things.
The top group. Who are they, and what do they have in common (size, channel, order pattern, industry)? Protect them.
The middle. Often the largest opportunity, because small changes in terms or process move many accounts.
The bottom. Some are loss-making. Ask why: price, volume, or avoidable service demands?
Measuring at customer level also gives you new metrics. The Harvard Business School Working Knowledge piece notes this allows measures such as the percentage of unprofitable customers or the total dollars lost in unprofitable relationships. Track those quarterly.
Treat the first run as a conversation starter, not a verdict. Check the largest outliers with the people closest to them. A customer who looks unprofitable may be a reference account, a door to a new market, or simply mis-coded in your data. The value of the exercise, as one management accounting paper puts it, is that results replace intuitive impressions with fact-based information that can guide you to add value for profitable customers, improve less profitable ones, and stop profit erosion.
How to reduce cost to serve without losing good customers
Cutting cost to serve does not mean firing customers. The goal is to match service to value. Options, roughly from least to most disruptive:
Fix internal waste first. Rework, duplicate data entry, manual re-keying, chasing missing information. These raise cost for every customer and none of them improves what the customer receives.
Change the process for small accounts. Harvard Business Review describes adjusting delivery frequencies for small accounts, refining restocking policies, and educating customers on efficient product selection to lower administrative and support burdens.
Introduce tiers. Standard service for everyone, premium service at a premium price. Minimum order values, delivery fees under a threshold, paid rush options and paid custom reports all make the extra effort visible.
Reprice. If a segment is structurally unprofitable at current prices, raise them, or restructure what is included.
Target your market. The same HBR piece stresses that success requires focusing on specific, defensible segments instead of serving everyone with a one-size-fits-all model.
Let go, last. Only after the other steps fail, and with a plan for the revenue you lose.
For more ideas on the broader cost picture, our guide to cost reduction strategies for small businesses puts operations ahead of cuts to headcount or tools.
The accuracy of your costs also affects everything downstream. IMA guidance on activity-based costing notes that accurate cost information is critical for profitable pricing, product-mix planning and sales and marketing strategy, including how sales compensation is structured. If your salespeople are paid on revenue alone, a customer-level view will show where that incentive pushes them toward unprofitable deals.
Common mistakes
Starting with software. A tool automates a model you have not designed. Build the logic in a spreadsheet first. If a tool is needed later, you will know exactly what it must do. Our AI readiness assessment explains why mapping revenue workflows comes before buying.
Allocating by revenue. Spreading costs as a percentage of sales brings you back to the average-cost problem.
Aiming for precision. A model within 10% of the truth that people trust is more useful than a perfect one nobody finishes.
Skipping the people. Management resistance is common, as with any organizational change. The research notes that success requires early involvement and buy-in from the managers and employees affected. Share the method before you share the rankings.
One-off analysis. Costs and customers change. Refresh the model at least once or twice a year and after any major pricing or process change.
Our perspective: start with how the money is made
We approach cost to serve from the revenue side, because that is where we begin with every client. Before anyone talks about tools, we map how a business actually earns money: who pays, for what, through which steps, and where effort or cash leaks out along the way. We call that map the Money Map, and a cost-to-serve analysis is one of the sharpest lenses you can hold up to it.
What we see again and again is that high cost to serve is rarely about the customer. It is about a workflow: the quote that gets re-typed three times, the order that needs a phone call because the form is unclear, the invoice that is wrong because two systems disagree. Those costs hit every account, but they hit the awkward ones hardest. Fixing the workflow often reduces cost to serve across the board, with no change to what customers get.
The second thing we see is the urge to buy something. A new platform feels like progress. But software that automates a messy process gives you a faster mess. We do not sell software or take vendor commissions, so we can say plainly when a spreadsheet, a changed policy or a simple automation is enough, and when a tool is justified. If you want a deeper look at how that differs from conventional engagements, we wrote about strategic advisory versus traditional consulting.
Our advice: run your first cost-to-serve model in a week, not a quarter. Pick your top 30 customers, list five activities, estimate minutes, and see what the ranking tells you. You will learn more from a rough answer than from months of planning.
— Mirko
Want a second pair of eyes on your numbers?
If you would rather not build this alone, our Your Digital & AI Office plan is a monthly advisory retainer at €995 per month. We map how your business makes money, find where revenue and cost leak, and advise on what to fix, automate or leave alone, with no software to sell you. Bigger, hands-on initiatives are scoped separately as larger projects.

You can see how it works and get in touch on the Raremind site.
A simple 30-day plan
Week | Focus | Output |
|---|---|---|
1 | Pick the customers and activities, pull net revenue and order data | Customer list with revenue, orders, tickets |
2 | Estimate time per activity with the people who do it, compute cost per minute | Cost per activity |
3 | Assign costs to customers, rank by profit | First profitability ranking |
4 | Review outliers with sales and operations, choose three changes | Action list with owners and dates |
What counts is the fourth week. A ranking that sits in a spreadsheet changes nothing. Pick three actions: one process fix, one pricing or terms change, one segment decision, and measure again in a quarter.
FAQ
How do I calculate cost to serve?
List the activities each customer triggers, such as order handling, delivery, support and invoicing. Multiply each customer’s usage of those activities by the cost per activity, then subtract the total and the product cost from the net revenue the customer paid. The time-driven method makes this easier: cost per minute of a team multiplied by the minutes each activity takes.
What is the difference between cost to serve and customer profitability analysis?
Cost to serve is the cost side: what it takes to sell to, deliver to and support a customer. Customer profitability analysis is the full result: net revenue minus product cost minus cost to serve. You need the first to produce the second.
Do small businesses need special software for a cost-to-serve model?
No. A spreadsheet with revenue, order counts, ticket counts and a few cost rates is enough for a first model. Consider software only when the logic is proven and the manual updating becomes the bottleneck.
How often should I update the analysis?
Refresh it once or twice a year, and again after a price change, a new channel or a large process change. Customer behavior and cost rates shift, so an old ranking slowly stops being true.
Sources
Challenges in advanced management accounting, The Open University
How to Create a Winning Post-Pandemic Business Model, Harvard Business Review
Activity-based cost management, Institute of Management Accountants
Using Activity-Based Costing, Institute of Management Accountants
Customer Profitability Management, Strategic Management Accounting