The Discount Everybody Applies and Nobody Approves

It is worth being precise about this because a discount is often treated as a field, and a field is not a decision. A field holds a number. A decision holds a number plus the conditions under which it was available, the person who could exercise it, the person who did exercise it, and when. Those four things are what separate a discount from a correction.

The person matters first, and not only for accountability. Who applies a discount reveals what the discount is actually for. If the same person applies small discounts on every sale, it is a commercial habit and the real question is whether it belongs in a rule. If it rotates across the counter, the discount is being used to close transactions, and the counter is now running a pricing policy nobody wrote down. If it always comes from one manager, it is already a decision and it merely lacks the record.

The rule matters because it decides which discounts need a second person. A business with a written threshold has an answer ready for every case, including the ones nobody anticipated, and the answer is available in a second rather than by telephone. A business without a threshold has to have a conversation every time, and conversations at a counter are the mechanism by which inconsistency enters a business: the same customer gets two different treatments depending on who is standing there, and both decisions are defensible in isolation.

The approver matters because that is what makes it a decision rather than an action. A person approving something is making a judgement about a business outcome — is this worth the margin to keep the customer — and the value of the record is that it can be looked at later. Which of them, over a month, cost more than they were worth. The time matters last but is not optional, because an approver who cannot be asked what they approved has approved nothing; they have received a notification and dismissed it.

Here is why this is worse than it looks. Every other error at a counter is a discrepancy between two records that both exist, and discrepancies are detectable because there is something to compare. A discount with no approver is not a discrepancy. It is a perfectly valid reduction, correctly calculated, correctly taxed, correctly recorded, and indistinguishable in every respect from a decision that was properly made.

That is exactly what makes it dangerous. It will not surface in reconciliation, because the totals agree. It will not surface in the stock count, because the movement is correct. It will not surface in an invoice check, because the invoice is right. The only place it surfaces is a review of what people did, and a review is a thing that only happens if somebody has decided to look, which is precisely the thing that is missing.

The practical consequence is that the discount stops being a decision made under a rule and becomes a component of the price. Which means it stops being visible as a cost. Margin analysis, if you do it, will show a lower margin on a category of sales and the natural reading will be that the pricing is wrong, when the actual cause is that a number nobody approved is now the price. A business can spend a year repricing because of a habit that was never a policy.

There is a second-order effect on the people at the counter, and it is the one that keeps the habit alive. A discount applied with no rule is a discount the operator knows they can get away with. That is not a moral observation, it is an observation about what the system teaches. Once nobody has ever been asked to justify a discount, the absence of a question is experienced as permission, and the next person at the counter learns the same thing from the same silence. A control that has never once been observed to work is indistinguishable from permission.

First, there is a written limit, and it is a number rather than a description. Anything describable in words will be interpreted differently by four people, and a rule that requires interpretation is not a rule. The number needs to be on the same scale as the transactions it governs, which means deciding whether it applies per line, per invoice, per customer, or per month per customer, and writing that down, because the four give completely different answers on the same sale.

Second, the limit is enforced by permission rather than by knowledge. If applying a discount inside the limit requires a capability the operator has and applying one outside it requires a capability they do not, the boundary is structural and it holds on a busy Saturday. If it requires the operator to remember the number, it will not hold on a busy Saturday, and nobody is at fault when it fails.

Third, above the limit there is a second person, and the second person is identified. Not a role in a policy document that names no one; a person who exists, is reachable at the counter, and has the capability. A rule whose approver is on leave is a rule that will be broken or a queue that will stop the counter, and the correct response is a named backup rather than a hopeful assumption.

Fourth, the reason is captured when the discount is outside the routine. A reason is not needed for every discount — that is what the threshold is for, and requiring a reason on every small discount is how you get a reason field full of the letter x. A reason is needed for the ones that are unusual, and unusual is precisely the set worth being able to read next month.

Fifth, the review is scheduled, which is the part everybody skips and the part that makes the rest worth doing. Without a recurring read of the discount records, the limit is a gate and nothing more. With one — say, every discount over the limit, listed, with the operator, the approver and the amount — a pattern becomes visible within weeks: one operator and one customer account, or one product family where the price is genuinely wrong. That review is also the only mechanism by which anybody ever finds out that a control works, and a control nobody has ever observed to work will be quietly removed by the first person who finds it annoying.

Two boundaries to state here rather than leave implied. The first is about what a discount is not. In NoxOrigin there is no change-request record, no milestone record, no deliverable record, no contract editor and no e-signature. If the thing being sold is project work and the customer asks for a reduction in scope, that is not a discount to be approved at a counter; it is a new quote raised against the same project, and the old quote stands as it was. Applying a percentage to an invoice is a different act from renegotiating what is being delivered, and conflating the two is how a scope change ends up as a small unexplained number on a document.

The second is about people, and it is a boundary of the product rather than of the method. There is no timesheet record and no payroll module in NoxOrigin. That means a discount record can tell you what an operator did at the counter, and it cannot tell you their hours, their shift, or their pay, and nothing here should be read as though it could. The question of who was on the till and when they were on it has its own shape and its own article on shifts and the business day, and the honest position is that the discount record answers who did this thing, while the business day answers when the day was open, and neither one substitutes for the other.

The last thing worth saying is that a discount policy is a margin policy. Once discounts are recorded properly, the question stops being who gave what away and becomes which kinds of discount pay for themselves. A discount to keep a customer is a customer-acquisition cost that happens to be booked at the counter. A discount to move stock is a stock decision, and stock movement and purchasing live in the Commerce area rather than in the invoice. A discount applied because the price was wrong is a pricing correction. None of those is a counter decision, and recording them as one is how a pricing problem gets reviewed by people who cannot fix it.