- 1. Which variety do customers actually choose on?
- 2. Which variety did you inherit rather than choose?
- 3. What does one more option cost downstream?
- 4. Two kinds of customization, and only one of them scales
- 5. How far back can you standardize without touching the choice?
- 6. When does standardizing actually lose you a sale?
- 7. What if customization is the thing you sell?
- 8. The test to run on your own line
- 9. What do you do when sales and operations disagree?
You have been told twice this year that complexity is killing your margins, and twice that your competitors are winning deals because they will build what the customer asks for. Both statements come from people you trust, both are supported by examples from your own business, and neither one tells you what to do on Monday. So the catalog keeps growing, because growing is what happens when nobody can say where the line is.
There is a line, and it can be located. It does not sit at a SKU count, and it is not a philosophical position about standardization. It sits at the point where additional variety stops changing what customers choose, and everything past that point is cost you are paying for nothing. What follows is a nine-step method for finding that point on your own product line. If your immediate problem is a catalog that has already grown past it, the cutting method is in how to cut a product line without losing revenue.
1. Which variety do customers actually choose on?
Start by separating the options customers select from the options they merely accept. A customer who specifies a voltage, a connection size, or a material is choosing. A customer who takes whatever configuration the quote came back with is accepting, and the variety on that dimension is doing no commercial work at all.
The evidence is in your order history rather than in anyone’s opinion. For each option dimension, look at how the orders distribute. A dimension where customers spread across five choices is a dimension they care about. One where ninety percent of orders land on the same value, and the rest are scattered, is usually a dimension where you offered a choice and customers took the default.
Watch for the dimension that customers appear to choose on because your quoting process asks them to. If a form requires a value before the quote can be produced, customers will supply one, and the resulting distribution says more about the form than about their preferences.
Ask the sales team the same question separately, and expect the answers to conflict. Sales will name dimensions that come up in negotiations, which is real information about what gets discussed but not always about what gets decided. Where the order history and the sales view disagree, the useful move is to find the specific deals that prove the point: orders where the customer took an unusual value on that dimension and would not have bought otherwise.
Write the answer down as a short list. Most industrial product lines have two or three dimensions that genuinely drive selection and a long list of others that accumulated for reasons nobody remembers. That short list is the part of your variety that earns its cost, and everything else is a candidate for the rest of this method.
2. Which variety did you inherit rather than choose?
Every product line carries options that exist because a customer asked once, an engineer solved a problem locally, or a competitor had it in their catalog. None of those is a reason to carry an option now, and none of them will be defended by anybody if you ask directly.
Go through the option list and try to name the origin of each one. You will find three categories quickly: options with a current commercial reason, options whose reason has expired, and options nobody can explain at all. The third category is usually larger than the second, which surprises people and should not, because nothing in a normal product process ever removes an option once it exists.
Dates help here more than debate. Sort the option list by when each value was last ordered, and anything untouched for two years moves the conversation from opinion to evidence. Nobody defends a configuration that has not been sold since the last recession.
Pay particular attention to options created for a single customer who has since left, been acquired, or redesigned around something else. These are the purest form of inherited complexity: real cost, no current demand, and no defender in the room. They are also the easiest place to start, because removing them requires no commercial judgment at all.
Nothing in a normal product process ever removes an option once it exists. That asymmetry, not customer demand, is what builds most catalogs.
3. What does one more option cost downstream?
More than the option itself, and in places that never get charged to it. An additional value on one dimension multiplies against every other dimension, so the effect on your part count is not additive but combinatorial, and the cost lands in forecasting, inventory, changeovers, documentation, and quoting.
Work out where your particular business pays. In a business that stocks finished goods, the cost concentrates in inventory and forecast error, because demand per configuration falls as configurations multiply and small numbers forecast badly. In a make-to-order business, it concentrates in changeovers, engineering time, and quoting effort, since every unusual configuration needs someone to work out how to build and price it.
The forecasting effect deserves particular attention because it compounds quietly. Splitting the same demand across more configurations means each one is forecast from a smaller and noisier history, which produces more safety stock in total to hold the same service level, and more write-offs when the forecast is wrong. You pay twice, in working capital and in obsolescence, and neither charge appears against the option that caused it.
Both effects are invisible in standard costing, which spreads overhead by volume or labor hours and therefore charges your highest-volume configurations for complexity the rare ones create. The rebuild that exposes the real figure is in what a SKU actually costs you, and it is worth doing for two or three representative configurations before you continue with this method.
One number is usually enough to change the conversation: the share of setups on your constrained equipment consumed by configurations in the bottom fifth of volume. If that number is large, you have located where the variety is being paid for, and you have an argument that does not depend on anybody’s cost allocation being correct.
4. Two kinds of customization, and only one of them scales
Customization that happens by combining standard elements is cheap and repeatable. Customization that requires someone to design, specify, or engineer something new is expensive and does not get cheaper with volume, because each instance consumes the same scarce technical attention as the first.
The distinction is not about how different the result looks to the customer. Two products can be equally distinctive to the buyer while one is assembled from a catalog of standard modules and the other is drawn from scratch. What matters is whether producing the variant consumes engineering judgment or only production time.
This is why the argument between standardization and customization is usually framed wrongly. The real choice is not how much variety to offer. It is where the variety is created in your process, and a business can offer a great deal of apparent choice cheaply if the choice is assembled late from standard pieces rather than engineered early into unique ones.
The bucket boundary is also where your pricing should change. Assembled variety can be priced from a list, because its cost is predictable. Engineered variety cannot, and quoting it from a list is how a business ends up selling its scarcest resource at an average price.
Sort your current variety into those two buckets before doing anything else. The engineered-to-order bucket is where your complexity cost actually lives, and it is usually a small share of orders consuming a large share of technical capacity, which is precisely the trade you want to examine.
5. How far back can you standardize without touching the choice?
Move the point where variety is introduced as late in the process as it will go. The customer keeps every option they actually select on, and your plant makes fewer distinct things, for longer runs, before the variant appears at all. Most lines can push that point further back than anyone assumes.
The practical version of this is a common base with differentiation added late: the same core built to one specification, with the customer-visible variety coming from modules, fittings, controls, finishes, or packaging applied near the end. Your buyer’s experience of choice is unchanged, and your forecast, inventory, and changeover burden all fall, because you are now forecasting bases rather than configurations.
Whirlpool’s side-by-side refrigeration line was reorganized on this principle as part of a platform launch, with a good, better, best structure and premium features migrated up into the higher tiers. The customer-facing effect was a clearer choice with more reason to pay for the upper tiers. The internal effect was a smaller number of distinct things to build and a lineup where every surviving product had a job that someone could state.
Late differentiation has an operational precondition that is easy to underestimate: the late steps have to be fast and reliable, or you have simply moved the bottleneck to the end of the line. Check what the finishing, assembly, or configuration steps can actually absorb before redesigning the product around them.
Ask the design question explicitly: for each dimension customers choose on, how late in the process could that choice be applied? Some will be fixed at the very beginning by physics or by the nature of the product, and those are the genuine constraints. Others are fixed early only because that is how the product was originally drawn, and those are available to you.
6. When does standardizing actually lose you a sale?
When the dimension you removed is one customers genuinely choose on, and no remaining configuration meets a requirement they cannot compromise on. Both halves have to be true. That is a considerably narrower condition than the one sales will describe in the meeting, and the useful thing about it is that it is testable.
The test is to find the lost order rather than the lost argument. When someone says a standardization will cost sales, ask for the specific deals from the last year that would have gone the other way, with customers named. Sometimes those deals exist and the objection is correct, in which case you have just learned where the line sits. Often the examples turn out to be quotes that were lost on price or lead time and remembered as lost on configuration.
There is a real asymmetry to respect here. A wrong decision about removing a customer-facing option is visible immediately and blamed on a person, while the cost of carrying unnecessary variety is spread across overhead and blamed on nobody. That asymmetry is why catalogs grow, and it is why the burden of proof has to be set deliberately rather than left to whoever speaks with the most conviction.
Take the timing into account as well. A configuration removed from the catalog while a large quotation is in progress will be remembered as the cause if that deal is lost, whatever the real reason was. Time these changes between bidding cycles where you can, and tell the sales team before customers hear it.
The safest way through it is not to guess. Price the rare configurations at what they cost before you remove them, and watch what customers do. The ones who need them will pay, which tells you the dimension matters, and the ones who do not will move to a standard configuration, which tells you the option was never doing the work it was credited with.
7. What if customization is the thing you sell?
Then the line moves, but it does not disappear. It sits somewhere different: not at what you are willing to build, but at what you are willing to build without charging for the engineering hours that go into it. Configurable businesses fail on the second question, not the first.
Businesses whose advantage genuinely is configurability should standardize hardest in the places customers never see. Common platforms, a disciplined module catalog, reusable engineering, and a quoting process that can price an unusual configuration quickly are what let a business say yes often without the cost of yes accumulating invisibly.
The failure mode here is not offering customization. It is offering it without a price that reflects the engineering hours it consumes, so that each accepted special quietly transfers technical capacity from the products that fund the business to the ones that flatter it. A configurable business needs a floor price for engineering attention in the same way a standard business needs a minimum order quantity.
There is also a portfolio question underneath this one. If configurability is genuinely your advantage, then engineering capacity is the resource your whole business runs on, and allocating it deliberately is a portfolio decision rather than a scheduling one. The criteria for that are in which product lines deserve investment.
The other discipline is a real module catalog rather than a nominal one. Many businesses that describe themselves as configurable are in fact engineering most orders from scratch while calling the result a configuration, and the way to tell is to count how many orders in the last quarter required a drawing that did not previously exist.
8. The test to run on your own line
Take one product family and one option dimension at a time, and run the path in the diagram above. It takes an afternoon per dimension and it produces a decision rather than a discussion.
For each dimension, answer three questions in order. Do customers choose on it, with order history as the evidence rather than opinion? If not, it is past the line and it gets standardized. If they do, can the choice be applied later in the process, so the variety survives but the internal cost falls? If it can, redesign toward a common base. If it cannot, price that configuration at what it actually costs and let the answer come from customers.
Do this per dimension rather than per product, because that is the level where the decisions are actually made. A product line with four option dimensions and five values each has hundreds of possible configurations, and arguing about those one at a time is how these projects consume a year and end with nothing removed.
Start with the dimension where you expect the least argument rather than the one that costs the most. Clearing one dimension end to end teaches the organization what the evidence standard is and what actually happens afterward, which makes the contested dimension a much shorter conversation when you reach it.
Record the result somewhere the product process will see it. A decision that a dimension is standard has to be visible at the point where the next special is requested, or the option will be reintroduced within two quarters by somebody acting in good faith on behalf of a customer.
9. What do you do when sales and operations disagree?
Give them different questions rather than a shared opinion. Sales owns whether customers choose on a dimension, with named deals as evidence. Operations owns what the dimension costs to carry, with setup and inventory figures. One person owns the decision once both answers are in.
Give each side the same deadline, too. These disagreements can run for months on the basis that one party has not finished gathering evidence, and a date does more to resolve them than any amount of further discussion.
Most of these arguments persist because both sides are answering the whole question, which neither one can settle alone. Splitting the question by evidence ends the loop, because each side is now responsible for a claim that can be checked rather than a position that can be restated.
Put a review date on the decision as well. An option standardized away this year can be reconsidered next year if a named customer asks for it and the numbers have changed, and knowing that a decision is revisitable makes it far easier to take. What you are trying to prevent is not reconsideration, which is healthy, but quiet reintroduction through the specials process.
Set the burden of proof explicitly and in advance. The workable default is that an option must be defended with a named customer and a number by a date, or it is standardized. That puts the burden on keeping rather than removing, which corrects the asymmetry described in step six, and it does so by agreement rather than by anyone winning a meeting.
Sales owns whether customers choose on a dimension. Operations owns what it costs to carry. One person owns the decision once both answers are in.
One limitation of this method, stated plainly. It relies on order history to reveal what customers choose on, which works well for a line with a reasonable volume of orders and badly for one where you sell a small number of large systems a year. With twenty orders a year there is no distribution to read, and the answer has to come from direct customer conversations and from your own commercial judgment instead. The diagram still applies; only the evidence for step one changes.
About the Stagnation Assassin
Todd Hagopian is a Fortune 500 transformation executive who has generated more than $3B in documented shareholder value across Berkshire Hathaway, Illinois Tool Works, and Whirlpool. 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.
This one divides a room faster than almost any other operating question, which is why it works well as a live session. Bring the standardization argument to your leadership team or association program through the speaking page, and let sales and operations make their cases with the evidence rules set in advance.