The approval request arrives at four in the afternoon on the last day of the quarter. Three points, on a deal that has been in the forecast for two months, with a note explaining that the customer is deciding today and the competitor has quoted lower. Whoever has to answer it has ninety minutes, no data, and a number they are also being measured on.
That moment is where your margin is decided, and it recurs every quarter in every industrial business I have ever worked in. Everything below is about making sure the answer at four in the afternoon was already determined at nine in the morning three months earlier.
This sits downstream of the customer and product profitability work, which tells you which accounts have room and which do not.
Why does your sales team discount?
Because discounting works, immediately and reliably, at solving the problem they are actually being asked to solve. A salesperson under a revenue number with a deal slipping has one lever that closes a gap this week, and it is price. Everything else takes longer than the quarter they are in.
Start here rather than with a discipline speech, because the behaviour is rational. Nobody is being careless. They are optimizing the thing you measure them on, using the only instrument that moves fast enough, and if you change nothing else the behaviour will not change either.
There is a second reason and it is structural rather than motivational. In most industrial businesses nobody has told the salesperson what the floor is. They know the list price and they know their quota. Between those two numbers is an undefined space in which they are expected to exercise judgement, with no information about where judgement stops being judgement and starts being value destruction. Given that brief, a reasonable person discounts until the deal closes, because no other signal exists.
The third reason, and the one nobody says in the room, is that discounting is free to the person doing it. The cost lands on a gross profit line the salesperson is not measured on, in a period after the one they are trying to save, attributed to nobody. Every incentive in that arrangement points one way, and it will keep pointing that way regardless of what anyone says in a kickoff meeting.
Is this a discipline problem or an authority problem?
Almost always authority. A discipline problem means people know the rule and break it. An authority problem means the rule exists, somebody senior overrides it under pressure, and everyone learns within two quarters that the rule is advisory.
The test is simple and uncomfortable. Count the exceptions granted above your threshold in the last quarter, and count how many were refused. If the refusal rate is near zero, you do not have a threshold, you have a notification step. The rule is not being broken. It is being honoured in a form that costs nothing to honour.
The distinction matters because the two problems have opposite remedies. A discipline problem is fixed with training, communication and consequence. An authority problem is fixed by moving the approval to somebody who can absorb the pressure, and no amount of training touches it. Diagnose before prescribing, because a training programme aimed at an authority problem consumes a quarter and changes nothing, and the organization concludes that the whole subject is unfixable.
One more test worth running before you design anything. Ask three salespeople, separately, what the lowest price they are allowed to quote is. If you get three different answers, or three versions of “it depends on the situation,” you have your diagnosis and you did not need the exception data. A floor that cannot be stated by the people expected to hold it does not exist.
What does a discount actually cost you?
Far more than the percentage suggests, because it lands on gross profit rather than on revenue. At a thirty percent gross margin, three points of discount removes ten percent of the gross profit on that transaction, and to replace it you would need to sell ten percent more volume at full price.
That arithmetic is the single most useful thing you can put in front of a commercial team, because it converts an abstraction into something they can feel. Most salespeople have never been shown it, and the ones who have stop treating three points as a rounding adjustment.
Run the same arithmetic at your own margin before you present it, because the multiplier varies sharply and the low-margin version is startling. At twenty percent gross margin, three points of discount removes fifteen percent of the gross profit on that transaction. At forty percent it removes seven and a half. The businesses with the least room are the ones where the conversation is hardest to have, which is the recurring shape of this entire subject.
Then take it one step further and express it in units. If a salesperson understands that a three point concession on one order requires ten additional orders at full price to replace, the concession stops being a negotiating courtesy and becomes a trade they can evaluate. That framing does more than any policy document.
There is a version of this that is worse than a discount and almost never counted, which is the free addition. Expedited shipping absorbed, a sample provided at no charge, an engineering change waived, extended terms granted to close the same deal. Each is a price concession that never appears in your discount data because it was never expressed as a price. When you set thresholds, set them on total concession value rather than on the discount field alone, or the behaviour simply migrates to the channels you are not watching.
Where should the threshold sit?
At the point where your own realization distribution says the exceptional begins, not at a round number somebody picked. Pull invoiced price against list by transaction for the last year, find the eightieth percentile of discount, and set the threshold just above it.
Set it there and roughly a fifth of transactions require an approval, which is an administrable volume. Set it at a round five percent without looking at the distribution and you will either approve everything, which teaches people the control is theatre, or block half your business, which teaches them to route around it.
Revisit the threshold annually and after any significant cost movement, because a discount percentage measured against a stale list becomes more permissive every year without anyone changing it. The threshold is a live number, not a policy you write once. Put its review on the same annual cycle as the price list rebuild so the two never drift apart.
Publish the threshold rather than operating it quietly. A control people know about changes behaviour before any request is ever submitted, because the salesperson prices differently knowing what will trigger a conversation. A control operated invisibly only catches things, which is far less valuable than preventing them. The pre-emptive effect is usually larger than the refusals.
Who should approve, and who should not?
Somebody who does not carry the revenue number being protected. That single rule eliminates most of the failures in this area, and it is the rule most commonly broken because the obvious approver is the sales manager.
A sales manager approving a discount to hit a sales number is not a control, it is a formality with an extra signature. Put the approval with finance, with the general manager, or with a pricing function reporting outside sales. The approver does not need to know the account. They need to be able to say no without it costing them personally.
The second requirement is that the approver has access to the account’s actual profitability at the moment they decide. An approver with no data will approve, because refusing without evidence is professionally uncomfortable and approving is not. Give them the cost-to-serve number and the account’s realization history on one screen, and the refusal rate changes immediately without anybody being instructed to be tougher.
One structural caution. Whoever holds this approval will become unpopular, predictably and through no fault of their own, and if nobody above them defends the role it will quietly be reassigned within a year. Decide in advance who backs the approver when the first serious escalation reaches the general manager, and say so publicly when you set the control up. The role survives on that commitment rather than on its job description.
How fast does the approval have to be?
Same day, and this is the requirement most governance designs fail. A control that takes three days to produce an answer will be bypassed, and it will be bypassed by good people trying to serve a customer, which makes it impossible to enforce afterwards.
Commit to a turnaround in writing when you set the threshold. Four working hours is achievable in most businesses. Name a deputy for when the approver is travelling. The speed of the control is what determines whether it is respected, far more than the seniority of the person operating it.
Build the deputy into the design rather than treating absence as an exception. Most bypass events I have traced started with an approver who was travelling, a salesperson who could not wait, and a manager who signed instead because the customer was on the phone. That single episode establishes a precedent and the precedent is used from then on. Two named approvers, either of whom can decide, removes the entire failure mode.
Speed also has to be honest about what it is buying. A same-day answer that is always yes is not fast governance, it is fast capitulation with a timestamp. Measure turnaround and refusal rate together, always, because either one alone can be gamed and the pair cannot. A control answering within four hours and refusing a quarter of what it sees is functioning. Any other combination needs a look.
What happens when the approver always says yes?
The threshold stops functioning within about two quarters, and the failure is invisible because the process is still running. Requests go in, approvals come out, the report shows full compliance, and realization continues to drift exactly as before.
This is why the refusal rate is the metric rather than the compliance rate. A healthy control refuses somewhere between a fifth and a third of what reaches it. Below that and the threshold is set too low or the approver is not empowered. Above that and it is set too high and you are wasting everybody’s afternoon.
The remedy is to make the refusal rate visible to the approver themselves, monthly, alongside the realization trend. Nobody sets out to become a rubber stamp. It happens gradually, under pressure, one defensible decision at a time, and the person doing it is usually unaware of the pattern until somebody shows them the count. In my experience showing them the number is sufficient, and no conversation about firmness is required.
There is a related pattern worth watching for, which is the approver who says yes but shrinks the grant. Three points requested, two points approved, every time, without reference to the account. That looks like governance and is actually a negotiating reflex applied internally. It produces a consistent haircut rather than a decision, and over a year it costs almost as much as approving in full while consuming everyone’s time.
Should the threshold be a percentage or a floor?
A floor, expressed in margin rather than in discount, wherever your data supports it. A percentage off list treats a high-margin product and a low-margin one identically, which means the same discount is trivial on one and fatal on the other.
A margin floor asks the only question that matters: does this transaction clear the number below which we do not sell. It also survives cost movement automatically, whereas a discount percentage silently becomes more permissive every time input costs rise and nobody rebuilds the list.
If your standard costs are too unreliable to support a margin floor, use a discount percentage in the interim and treat fixing the costs as the actual project.
Whichever you choose, express it per product family rather than as one number for the business. A single company-wide floor is either too low to protect your differentiated products or too high to let you compete on the commodity ones, and in practice it becomes whichever of those causes the least argument. Three or four floors, matched to how differentiated each family actually is, take an afternoon to set and remove most of the objections before they are raised.
If you do use a percentage, hold it against the current list and rebuild the list annually. A discount threshold anchored to a price list that has not moved in four years is a threshold that has quietly loosened by whatever your costs did in that time, and nobody involved will have noticed, because the number on the policy document never changed.
How do you handle the last day of the quarter?
By removing the incentive that created the moment, and until you do, by refusing to make exceptions on a compressed timeline. A deal that requires a decision in ninety minutes is a deal being used to manufacture urgency, and the urgency is usually internal rather than the customer’s.
Two practical measures. Freeze threshold exceptions in the final week of a quarter, announced in advance, so nobody is surprised. And look at the distribution of your own discounting by week of the quarter. If it spikes in the last two weeks, the pattern is being produced by your forecast process rather than by your market, and no approval workflow will fix that.
The deeper fix is to stop the quarter from being a cliff. Businesses that forecast monthly with a rolling horizon generate far less end-of-period discounting than businesses running a hard quarterly gate, because the pressure is distributed rather than concentrated. That is a planning change rather than a pricing one, and it belongs to whoever owns the operating cadence, but it will do more for your realization than any threshold you set.
The other quarter-end pattern to kill is the retroactive approval. A deal closed at a price below the floor, with the approval requested afterwards because the customer needed an answer and the approver was unreachable. Grant one of those and you have taught the organization that the control is a formality that can be satisfied later, which is functionally the same as not having it. Refuse the first one publicly, and the practice does not recur.
What do you do about compensation?
Change it, last, after the controls have produced evidence. Commission on revenue at any price will eventually defeat every control above it, because the incentive is continuous and the control is episodic.
The change does not have to be dramatic. Paying on gross profit rather than revenue is the cleanest version and it is a large cultural shift. A smaller step that works: keep revenue commission but attach an accelerator above the margin floor and a reduction below it. That preserves the familiar mechanic while making the floor something the salesperson has a reason to care about.
Do this after two quarters of the controls running, not before. You will need the data to defend the design, and your commercial team will need to have seen that the floors are set sensibly before they will accept being paid against them.
One warning about doing this badly. A compensation change announced alongside a new set of controls reads to a sales organization as a pay cut delivered with extra paperwork, and you will spend your credibility on the wrong argument. Separate them by at least two quarters, let the controls demonstrate that the floors are reasonable, and introduce the compensation design as an opportunity to earn more above the floor rather than as a penalty for falling below it.
Whatever you do, change the plan design in a normal planning cycle rather than mid-year. Mid-year compensation changes are read as punishment regardless of their content, and the resulting resistance attaches itself to the pricing work rather than to the plan. The controls can run for two or three quarters perfectly well under the existing plan, and the evidence they generate is what makes the new design defensible when the cycle comes around.
What should you measure weekly?
Four numbers, on one page, taking five minutes in a meeting that already exists. Realization against list by segment. Exception requests received. Exception refusal rate. Average approval turnaround.
The first tells you whether the controls are working. The second tells you whether the threshold is set sensibly. The third tells you whether the approver is empowered. The fourth tells you whether the control is being respected or bypassed, because a turnaround that drifts past a day means people have started going around it.
Weekly for the first quarter, then monthly. A weekly review that persists forever becomes a meeting people prepare for rather than a control.
Add a fifth number once a quarter rather than weekly: the list of accounts that appear above the threshold most often. That list is the actual finding. In most businesses five or six accounts generate a wildly disproportionate share of all exception requests, and the right response is not a faster approval process but a repriced relationship. The weekly numbers keep the control honest. The quarterly list tells you what to go and fix.
What does this cost you in volume?
Some, and you should say so out loud before you start rather than discovering it in month three and losing your nerve. A business that stops discounting discretionally will lose transactions it previously won, and the ones it loses will be concentrated in the most price-sensitive part of the book.
Size it in advance with the breakeven arithmetic. At a thirty percent margin, a two point improvement in realization leaves you better off unless you lose more than roughly six percent of the volume it applied to. Run that against the accounts that sit above your threshold most often, and you will usually find the trade is comfortable by a wide margin.
What you must not do is judge the programme on volume alone in the first quarter, because the volume effect arrives before the margin effect and the two are measured in different places.
Note the expiry date on the grant, because it is the element everyone omits. A discount approved without an end date is a permanent price change made by somebody who thought they were making a temporary exception, and in four years it will appear in your realization analysis as unexplained drift.
One limitation worth naming. Everything above governs discretionary discounting, which is the largest but not the only source of price erosion. It does nothing about a list price that has not been rebuilt against current cost, and a perfectly disciplined organization selling from a stale list will hold its discipline all the way to a bad margin. Rebuild the list annually and then govern the discounting. In that order.
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.