A discount looks like a local decision. This deal, this quarter, this one client who is nearly there. The arithmetic supports it: the marginal deal at a lower price is still better than no deal, and that is usually true of any single instance.

It is the series that does the damage, because the buyer is running a series too.

The mechanism

Round one. A discount closes a deal that was stalling. It works, which is the problem — it is now a known tool.

Round two. The same tool is reached for earlier, because it worked. The concession that was a closing move becomes a negotiating position.

Round three. Buyers learn. The discount is now expected, so the list price is read as an opening bid rather than a price. Sophisticated buyers wait for it; unsophisticated ones hear about it from sophisticated ones.

Round four. The discounted price is the real price and the list price is decoration. The margin is permanently lower, and the next concession has to come out of a base that has already been cut.

Each round is individually rational. The sequence is not, and nobody is in a position to see the sequence except the person looking at annual realized price — which is why the pattern is usually diagnosed from a spreadsheet a year late.

Why the margin arithmetic is worse than it feels

On a 30% gross margin, a 10% discount removes a third of the profit on that sale. Not a tenth — a third, because the discount comes entirely out of the margin and not out of the cost.

Which means the volume required to compensate is large and is usually not calculated. Recovering the profit lost to a 10% discount at that margin requires roughly 50% more volume, at a cost base that also has to grow. The deal that was closed on the argument that some margin beats none was, in aggregate, an argument for doing considerably more work for the same money.

Why sales incentives accelerate it

If the person negotiating is measured on revenue closed rather than on margin, discounting is free to them and expensive to the business. They will discount, correctly by their own measure, and the metric will look excellent while the thing it was meant to proxy for gets worse. This is Goodhart's law in its most common commercial form.

The fix is not exhortation. It is either measuring margin instead of revenue, or removing discounting authority from the person carrying the number.

Getting out

Make concessions structural, not numerical. If a price has to move, move something else with it — a longer term, a smaller scope, a faster payment, a case study, fewer revisions. A concession that costs the buyer something teaches a different lesson than one that does not, and it does not reset the price.

Hold a published floor. A price below which the answer is no, decided in advance and not in the room. The value of deciding in advance is entirely that the room is a bad place to decide.

Walk from one. The single most effective act, and the one most avoided. A business that has never walked away from a deal on price has no evidence that its price is real, and neither do its buyers.

Change what is being priced. If the discount is being demanded because the buyer is comparing your day rate to someone else's day rate, the problem is the comparison, not the rate — which is commoditization and has a different answer.

The asymmetry worth knowing

Prices are much easier to lower than to raise. A discount is granted in one conversation; recovering it takes a year, a reason, and usually the loss of the clients who came for the discounted price.

Which is a reason to treat the first concession in a relationship as the expensive one, whatever its size. It is not setting a price for that deal. It is setting an anchor for every conversation after it.