Everyone knows discounting costs margin. The part I did not see for years is that it also costs the number it was supposed to save, and the mechanism is arithmetic rather than willpower.
Managers chasing a monthly number discounted heavily in the last week to get deals closed. Each individual decision was defensible. The deal was real, the quarter mattered, the concession was small, and the alternative was a miss.
Then the market learned. Pressure applied in week four worked, so pressure got applied in week four. Within a few cycles it was not our behaviour any more, it was part of how our buyers bought.
“It is three points. We will make it back on the next one.”
“They were always going to close. This just gets it in this month.”
“Everyone discounts at month end. That is the business.”
“We need approvals on this.”
The last one is the one that felt like a fix and was not. Approval discipline decays under pressure, and the pressure is at its highest in exactly the week the approvals are meant to hold.
Teaching buyers to wait does not just lower your price. It moves your month. More of the number ends up closing in the final week, because that is when you have trained everyone to transact.
And the final week is the week with the most slippage in it. A deal that was going to close on the twenty-eighth has nowhere to go if it slips. A deal that was going to close on the tenth has three weeks of room.
So the discount buys you this month and costs you the shape of every month after it.
Two companies, same pipeline, same slippage rate in the final week. The only difference is how much of the month has been pushed into that week.
| Closes in week four | Slips | Books | |
|---|---|---|---|
| Company A a month that lands evenly | 35% | 20% | 93% |
| Company B a month trained to wait | 75% | 20% | 85% |
A model, not measured actuals. Both companies slip the same share of their week-four deals. The only variable is how much of the month is sitting in that week.
Eight points of the number, given away by the calendar rather than by the price. And it compounds: further behind, so discount harder, so train the market harder, so more concentration, so more slippage, so further behind.
This is the part that changed how I argue about discounting, and it is not the part anyone expects.
Operating costs are fixed and the profit number is committed, so the whole hit lands on the only discretionary bucket you have: the one that funds next year's product.
Worked illustratively. Take a business at forty points of gross margin, allocating roughly twenty-six points to operating cost, eight to profit and six to reinvestment.
| Base | After a 7-point giveaway | |
|---|---|---|
| Gross margin | 40.0% | 33.0% |
| Operating cost | 26.0 | 26.0 unchanged |
| Profit | 8.0 | 8.0 committed |
| Reinvestment | 6.0 | −1.0 |
An illustrative allocation, not measured actuals. The point is the mechanism rather than the split.
Seven points off a forty point margin is not an eighteen percent haircut on profit. It is the whole reinvestment budget, and then some. The company does not look less profitable that year. It looks the same, and stops improving. There is no line item anywhere called things we did not build this year.
Which also tells you what your real discounting capacity is. It is not your gross margin. It is the discretionary points, and for most companies that is a much smaller number than they think.
First, the obvious version, which helped and did not solve it. Programmatic and measured discounting, and no rep discount without manager approval. Better than nothing, and still leaning on willpower in the week willpower is thinnest.
Then the version that worked: budget the discount and pre-fund it with price. A one-time five percent increase, taken a couple of years ahead of needing it, to make room for a planned five percent discount programme. The discounting was not eliminated. It went from an unbudgeted seven to ten point leak to a budgeted five point instrument with its funding already in place.
It drives volume, the volume makes up the margin, and it usually hits the objective. The same points off, spent at the start of the period instead of the end, stop being a loss and start being demand generation.
That is the whole distinction in one sentence. Announced early, a concession changes what buyers do. Reached for late, it changes only what you keep, and by then the deals it is aimed at were never going to close anyway.
And knowing the leak was seven to ten points is what made five the right size. Pick the number without measuring and you will still leak, just more tidily.
Which gives the question worth taking away from this whole page. Of your gross margin, how many points are actually discretionary? That number, not your margin, is your real discounting capacity, and most leaders have never separated the three buckets well enough to answer it.
If you do not know your realised discount rate by segment, by rep and by motion, you are not pricing it in. You are raising price and still leaking.
Years later, working through a forecasting problem, the real answer arrived and it was not a pricing answer at all.
That is what makes it a reaction rather than a lever. You are behind because the pipeline was not real. You discount to close the gap. And the deals you are discounting are the same unreal deals that put you behind.
You will pick off a couple. That is what keeps the habit alive. But they come in at a lower margin, they are less profitable than the deals you planned for, and there are never enough of them to close the gap. I have not once seen last-minute discounting save a month.
Which is why it fails so consistently. It is not that the concession was too small or arrived too late. It is that the thing being bought was never for sale. A buyer who was price-checking you does not become a customer at a lower price, they become a cheaper price-check.
The spiral was a detection failure being paid off at the register. Deals were called at ninety percent that were never ninety percent, nobody found out until the month was nearly over, and by then the only lever left was price. Once the forecast got honest, the month-end discounting mostly stopped on its own, because there was nothing left to rescue at the last minute.
One is modelled before the period starts and is a demand instrument. One is a reach in the final week and is a forecasting failure wearing a pricing costume. The test is the calendar, not the amount.
Which makes reaching for price in the final week a phantom-pipeline detector rather than a pricing decision. The count of month-end discounts is a reading on forecast quality, not on your price list, and it is the one number I would watch if I could only have one.
It is the cheapest diagnostic in this entire list and almost nobody runs it. Take the last six months, and for each one work out what share of bookings closed in the final week. If that share is climbing, you are not managing a pricing problem. You are watching a forecast problem arrive at the register.
Run the model, free No account. Your numbers stay in the link.