- Why the SKU’s own revenue is the wrong starting number
- The three places the revenue can go
- How to classify each customer
- Building the estimate customer by customer
- Testing the estimate with price instead of guesswork
- What to do with the number once you have it
- Where these estimates usually go wrong
- Frequently asked questions
The analysis is finished. The cut list is sitting in front of you, the case for it is sound, and someone senior has asked the only question that matters to them: how much revenue does this cost us? You do not have a defensible answer, and the number in the cut list, the revenue of the SKUs themselves, is not it.
That question is where most rationalizations stall, and usually for good reason. Nobody wants to authorize a decision whose downside is unknown. The good news is that the estimate is not hard to build. It just has to be built at the level of the customer rather than the product, which is also the level the pillar guide on how to cut a product line without losing revenue uses throughout.
Why the SKU’s own revenue is the wrong starting number
Because the SKU’s revenue measures what you sell, and the risk is what the customer might stop buying. Those are the same number only when a product is bought by customers who buy nothing else from you, which is rare in an industrial business.
The number that gets quoted
Adding up the cut list gives a total that feels like the answer. It is precise, it comes straight from the system, and it is usually far too small. It counts the sales of the products going away and nothing about the relationships those products sit inside.
Why the small number survives anyway
It survives because it is the only number anybody can produce quickly, and because the people who would dispute it, the account managers, are rarely asked before the list is built. The result is a business case whose downside figure is precise, defensible in the meeting, and wrong in the one direction that costs money.
The number that matters
Industrial buyers consolidate suppliers, because every extra vendor costs them qualification, paperwork, and attention. A customer who buys forty items from you and loses one they depend on may move all forty rather than manage two suppliers. The SKU’s sales were small. Your exposure was the entire account.
The SKU’s revenue measures what you sell. The risk is what the customer might stop buying. Those are rarely the same number.
The three places the revenue can go
Every dollar currently attached to a cut SKU ends up in one of three places: on a substitute you still sell, with a competitor who makes that item, or gone along with the rest of that customer’s business. Your estimate is simply a sorting of each customer into one of the three.
To a substitute you still sell
This is the outcome the cut assumes. The customer moves to a nearby product in your line, and the revenue stays with you at a slightly different price. For genuinely redundant variants, most of the revenue lands here.
To a competitor, for that item only
Your nearest remaining product is not close enough, and someone else makes what they need. You lose the item’s revenue and keep the rest of the account. The loss is real and bounded.
Out the door with the basket
The customer consolidates the whole relationship elsewhere. This is the destination that makes rationalizations dangerous, and the only way to estimate it is account by account, with names attached.
How to classify each customer
Pull every customer who bought each cut candidate in the last year, then sort each one into migrate, might move, or basket risk. Three questions do the sorting, none of them requires modeling anything, and the whole exercise takes an afternoon for a typical cut list.
The three questions
Is there a substitute in your line that genuinely does the job, in this customer’s application rather than in principle? How much else does this customer buy from you, and how important are they? And how hard would it be for them to replace you entirely, given qualification requirements, tooling, or approvals on their side?
Who gives the answers
Ask the person who owns each account, and get the answer in writing. Account managers are not always right, and some will classify every customer as a basket risk to protect the products they sell. A written call with a name on it is still the best input available, and it gives you someone to go back to when the outcome differs from the prediction.
What to do with the disagreements
Where an account manager’s call conflicts with what the data suggests, take the more cautious of the two and move on. The point of the classification is to rank risk, not to settle who knows the customer better, and an argument about one account can absorb more time than the whole rest of the exercise.
Where the grid does the work for you
If you have built the customer and product grid, most of this classification is already done. Customers in your top contribution tier who buy a cut candidate are basket risks until proven otherwise. Customers in the bottom tier who buy one item from you are rarely a basket risk at all.
Building the estimate customer by customer
Sum the exposure by category rather than by product, and present it as a range. What you are producing is not a forecast of lost revenue. It is a statement of how much revenue sits behind each level of confidence.
The arithmetic
For migrate customers, count nothing, since the revenue stays. For might-move customers, count the revenue of the cut items only. For basket-risk customers, count their entire annual revenue with you, not just the item. That last line is the one people resist, and it is the whole reason the estimate exists.
What the output looks like
| Category | What you count | What the number means |
|---|---|---|
| Migrate | Nothing | Revenue expected to stay, at the substitute |
| Might move | Revenue of the cut items | Bounded loss, limited to those items |
| Basket risk | The customer’s full annual revenue | Exposure, not expected loss |
Why not weight the categories
You could multiply each basket-risk account by some probability of leaving and produce a single expected figure. Resist it. The probability would be invented, the resulting number would look authoritative, and it would hide the concentration that actually determines what to do. Three honest categories beat one false total.
Do it once, for the whole list
Classify every customer on every candidate in one sitting rather than product by product as each cut comes up for decision. The same accounts appear again and again across a cut list, and sorting them once gives you a single view of which relationships the whole exercise touches, which is the view the person approving it actually wants.
Exposure is not a forecast
Say that out loud when you present it, because the basket-risk figure will be uncomfortably large and somebody will read it as a prediction. It is the amount at risk if those specific accounts are handled badly, and handling them well is a choice you control. Its job is to tell you which cuts need a migration plan and a personal conversation before they are announced, and which ones do not.
Testing the estimate with price instead of guesswork
There is a cheaper way to find out which customers truly depend on a product, and it costs you almost nothing if your classification turns out to be wrong. Raise the price on the candidate first, give it a date, and watch which customers pay it.
How the test works
Set the price at a level where the product would genuinely pay for the capacity and complexity it consumes, and give it a date. Customers who need it will pay, and the product stops being a problem, since it now earns its place. Customers who do not will migrate themselves to the substitute you would have pointed them toward anyway.
Why it beats a model
Every customer sorts themselves into the right category with their own money, which is more reliable than any classification you or your account managers can produce. The catch is that the increase has to be real, with a date and a written notice. A price floated verbally and withdrawn at the first complaint teaches everyone that nothing was ever at risk.
What to do with the number once you have it
Use it to sort the cut list into three actions rather than to decide whether to proceed at all. Cuts with no basket risk go ahead on schedule. Cuts with concentrated basket risk get a migration plan first. Cuts where a single account carries the exposure go back for a decision about that account.
The concentration question
Look at how the exposure is distributed before you look at its total. Exposure spread thinly across many small accounts is a manageable risk. The same total sitting behind two large accounts is a different situation, and it usually means the product is not really the decision in front of you.
Exposure spread across many small accounts is a risk. The same total sitting behind two accounts is not a product decision at all.
Where these estimates usually go wrong
Four failure modes account for most bad estimates, and all four are avoidable once you know to look for them. Check for each one before you present the number, because every one of them is far easier to fix beforehand than to defend in front of the people approving the cut.
Substitutes that only work on paper
A substitute that requires the customer to requalify, retool, or revalidate is not a substitute in the year you need it to be one. Ask what the customer would actually have to do, not whether the specifications overlap.
Counting the item instead of the account
The most common error, and the most expensive. It produces a small, confident number that turns out to be wrong in the only direction that matters.
Treating a quiet customer as a safe one
An account that never calls and never complains is easy to classify as low risk. It is just as likely to be an account that switches without a conversation, because nothing in the relationship gives them a reason to warn you first.
Missing the customers behind a distributor
When a distributor buys the product, the customer who depends on it is often invisible to you. Ask the distributor which end users are behind the volume before you classify them.
Assuming the cut happens all at once
Revenue does not stop on the announcement date. It runs down through last-time-buy orders and contractual obligations over quarters, which affects both the timing of the loss and the timing of the savings.
Frequently asked questions
How accurate can this estimate be?
Accurate enough to rank the risk, which is what it is for. It will not predict a total, and it should not be presented as though it could. What it does reliably is separate the cuts you can make on schedule from the ones that need work first.
What if we do not have customer-level data?
You almost certainly do. Invoice lines carry customer and product together, and a year of them is enough. The obstacle is usually that nobody has put them in one place before, not that the data is missing.
Should we tell customers before the estimate is done?
No. The classification tells you who to speak to first, and a general announcement before you have that list is how a basket-risk account learns about it from a form letter. The sequence for those conversations is in how to sunset a product without losing the customer.
Does this apply to a distributor-heavy business?
Yes, with one extra step. Your direct customer is the distributor, but the dependency usually sits with an end user you cannot see, so a distributor buying a candidate should be treated as unclassified until they tell you who is behind the volume.
What if the exposure is too big to accept?
Then you have learned something useful before spending anything. Shrink the list to the products with no basket risk, run those, and revisit the rest once you have seen how the first round actually behaved against your predictions.
About the Stagnation Assassin
Todd Hagopian is a Fortune 500 transformation executive who has generated $3B in shareholder value across Berkshire Hathaway, Illinois Tool Works, Whirlpool, and JBT Marel, where he serves as Global 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.
Here is the question to answer yourself first: of the customers who buy the products on your cut list, how many are in your top twenty by profit contribution? If you cannot answer that in ten minutes, the estimate is not your first problem. Send me what you find through the contact page and I will tell you what I would check next.