- 1. What does 80/20 actually mean inside Illinois Tool Works?
- 2. What does the sort actually cover?
- 3. Why does the simplification have to run front to back?
- 4. What happens to the customers in the tail?
- 5. Why does the sort need a structure change to hold?
- 6. Where do most copies of this fail?
- 7. What can you take from this if you are not Illinois Tool Works?
- 8. How far past 80/20 can you push this?
Every 80/20 review I have sat in opens the same way. Somebody brings a revenue ranking. Customers down the left, twelve months of sales across the top, sorted descending. The top of the page is familiar and comfortable. The bottom of the page runs for four more sheets and nobody in the room has looked at it in a year.
Then somebody says the sentence that kills the whole exercise: “so we focus on the top twenty percent.” And the meeting ends, and a slide gets made, and eighteen months later the tail is longer than it was before.
That is not what Illinois Tool Works does, and the gap between those two things is the entire subject of this article. The public version of 80/20 is an analysis. The version that produced a multi-decade compounding machine is a restructuring process that happens to start with an analysis. Almost everything written about it describes the first one. I have run the second one, and the difference is not subtle.
1. What does 80/20 actually mean inside Illinois Tool Works?
It means a set of tools that divisions use to rebuild how the business is structured and operated, not a ranking exercise that produces a focus list. The company calls it the 80/20 Front-to-Back Process and treats it as a trade secret. The operative verbs in its own description are structure and operate, not analyze or prioritize.
Read the company’s own language on this and one thing jumps out. Illinois Tool Works describes the process as a set of proprietary tools its divisions use to structure and operate their businesses around their largest and most profitable customers, and to minimize the cost, complexity and distraction of serving small ones. Two verbs, both of them about how the business is built. Nothing in that sentence is about insight.
This is where nearly every copy goes wrong, and it goes wrong in the first meeting. A ranking tells you what is true. A structure decides what happens next quarter when a salesperson has a slow month and a tail customer calls with a custom request. If you did the ranking and left the structure alone, the salesperson takes the order, because every incentive in the building says take the order. The ranking changed nothing. It was never going to.
The second thing to notice is that the process does not stand alone. It is one of three elements of the operating model, alongside customer-back innovation and a decentralized culture that pushes real authority down to division leadership. That is not corporate garnish. The sort only survives because the person who ran it also owns the P&L, the pricing, the product roadmap and the plant, and therefore cannot escape the consequences of ignoring it.
2. What does the sort actually cover?
Customers and products together, plus every activity attached to them. A single-axis customer ranking is the most common way to do this badly, because it hides the actual problem: small customers buying non-core products, which is where complexity compounds fastest and where the cost is almost never allocated.
Build the matrix both ways. Customers down one axis, product families across the other. What you get is four populations, and they behave nothing alike.
Large customers buying core products are the business. Large customers buying tail products are the trap, because the volume makes the complexity feel justified. Small customers buying core products are usually fine and frequently growable, which is the quadrant most crude 80/20 exercises kill by mistake. Small customers buying tail products are where the money goes to die.
That fourth quadrant is almost always tiny on a revenue basis and enormous on a cost basis, and the reason is arithmetic rather than opinion. A part number consumes the same fixed load whether it ships once a year or once an hour: a drawing, a spec, a supplier, a routing, a setup, a place in the forecast, a line in the count, a quality record, a person who knows how to build it.
Hypothetical: a plant carrying 2,000 active part numbers where 1,400 of them produce four percent of revenue. If the fully loaded annual carrying load per part number is $900, which is a deliberately conservative figure, those 1,400 numbers absorb $1.26M a year to deliver four points of the top line. That math is illustrative, not drawn from any company’s reported figures. Run it with your own carrying cost and the shape will not change, because the shape comes from the fixed load, not from the rate.
The quadrant that gets killed by accident deserves one more sentence, because the mistake is expensive and quiet. Small customers buying your core products are frequently the cheapest customers you have: no custom work, no expedites, standard terms, standard lead time, and often a better price because they have less leverage. A crude revenue-ranked cut removes them first, which means you have just fired your highest-margin accounts and kept the large ones that grind you on terms. Sort on margin per unit of complexity and that population moves from the bottom of the list to somewhere near the top.
A part number consumes the same fixed load whether it ships once a year or once an hour. Fourteen hundred of them absorbing $900 each is $1.26M a year to deliver four points of revenue.
3. Why does the simplification have to run front to back?
Because the cost you are trying to remove does not live at the front. The front-to-back label means the simplification runs the length of the value chain: commercial structure, production, supply chain, and the administrative overhead attached to all of it. Stop at the front and you have changed who calls whom, without removing a dollar.
This is the failure I have watched most often, and it is seductive because the front is the easy half. Re-segmenting a sales force is a two-week project. You draw new territories, you name the key accounts, you change the call plan, and something visible has happened. Everyone can see it.
Meanwhile the plant still carries every routing. The supplier base still carries every vendor. The ERP still carries every part number, the quality system still carries every document, and the month-end close still consumes the same four days because the number of things being closed has not moved. The complexity did not leave the building. It just stopped being discussed in the sales meeting.
Run the same list back through the operation and the removals get specific. Routings retired. Tooling scrapped or sold. Suppliers consolidated so the remaining volume earns real leverage. Part numbers deactivated rather than left dormant, because a dormant part number still shows up in the forecast, the cycle count and the audit. Documentation retired. Changeovers eliminated rather than optimized.
None of those are analysis outputs. They are operational acts with dates and owners, and every one of them takes an argument with somebody who has a reason to keep the thing. That is the actual work, and it is why the process takes quarters rather than weeks.
4. What happens to the customers in the tail?
Most of them stay. The popular version of this process imagines a mass firing, and that is a caricature. Four outcomes are available, and exit is the last and least common of them, used when the first three have been tried and the arithmetic still does not work.
Standardize. The customer stays and the custom item goes. A large share of tail complexity exists because somebody said yes to a variant nobody has revisited since. Offer the standard product, and a surprising number of accounts take it, because their reason for the variant retired three years ago.
Reprice. The customer stays and pays what the complexity costs. This is the outcome people fear and it is the one that most often works, for a reason worth sitting with: a small customer buying a low-volume item frequently has nowhere else to go, and has been getting a price built on a cost allocation that never saw them.
Re-channel. The customer stays and stops being served directly. High-volume, lower-complexity items move to distribution, which is built for exactly this and does it more cheaply than a direct technical seller can. The account is not lost. It is served by the structure that fits it.
Exit. The customer goes, on terms, with notice, with a last-time buy and a named alternative. This is the right answer when the item cannot be standardized, the customer will not accept a price that covers its cost, and no channel fits.
Notice the shape of that tree. Three of the four branches end with the customer still buying from you. The version of this process that gets described in business books ends at the bottom box, which is the rarest outcome and the only one that makes a good anecdote.
5. Why does the sort need a structure change to hold?
Because a sort with no owner reverses itself. The decision has to be made and held at a level where one person owns the full consequence: the revenue lost, the cost removed, the price raised, the customer conversation. Split those across four functions and the sort dies at the first quarter-end that misses.
This is why the decentralized structure is not a separate topic from 80/20. It is the enforcement mechanism. Divisions run with real autonomy and a lean corporate center, which means the person deciding to retire a product family is the same person who has to explain the resulting revenue line, and who also captures the margin when the complexity comes out. Both halves land on one desk.
Now picture the common alternative. Corporate runs the analysis. Corporate distributes the tail list. The division is told to execute it, but still carries a revenue budget written before the sort, and still cannot change price without approval from a function three time zones away. That division will present a plan, execute the harmless twenty percent of it, and quietly protect everything else. It is not insubordination. It is the only rational response to the incentives in front of them.
If you are going to do one thing from this article, do this one: before you run the analysis, identify the single person who will own both the loss and the gain, and confirm they hold pricing authority. If no such person exists, build that first. The analysis will still be there in six weeks.
6. Where do most copies of this fail?
Five ways, and I have seen all five. They are ranked by how often they occur rather than by how much damage they do, though the first one does the most of both.
Run as a project instead of a process. Somebody owns the 80/20 initiative, it has a start date and an end date, and it produces a report. The businesses that compound on this run it as an annual operating cadence with the same rigor as budget season, indefinitely, forever. The complexity comes back every year because every year the organization says yes to new things. A one-time cut against a continuous inflow is a losing structure.
Sort on revenue instead of margin. Revenue ranking is available in twenty minutes and it is wrong, because your largest revenue account is frequently your most expensive one: the deepest discount, the most expedites, the tightest terms, the most engineering support. Until cost is allocated honestly, the ranking is a popularity contest.
Cut the front, keep the back. Covered above, and the most expensive of the five, because it produces all of the political cost of the exercise and none of the savings.
No one owns the no. The sort produces a list of things the company will stop doing. If nobody has the authority and the appetite to say no to the next request for one of them, the list is a suggestion. Within a year you will have re-added a third of what you removed, one reasonable exception at a time.
Treat it as a cost program. This is the subtle one. The point is not to be smaller. The point is to move capacity, engineering time, tooling, working capital and management attention out of the part of the business that cannot pay for them and into the part that can. A business that runs this properly usually grows, because the freed capacity goes somewhere.
There is a sixth failure that belongs in its own paragraph because it is the one that gets people fired. The exercise is announced before the analysis is credible. Somebody presents a preliminary tail list in a leadership meeting, the room reacts, word travels, and within a week your sales organization is telling customers that a rationalization is coming. Now every account in the tail is fielding calls from your competitors, the good ones start dual-sourcing as a precaution, and you have not yet decided anything. Do the arithmetic quietly, decide the four outcomes account by account, and communicate once, with specifics. There is no version of this that benefits from an early announcement.
A one-time cut against a continuous inflow is a losing structure. Remove 1,400 part numbers once and say yes 300 times a year, and you are back where you started inside five years.
7. What can you take from this if you are not Illinois Tool Works?
Three things, and none of them require scale. The process is often described as something that only works at a multibillion-dollar diversified manufacturer, which is backwards: it is easier in a small business, because one person already owns the whole consequence.
First, build the matrix both ways before you decide anything. Customers by product families, with cost allocated as honestly as you can manage in three weeks rather than perfectly in six months. A rough allocation that exists beats a precise one that never ships.
Second, push the list all the way to the back of the operation. Write the removals as dated operational acts with names attached: this routing retired by March, this tooling scrapped, these eleven suppliers consolidated to three, these 400 part numbers deactivated rather than dormant.
Third, install the cadence before you install the cut. Decide now, in writing, that you will re-run this every year in the same month, and decide who says no in between. Without that, you are buying a one-time benefit at full political price.
I ran a small industrial tank liner manufacturer for several years, bought it and eventually sold it to its largest vendor, and the same arithmetic held at a fraction of a percent of the revenue. The tail was proportionally just as long. The difference was that removing it took a conversation rather than a governance process.
8. How far past 80/20 can you push this?
One level further than almost anyone does, and that level is where the unusual money is. Apply the same sort a second time to the twenty percent that survived the first one, and the concentration compounds rather than flattening.
The arithmetic is straightforward and most people have never run it. If twenty percent of inputs drive eighty percent of output, and that relationship holds inside the surviving population, then four percent of inputs drive sixty-four percent of output. The method I use for this is 80/20 Squared, and its practical value is that the second cut lands on a population that has already passed the first screen, which means everything in it looked like a winner an hour ago.
That is the uncomfortable part and it is the point. The first pass is easy because the tail is obviously the tail. The second pass forces you to rank things you have already blessed, and the answers are less comfortable: inside your best customer segment, a small number of accounts carry the majority of the profit, and the rest are riding on the segment’s reputation rather than their own numbers.
Hypothetical, to make the shape visible: a business with 500 customers where 100 of them produce eighty percent of gross profit. Run the sort again inside those 100 and roughly 20 customers produce sixty-four percent of the total. Twenty relationships. That is a number a management team can actually do something about: name them, assign them, build the roadmap around them, and know precisely what happens if one leaves.
Illinois Tool Works built a compounding machine on the first cut applied relentlessly for decades. The second cut is available to anyone willing to rank their winners against each other, and most organizations will not do it, because the first cut already made everyone uncomfortable and the second one asks them to do it again to people they like.
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.
Run the 15-minute 80/20 Portfolio Performance Audit. Bring one quarter of revenue by customer and by product family. In fifteen minutes I will tell you how long your tail actually is, which of the four outcomes each part of it belongs in, and what the second cut would look like. Request the audit here.