Where should your discount thresholds actually sit?
Every discount policy we have seen was written the same way: somebody picked round numbers in a meeting. Ten, twenty, thirty-five. They are reasonable numbers. They are also completely disconnected from the one input that decides whether the policy will survive contact with a quarter — how much your team already discounts.
That disconnect is why so many approval ladders end up abandoned. Not because reps are cavalier, and not because the thresholds were too strict in principle, but because the first rung was set below the discount half the team's quotes already carry. From day one, the exception became the default path, and a process that fires on most quotes is not governance. It is a queue.
Start with the distribution, not the round number
Before choosing a single threshold, export the last two quarters of quotes and find one number: the median discount on quotes that carried any discount at all. Not the mean — a couple of desperate 60% deals will drag a mean somewhere useless. The median is the discount half your discounted quotes exceed, and it is the fulcrum the whole policy balances on.
Then look at the shape around it. In most B2B catalogues, discount depth decays roughly geometrically: if the median discounted quote is at 12%, then something like a quarter sit above 24%, and an eighth above 36%. You do not need this to be exact. You need it to be approximately true, because it tells you what each candidate threshold would cost in routing volume before you commit to it. Our discount policy builder does this arithmetic from three inputs and prints the assumption openly, which is the only honest way to publish a model like this.
Rule one: the first threshold goes above the median
This is the decision that matters more than the other three combined. Put the self-serve limit above the typical discount and the policy is invisible for most quotes — reps feel nothing, approvers see only the deals worth their attention, and the process earns the right to be strict where it counts. Put it below, and every quote becomes a request. Reps respond exactly as you would: they pad the opening number to leave approval room, which raises your average discount, which is the opposite of what the policy was for.
A useful target is that at least 70% of all quotes never route anywhere. If your first threshold cannot reach that, the answer is almost never "add another approver". It is that the threshold is in the wrong place, or that your list price is.
Rule two: three rungs above self-serve, at most
The instinct when a policy feels leaky is to add a rung. Resist it. Every rung adds latency to the deals under the most time pressure, and it does so at exactly the moment a competitor is also in the room.
A working shape looks like this — and note that the interesting column is the third:
| Band | Decides | Roughly how often |
|---|---|---|
| Up to the median + a margin | Nobody — self-serve | 70–85% of quotes |
| Next band | Sales manager | 10–20% |
| Next band | Regional director | 3–8% |
| Top band | Finance | 1–3% |
| Above the ceiling | Nobody — refused | <1% |
If a rung would decide fewer than a couple of quotes a month, it is not a control, it is a delay. Fold it upward. The director band is the usual casualty, and losing it costs almost nothing because the quotes it would have caught are already visible to Finance.
Rule three: the ceiling is not a threshold
Thresholds route: they decide who says yes. A ceiling states that a deal does not happen at that price, and its entire value comes from the fact that nobody can move it in the last week of a quarter. A ceiling with an override is a threshold wearing a costume — it will be overridden precisely when the pressure is highest, which is the case it existed for.
The test is simple and slightly uncomfortable: can your organisation name a discount it would refuse even on a logo it badly wants? If not, you have a routing policy and no floor under it. That is a legitimate position — but then say so, and put the enforcement effort into the margin floor instead.
Rule four: flag the margin floor, block the ceiling
A quote below your margin floor is sometimes the right call. Strategic logos, competitive displacements, and land-and-expand deals all justify thin first-year economics. The policy's job is not to prevent those; it is to make sure they are a decision with a name against them rather than something that surfaces in a QBR three months later.
So: the floor flags and records, the ceiling refuses. Keeping those two mechanisms distinct is what stops a policy from being either toothless or unusable. It also means the margin floor quietly does the work that per-product thresholds would otherwise do — a deep discount on a thin line trips the floor even when the percentage sits comfortably inside its band.
What to measure once it is live
Two numbers in the first month, and they are in tension on purpose:
- Self-serve share — the percentage of quotes that never routed. Falling below 70% means the first threshold is too tight, not that reps are misbehaving.
- Median approval latency, per rung — if one rung is slow, it almost always has one approver rather than too much process. Add a second person with the role before you touch the thresholds.
After a quarter, add a third: win rate by discount band. It is the number that tells you whether the deep discounts you approved actually bought anything. In more cases than anyone expects, the answer is that the deal closed for reasons the discount did not change — and that finding is worth more than the policy.
The part that is not about thresholds
You can get all four rules right and still end up where you started, because a policy that lives in a document is enforced by memory. The ladder only becomes a control when something refuses: a rep cannot approve their own quote, a step cannot be taken out of order, and past the ceiling nothing finalizes at all. Until then, what you have is a strongly worded suggestion with an audit gap in the middle of it.
Which is the honest reason to write the policy down first, before buying anything. Run it unenforced for a quarter and measure how often it is followed. That number is your business case — for us, or for whoever else you end up shortlisting.