Your gross margin is two and a half points lower than it was four years ago. Nobody made that decision. There is no meeting where it was approved, no memo announcing it, and if you ask four executives where it went you will get four confident answers that contradict each other. Sales says pricing pressure. Operations says input costs. Finance says mix. Procurement says all three.
All four are partly right, which is the problem. Margin does not leave through one hole. It leaves through eleven or twelve, each one small enough that no single one is worth an escalation, each one sitting in an accounting category that belongs to a function rather than to a person, and collectively adding up to the entire difference between a good year and a mediocre one.
This article names them, sorts them, and tells you which to chase first. It assumes you have already built the customer and product profitability picture, because that is what tells you which accounts and products each leak is concentrated in.
The leaks fall into three families. Commercial leaks are price you gave away. Transactional leaks are cost you absorbed on the customer’s behalf. Operational leaks are value you destroyed and then paid to replace. They differ enormously in size, in how fast they close, and in who has to be persuaded, and those three criteria are the only ones that matter when deciding what to do first.
One framing point before the list, because it determines how the conversation goes internally. None of what follows is an accusation. Every leak here was created by somebody doing their job inside the incentives you gave them, and presenting the findings as a discovery of waste will guarantee that the people who have to close the leaks spend the quarter defending themselves instead. Present it as a measurement gap, which is what it actually is, and the same people will hand you the answers.
Why does none of this show up on your P&L?
Because every one of these leaks is correctly classified as something else. That is not an accounting failure. It is what financial reporting is designed to do, and it is why the leak survives for years inside a business full of competent people looking at accurate numbers.
Take discount drift. The discount you granted is not recorded anywhere as a discount. It is recorded as revenue, at the lower number, and there is no line on any statement called price we did not charge. The only way to see it is to compare invoiced price against list price, by account, over time, and almost no management reporting package does that because the general ledger has no reason to.
Take absorbed freight. It appears in your freight cost line, aggregated across every shipment, where it looks like a logistics number and gets managed by a logistics person whose job is to reduce the cost of shipping. It is not a logistics number. It is a price concession granted account by account by salespeople, and the logistics person cannot fix it because they did not grant it and have no authority over the accounts that received it.
Take expedites. Each one is a scheduling decision made by a planner, recorded as premium freight and overtime, and entirely invisible as a commercial matter. Nobody sums them by customer. When you do, the distribution is brutally concentrated, and a handful of accounts turn out to be consuming a disproportionate share of your plant’s flexibility for free.
The pattern repeats across all of them. Every leak is a commercial decision recorded as an operating cost, or an operating failure recorded as a cost of doing business, and in both cases the number lands in a function that did not cause it and cannot close it. That structural mismatch is why these persist. It is also why the fix is almost always about ownership rather than analysis.
There is one more reason these persist that is worth naming because it is not an accounting matter at all. Each leak, taken individually, is below the threshold at which anyone escalates. A salesperson granting two points to protect an account is making a defensible call. A planner expediting an order for a customer who asked nicely is doing their job. A credit manager extending terms during a difficult quarter is managing a relationship. Nobody in this chain is behaving badly, and that is exactly why the aggregate is invisible: there is no misconduct to detect, only a hundred reasonable decisions with no mechanism summing them.
Which means the diagnostic question is never who is responsible. It is which number, if somebody looked at it monthly, would have made the pattern visible in month two rather than year four. Every leak below has such a number, and in most businesses none of them is on anybody’s report.
One more structural reason worth naming: the leaks are recorded at the wrong level of aggregation. Your freight line is a company number. Your expedite cost is a plant number. Your realization, if you track it at all, is probably a segment average. Every one of these is a distribution masquerading as a total, and a total is exactly the shape that hides a leak, because the accounts behaving badly are averaged against the ones behaving well until the whole thing looks acceptable. The single most useful analytical change available to most manufacturers is not a new metric. It is taking three metrics they already have and breaking them out by account.
One of the biggest leaks is pricing you plan to fix later. See The Recovery Myth vs. Blitzscaling.
Underpricing is one of the quietest margin leaks. See the 60-day price elasticity test.
Margin leaks hide in metrics that ignore the end user; read the $75 cart that cost retailers millions.
How do you compare one leak against another?
Three criteria, applied to all of them on the same basis: how big it is in dollars, how fast it closes, and who has to agree. The third one is the criterion people skip and it is usually the binding constraint, because a large fast leak owned by somebody who does not report to you is a harder project than a small slow one you control outright.
Size first, and size honestly. Every leak below has a way to estimate it from data you already have, and the estimate does not need to be precise. You are ranking, not budgeting. An order-of-magnitude number produced in an afternoon beats a precise one produced never, and the ranking rarely changes between the two.
Speed second. Some of these close on the next invoice, some need a notice period, and some need a contract cycle. This matters more than size when you are inside a year, which you usually are, and it is the reason the sequence below is not simply sorted by dollars.
Ownership third. Ask who would have to change their behaviour for this leak to close, and whether that person currently has any reason to. A leak that requires a salesperson to stop doing something their compensation rewards is not an analysis problem, and treating it as one will waste a quarter.
Look at the shape of that diagram and then look at where your improvement effort actually goes. In every business I have worked in, the operational band receives the most attention, the most headcount and the most executive time, and it is the smallest of the three. It gets the attention because it has a clear owner and a visible number, which are exactly the properties the other two lack.
One caution on sizing before you start. You will be tempted to build a comprehensive leakage model that quantifies all eleven items precisely, and it will take a quarter and produce a number nobody acts on. Size the three families roughly, pick the family with the largest fast-closing item, and go. The precision you gave up was never going to change which one you chased first, and the quarter you saved is a quarter of the leak actually closed.
Apply all three criteria as a single pass rather than sequentially, and write the answers down for every leak before deciding anything. The output is a one-page table with eleven rows and three columns, and it is worth more than any consultant’s diagnostic, because it is built entirely from what your own people already know and it forces the ownership question into the open while the conversation is still about analysis rather than about blame.
Expect the table to surprise you in one specific way. The leaks your leadership team nominates as the largest are almost always in the operational band, because that is where the visible numbers live. When you finish sizing them honestly, the ranking usually inverts. That inversion is the most valuable output of the exercise and it is worth making the table in a room together rather than distributing it afterward, because the argument as it flips is where the organization actually learns something.
Where do the commercial leaks sit?
In the gap between your list price and your invoiced price, which is the single largest leak in most industrial businesses and the one least likely to be measured. Three forms: discount drift, spec creep, and the quote-to-invoice gap.
Discount drift is the accumulation of concessions nobody revisited. A point granted to close a quarter in 2022. A second point given when a competitor quoted aggressively on one job and then applied to the whole account. A third absorbed during a supply disruption. None of these was wrong at the time and none of them had an expiry date, so all of them are still running.
Hypothetical, to size it: a business at $40 million in revenue with two points of accumulated drift across the book is leaking $800,000 a year, and at a thirty percent gross margin that is nearly seven percent of gross profit. Pull your own realization by account and the number will be specific rather than illustrative, and it will usually be larger than this.
Spec creep is the same phenomenon in engineering. A customer asks for a tolerance, a finish, a test report, a certification, a packaging change. Each is granted in a single conversation by somebody being helpful, each carries a real cost, and none triggers a price review because the part number did not change. Over five years you are manufacturing a substantially different product at the original price.
The quote-to-invoice gap is the third and it is the most embarrassing when you find it. Price quoted, price ordered, price invoiced, and price paid after deductions are four different numbers in most businesses, and the difference between the first and the last includes unauthorized deductions, unearned rebates, freight claims and short payments that nobody chased because each was too small to be worth the call.
All three close on the next transaction, which is what makes this band the right place to start. You are not renegotiating history. You are deciding that the next quote goes out differently, and no customer has a contractual right to a concession you have not yet granted.
A note on where the commercial band hides in businesses that believe they have no discounting. Plenty of manufacturers quote from a price list with no formal discount structure at all, and conclude that discount drift cannot apply to them. It applies through a different mechanism: the list itself. If your price list has not been rebuilt in four years while your costs moved, you are discounting relative to your own economics, silently, on every transaction, and the gap does not appear as a discount because there is nothing to compare it to. Rebuild the list annually against current cost and current value, and the drift becomes visible.
The other hiding place is the quote that was never won. Realization analysis looks at what you invoiced, which is a sample biased toward the prices customers accepted. Pull your lost quotes alongside the won ones and look at the price difference. If you are losing at a price barely below where you are winning, your problem is not price at all, and continuing to discount into that gap is the most expensive form of this leak.
Spec creep deserves one more paragraph because it is the leak most likely to be defended as good customer service, which it genuinely was at the moment it was granted. The failure is not the concession, it is the absence of a review point. A part that has accumulated four customer-requested modifications over six years is a different product, and the moment to reprice it is the next annual review or the next engineering change, not never. Build the review point into your product lifecycle process and the leak stops accumulating, without anybody having to refuse a customer anything.
The quote-to-invoice gap has a fast diagnostic worth running this week. Take fifty invoices at random from last quarter and trace each back to its quote. Compare the four numbers: quoted, ordered, invoiced, paid. In most businesses at least a fifth of that sample will show a difference somewhere in the chain that nobody can explain, and the total across those fifty, extrapolated, is the size of the leak.
Where do the transactional leaks sit?
In the costs you absorb on the customer’s behalf without ever having priced them: freight, payment terms, expedites and small orders. Individually modest, collectively substantial, and concentrated in a small number of accounts almost every time.
Freight first. Absorbed freight is a discount denominated in logistics. Pull freight cost as a percentage of revenue by account and you will find a spread far wider than any policy you have written, with the extreme cases usually explained by an arrangement made years ago for a customer who is no longer important.
Payment terms next, and this one is chronically underestimated because it never appears as a cost at all. Extended terms are a financing cost you are absorbing. Hypothetical: an account at $2 million a year moved from thirty to ninety days is sixty extra days of receivable, roughly $330,000 of working capital, and at any realistic cost of capital that is tens of thousands of dollars a year granted in a conversation that took five minutes.
Expedites are the third and the most concentrated. Every expedite consumes a setup, a schedule disruption, premium freight and somebody’s afternoon. Count them by account for a quarter. The distribution will be severe, and the accounts at the top will almost always be ones you had not suspected, because the planner absorbing the disruption has never been asked to report it commercially.
Small orders complete the set. An order line costs roughly the same to process whether it carries one unit or four hundred, so the cost per dollar of revenue on a small order is a multiple of your average. A minimum order quantity or a small-order fee addresses it directly, and both have the useful side effect of changing customer ordering behaviour rather than merely recovering the cost.
This band closes in one notice period, typically thirty to ninety days, because each item requires telling somebody that an arrangement is changing. That is slower than the commercial band and considerably faster than the operational one.
There is a fifth item in this band that rarely gets counted and is occasionally the largest of them: unearned rebates and deductions. Volume rebates calculated on forecasts that were never met and never clawed back. Promotional allowances that continued past the promotion. Chargebacks and short payments taken unilaterally by large accounts and written off because pursuing each one costs more than it recovers. Individually trivial, and in businesses selling through large retail or distribution partners this can be the single biggest line on the entire list.
Sizing it is straightforward and uncomfortable. Sum every credit, allowance, rebate and deduction taken against invoices for one year and express it as a percentage of gross revenue. Most businesses have never produced that figure, and the exercise takes an afternoon. The number is a price concession by any economic definition, and almost nowhere is it managed as one.
All four of these transactional items share a property that makes them unusually tractable: each one has a natural policy answer that applies to everyone and therefore requires no individual negotiation. A published freight policy above a threshold. Standard terms with a named exception process. An expedite fee. A minimum order quantity. Announcing a policy is a fundamentally different act from renegotiating an account, and it is why this band closes faster than its size would suggest.
It is also the band where the fix changes customer behaviour rather than merely recovering cost, which is worth more over time than the money. An expedite fee reduces expedites. A minimum order quantity consolidates order patterns. Both return capacity and predictability to your plant, and neither of those benefits appears in the recovery calculation you used to justify the change.
Where do the operational leaks sit?
In returns, rework, scrap and obsolescence, which is the band your organization is already working on and the smallest of the three. I am not arguing that this work is unimportant. I am arguing that it is over-resourced relative to the other two, and that the reason is visibility rather than value.
Returns and warranty carry a cost most businesses understate, because the recorded number is usually the credit issued and not the freight in both directions, the inspection, the disposition, the replacement production and the administrative handling. The full cost of a returned unit is frequently several times the credit, and none of the rest is attributed to the account that generated it.
Rework is similar and worse, because rework is often invisible by design. Work performed to fix something that should have been right consumes capacity that had already been counted as available, which means it shows up as a schedule miss rather than as a cost. In plants where rework is absorbed informally by experienced people, it does not appear anywhere at all.
Scrap and obsolescence are the most visible items on this list and the most diligently managed, which is precisely why they are usually the smallest remaining opportunity. If your operations team has been at this for years, the accessible scrap is gone. Obsolescence is the exception worth a second look, because it is generated by portfolio decisions rather than by manufacturing, and it therefore sits in the same ownership gap as everything in the first two bands.
This band closes slowest, requires capital or process change, and needs the most people to agree. Which is the argument for working it continuously through an operating cadence rather than through a margin recovery program, and for not letting it absorb the attention that belongs in the top band.
A business at $40 million with two points of accumulated discount drift is leaking $800,000 a year. At a thirty percent gross margin that is nearly seven percent of gross profit.
One more structural point about this band, and it is the reason I keep the three families separate rather than presenting a single ranked list of eleven leaks. The operational band is where cost is genuinely destroyed, meaning value disappears from the economy rather than moving to the customer. The commercial and transactional bands are transfers: the money went to your customer, who is enjoying it. That distinction matters when you are deciding what to chase, because a transfer can be reversed by a decision, while destroyed value has to be prevented by a process change. Decisions are fast. Process changes are not.
Obsolescence deserves a closer look than it usually gets, because it is the one item in this band that is not actually an operational failure. Inventory becomes obsolete because somebody launched a product, forecast it optimistically, kept it in the catalogue past its useful life, or agreed to a customer-specific variant that was ordered twice. Every one of those is a portfolio or commercial decision, and writing the resulting inventory off as an operations number both misattributes the cause and guarantees it recurs. Track obsolescence by the decision that caused it rather than by the warehouse it sat in.
Which one should you chase first?
Discount drift, every time, and it is not close. It is the largest, the fastest to close, and the only one that requires no customer conversation about the past, because you are changing what happens on the next quote rather than renegotiating an existing one.
The sequence I would run, and the reasoning behind it: realization on new quotes first, because it starts the same week and costs nothing. Then expedite and small-order fees, because they recover cost and change behaviour simultaneously. Then freight and terms, account by account, because each needs a conversation. Then spec creep at the next product review, because it needs engineering involvement. Then the operational band, continuously, as an operating cadence rather than a project.
Notice that this sequence is sorted by speed and ownership rather than by size, with one exception: discount drift happens to be both the biggest and the fastest, which is why it is unambiguous. Everything after it involves a trade between how much is there and how long it takes to get.
The mistake to avoid is running all of these at once as a comprehensive leakage program. A program with eleven workstreams produces a governance structure and a status report. Four leaks closed in sequence by named people produces money, and the first one funds the attention required for the rest. The ordering logic here is the same one that applies to margin levers generally, which sort by speed rather than by size.
A word on sequencing inside the first move, because even discount drift can be run badly. Do not attempt to reverse existing discounts across the book in month one. Start with new quotes only, where you have no history to renegotiate and no customer has a claim on a concession you have not yet granted. Let that run for a quarter and measure it. Then, and only then, take the ten largest historical concessions individually, with the data, one conversation at a time. Businesses that invert this order start with the difficult conversations, generate resistance internally and externally, and abandon the effort before reaching the easy money that was available the whole time.
One question always comes up at this point and it deserves a direct answer: how much of the two and a half points can you realistically get back. Not all of it, and anyone promising otherwise is selling something. Some of the drift was genuine competitive response and reversing it would cost you the volume. Some of the spec creep is now the product and the customer will not accept the original. A reasonable working assumption is that the commercial band is roughly half recoverable, the transactional band is mostly recoverable because it is policy rather than negotiation, and the operational band is whatever your process work has not already taken. Build your plan on the recoverable share rather than on the leak total, because a target built on the full number will be missed and the programme will be judged a failure while delivering real money.
How do you keep them closed?
By giving each leak an owner and a number that somebody reads monthly, which is the only mechanism that has ever worked and the one most businesses skip because it is unglamorous. Every leak on this list reopened somewhere because the person who closed it moved on and nobody inherited the measurement.
Build one page. Realization against list by account. Freight as a percentage of revenue by account. Expedite count by account. Average days to pay by account. Order lines below your minimum. Returns and rework as a percentage of output. Six numbers, monthly, with a name attached to each and a standing five minutes in an existing meeting.
Then set a threshold on each and require an explanation when it is crossed, rather than reviewing all six every month in detail. Exception-based review survives busy quarters. Comprehensive review does not, and a comprehensive review that quietly stops being read is worse than no review, because it creates the impression of control.
The last piece is compensation, and it is the one that determines whether any of this holds. If your commercial organization is paid on revenue at any price, every leak in the commercial and transactional bands will reopen, steadily, in a way nobody intends and everybody can predict. You do not have to solve that before starting. You do have to solve it before year two, or you will be reading this article again with the same numbers.
The question was where margin actually leaks in a manufacturing business. The answer is that most of it leaks through the commercial band, in gaps that your financial reporting is structurally incapable of showing you, owned by nobody in particular and closable on your next quote.
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.