- 1. What are you actually trying to measure?
- 2. The four drivers that matter
- 3. How to set the rates
- 4. The three-week plan
- 5. Where the rough version misleads you
- 6. What to do when the data is missing
- 7. When you actually need the full version
- 8. Questions people ask
The question is usually asked as how do we implement activity-based costing, and that framing is what kills it. Implementation implies a system, a project team and a charter. The real question is narrower and answerable: how do I get a cost-to-serve number good enough to decide with, without starting something that will not finish.
Three weeks, four drivers, one spreadsheet. That is the honest answer, and the reason it works is that you are not pricing a government contract. You are sorting accounts into profitable, marginal and value-destroying, and the distance between those three states is wide enough that an approximate number resolves it. This is the practical layer underneath the customer profitability method.
1. What are you actually trying to measure?
The cost of serving one customer that differs from the cost of serving another. Not total cost, not fully absorbed cost, not a number that ties to the general ledger to the dollar. The difference, because the difference is the only part that changes a decision.
This distinction saves you most of the work. A cost that every customer consumes identically in proportion to revenue can stay allocated on revenue, because allocating it more precisely will not change anybody’s ranking. Only the costs that vary by customer behaviour need a driver, and there are fewer of those than the textbook suggests.
Watch a related video: Contribution Margin vs P&L: The $4M Mistake.
2. The four drivers that matter
Four, in every industrial business I have worked in, and together they capture most of the cost difference between customers. Add a fifth only when you can name the specific decision it would change.
Order lines
The workhorse. Order entry, picking, packing, documentation, invoicing and a share of planning all scale with lines rather than with dollars. Forty orders of one unit and one order of forty units produce identical revenue and wildly different cost, and this driver is what makes that visible.
Shipments
Separate from lines, because one order can ship five times. Freight, handling and the administration attached to each delivery scale here. In businesses with release schedules or blanket orders this is frequently the largest of the four.
Setups and changeovers
The production driver. Every distinct item a customer buys in small quantities consumes a setup that a high-volume item amortizes across thousands of parts. Attribute setups to the customers whose order patterns caused them.
Exceptions
Expedites, engineering changes, quality escalations, returns and credit notes. Count events rather than measuring cost, then apply an average. This driver is always the most concentrated and it is the one that produces the surprises.
3. How to set the rates
Take the total annual cost of the function, divide by the total annual count of the driver, and accept the result. That is it. Order entry costs you the loaded cost of the people doing it plus their systems, divided by the number of order lines they processed last year.
The instinct at this point is to refine, and the instinct is wrong. Refining the rate changes every customer’s number by roughly the same proportion, which leaves the ranking unchanged. You are not trying to be right about the rate. You are trying to be right about which accounts consume ten times the average of it.
Round the rates aggressively. A number like $14 per order line is easier to explain, defend and remember than $13.87, and the difference is invisible in the conclusion.
4. The three-week plan
Five steps across three weeks, run by one analyst with access to people who know how the business works. Anything longer than this is a project, and projects in this category have a poor completion record.
- Days one to three: pull the revenue and gross profit base. By customer and product family, twelve months, with standard cost applied. Accept whatever your standard costs say for now and note which families look suspect.
- Days four to seven: count the drivers. Order lines, shipments, setups and exception events, by customer, for the same twelve months. Most of this is in your ERP. The exception count usually is not, and you will have to ask people.
- Week two: build and apply the rates. Four rates, four multiplications, one column of allocated cost per customer. Subtract from gross profit and you have cost-to-serve profitability.
- Week three, first half: test the outliers. Take the ten accounts that moved most and walk each one with the salesperson who owns it. You are looking for cases where the model is wrong, and you will find two or three.
- Week three, second half: fix what the test found and stop. Do not start a second iteration. The model is now good enough to act on and further precision will delay every decision that depends on it.
5. Where the rough version misleads you
Three places, and knowing them in advance is what separates a useful approximation from a confident error.
It misleads when a customer’s cost is driven by something outside the four drivers. A single account requiring a dedicated inspector, a bonded warehouse or a unique packaging line carries a cost no per-line rate will capture. These are rare and obvious, and you handle them by adding the specific cost directly rather than by adding a fifth driver for everybody.
It misleads on shared capacity. If a setup would have happened anyway for another customer, attributing its full cost to the small account that also uses that item overstates the case for killing it. Check the outliers for this specifically, because it is the most common way a rough model condemns something it should not.
If a setup would have happened anyway for another customer, attributing its full cost to the small account overstates the case for killing it. Check every outlier for shared capacity.
And it misleads when standard costs have drifted. Your allocation can be perfect and your answer still wrong if the underlying product cost is four years stale. Note the families where standard and actual diverge most, and treat that as a separate finding rather than as a reason to stop.
6. What to do when the data is missing
Estimate it, label the estimate, and move on. Missing data is the single most common reason this exercise stalls, and in nearly every case the missing item is the exception count, because nobody has ever been asked to record expedites by customer.
Reconstruct it from whatever proxy exists: premium freight records, overtime authorizations, schedule change logs, or a planner’s memory over a quarter. A planner asked which five accounts cause the most disruption will give you a list in ninety seconds, and that list will be right. Use it, mark it as an estimate, and start counting properly from this month forward.
7. When you actually need the full version
Three situations, and outside them the rough version is not a compromise, it is the correct tool. You need the full implementation when cost-to-serve has to be auditable for contractual or regulatory reasons, when your product mix is so configured that the four drivers genuinely do not capture the variation, or when the output will drive pricing across thousands of transactions automatically rather than informing a few dozen decisions.
Even then, build the rough version first. It tells you whether the full one is worth commissioning, it produces the driver definitions the full system will use, and it gives you a result to sanity-check the expensive answer against.
8. Questions people ask
Will finance accept a model that does not tie to the ledger?
Frame it as a management view rather than a financial one and most will. It is not intended to replace anything they produce, and it should not be reconciled to the penny. If reconciliation becomes a condition, you are back in a six-month project.
How often should we rerun it?
Annually, with the same four drivers and refreshed rates. The interesting information after year one is which accounts moved and why, which takes an afternoon rather than three weeks.
Should we tell sales before or after?
Before, and have a commercial person in the room from day one. A model delivered to sales as a conclusion will be challenged on method. A model built with them will be challenged on specifics, which is what you want, because the specifics are where it is actually wrong.
What if the answer is that almost everything is profitable?
Then either your allocation is too generous or you genuinely have a clean book, and the way to tell is the shape of the curve rather than the count. If the cumulative profit curve never exceeds one hundred percent, nothing is subsidizing anything, and that is a real and unusual finding worth knowing.
The question was how to get activity-based costing done without a six-month project. The answer is that four drivers, roughly rated, applied over three weeks by one person, will move nearly every decision the precise version would have moved, and it will do it in time to matter.
About the Stagnation Assassin
Todd Hagopian is a Fortune 500 transformation executive who has generated $3B+ shareholder value across five turnarounds at Berkshire Hathaway, Illinois Tool Works, and Whirlpool. He is VP of Product Strategy and Marketing. Known as The Stagnation Assassin, he is the author of two published books: The Unfair Advantage: Weaponizing the Hypomanic Toolbox and Stagnation Assassin: The Anti-Consultant Manifesto. His blog is published in 15+ languages and read by operators worldwide. Bring him to your stage via his speaking page or connect with him on LinkedIn. His newest book is 10-Minute Transformation (February 23, 2027).