Every company has an account that everyone treats as untouchable. Run a real customer profitability analysis on it and you may not like what you find.
It’s the biggest one. It’s been there the longest. A quarter of revenue rides on it, and everybody in the building knows you don’t lose it. When the schedule is tight, that account goes first. When they change a date, you move everything else to cover it.
On the last one of these we ran, that account turned out to be the third-most profitable customer in the book. Not the third largest. The third best, behind two accounts a fraction of its size that nobody in the company ever thought about.
Nothing was wrong with the pricing. The gross margin on that account looked normal. Gross margin just wasn’t the right number to look at.
Gross margin is where the analysis starts, not where it ends
Gross margin answers a narrow question: after the direct cost of delivering the work, what’s left? It’s a real number, and it’s the right number for your financial statements. For management purposes, though, it leaves out most of what actually matters, because it only counts the costs that are easy to attribute.
The costs that decide whether a customer is good are mostly the ones nobody can cleanly assign. They’re real, they’re substantial, and because no one can pin them to a specific account, they land in overhead, where every customer absorbs an even share, including the ones who never caused them.
Overhead allocation is where difficult customers become invisible. Their costs get quietly paid for by the customers who didn’t create them.
None of that is fraud or error, and your accountant did nothing wrong. A reporting framework simply can’t answer a management question, because nobody designed it for that job.
This is managerial analysis, not statutory reporting. It doesn’t belong on your financial statements, GAAP doesn’t require it, and no auditor will ever ask you for it. It’s for you, and that changes how you build it: you’re optimizing for a decision, not for compliance.
What cost to serve actually includes
Here’s what you have to strip out of gross margin before you know what a customer is really worth.
Rework and warranty. Some customers generate more of it than others, sometimes because of their spec, sometimes because of the pace they demand. Companies rarely track this by customer, and it’s almost never small.
Expedite cost. Overtime, rush freight, buying material on the spot because the schedule slipped. Each instance feels like a one-off. Add them up by customer over a year and a pattern shows up fast.
Scheduling drag. The hardest one to see, and often the largest. A customer who changes dates costs you more than their own inefficiency; the disruption ripples through everything else you moved to cover them, and those costs land quietly on other customers’ jobs.
Working capital. A customer who pays in 60 days is not the same customer as one who pays in 15, even at identical margin. You’re financing them. That has a real cost: your cost of capital, times the balance, times the time, a number most owners have never once run for a single account.
Sales and account management drag. Some accounts eat enormous amounts of your best people’s time. That’s the most expensive resource in the building, and it’s the one nobody puts on a job cost.
Run those five through a large account and it’s routine to find the real margin is several points lighter than the statement shows. The account may still hold up, or it may be underwater. Either way, you don’t currently know, because the number that would tell you doesn’t exist anywhere in your accounting system.
How to Build a Real Customer Profitability Analysis
Five steps.
1. Define the unit. Customer, job, crew, lane, SKU. Pick the one you actually manage by. Most companies pick the wrong unit here and end up with a model that’s technically correct and operationally useless.
2. Get direct cost right first. Before you allocate anything, make sure you’ve actually coded the costs you can attribute directly. In most books we review, a meaningful chunk of directly attributable cost sits in a general bucket simply because nobody set up the coding for it.
3. Pick allocation drivers that reflect causation, not convenience. Allocating overhead on revenue is the default choice, and it’s the worst one, because it guarantees your biggest customer looks average by definition. Allocate on whatever actually drives the cost: touches, hours, changes, deliveries.
4. Add the working capital charge. Average balance times days outstanding times your real cost of capital. It’s one line, and it reorders the rankings more often than people expect.
5. Make it formula-driven and traceable. Every number needs to trace back to a transaction in under two minutes. If it can’t, you won’t trust it, and if you don’t trust it, you’ll go straight back to your gut, which is exactly what you were doing before you built the model.
What you do with the answer
The instinct is to fire the customer. That’s almost always the wrong move, and it’s usually a sign the customer profitability analysis stopped one step too early.
An account that’s marginal after cost to serve is still absorbing real overhead. Cut it and the overhead doesn’t disappear; it lands on whoever’s left, and those customers now look worse too. That’s how a company talks itself into a slow decline, one reasonable decision at a time.
Three real moves instead:
- Reprice, but not across the board. Target the specific thing driving the cost: change fees, expedite fees, terms. Most customers accept this when you tie it to a behavior they recognize, rather than roll it out as a general increase.
- Re-scope. Change what you deliver, not what you charge. Often the expensive part of an account is a service nobody ever priced, because it started as a favor.
- Re-sequence. If the cost is scheduling drag, the fix is operational, not commercial. It doesn’t require a single conversation with the customer.
The account we opened with is still a client today. We changed the terms and added a change fee tied to late schedule changes. That account moved from third to first in profitability within two quarters, and the customer never pushed back, because the fee tracked something they were actually doing, and they knew it.
“Biggest” and “best” are different words. Most companies have never checked whether they mean the same thing.
Frequently Asked Questions
Why doesn’t gross margin tell me whether a customer is actually profitable?
Gross margin only nets out the direct costs that are easy to attribute. The costs that actually separate a good customer from a bad one, rework, expedites, scheduling disruption, working capital, and account management time, are hard to attribute, so they land in overhead and spread evenly across every customer, hiding the true picture.
What costs should I include when I calculate true customer profitability?
At minimum, rework and warranty cost, expedite cost, the scheduling disruption you cause other jobs, a working capital charge based on days outstanding and your cost of capital, and the account management time your best people spend on the account. Most of these sit in overhead today, uncounted.
Should I fire an unprofitable customer once I find one?
Almost never as the first move. That customer is usually still absorbing real overhead; remove them and that overhead lands on the customers who remain, making those accounts look worse. Try repricing the specific behavior driving the cost, re-scoping what you deliver, or fixing the scheduling problem operationally before you consider cutting the account.
How is customer profitability analysis different from what’s in my financial statements?
It’s managerial analysis, not statutory reporting. GAAP doesn’t require it and no auditor will ask for it. It exists purely to help you make a decision, so you build it around the unit you actually manage by, not the categories your financial statements use.
What would a real customer profitability analysis find in your book?
The Growth Readiness Review builds one at the level your business actually runs: formula-driven, auditable, every number traceable back to a transaction. Three weeks, fixed fee, no retainer attached. You keep the model either way.
