In an earlier article on discounting in CPQ, I explained the mechanics: schedules, manual discounts, and approval thresholds. Today I want to go one level deeper, into the design question the mechanics cannot answer for you. Where do the thresholds go, who approves what, and why do so many approval processes end up being quietly ignored by the very people they were built to govern?
This one comes from a real build. In my TechnoStore project I designed a multi-tier discount approval matrix: discounts above 30% route to the Finance Director, and above 50% they go to the VP. Simple numbers on paper. But the thinking behind those numbers is the actual lesson, so let me walk you through it.
Why “manager approves everything” fails
The most common approval design I see is also the laziest: any manual discount, of any size, routes to the rep’s manager. It feels safe. Everything gets a second set of eyes. What could go wrong?
Two things, and they feed each other.
Rubber-stamping. When a manager receives forty approval requests a week and thirty-five of them are routine 5% discounts, they stop reading. Approval becomes a reflex click between meetings. And here is the dangerous part: the reflex does not pause for the one request that actually mattered. The 45% discount sails through on the same muscle memory as the 5% ones. You built a checkpoint and trained everyone to sprint through it.
Deal delays. Meanwhile, every small, harmless discount now waits on a human. The manager is traveling, the quote sits for two days, the customer cools off. Reps learn that the approval process is where deals go to get slower, and they start resenting it.
An approval step that fires on everything protects nothing. Attention is a budget, and a flat rule spends all of it on requests that never needed it.
The fix is not more approvals or fewer approvals. It is placed approvals.
Design thresholds around real margin risk
Here is the question that should decide every threshold: at what discount level does this deal start genuinely hurting us, and who in the company actually understands that harm?
In the TechnoStore matrix, that thinking produced two tiers:
- Above 30% → Finance Director. Around this level, a discount stops being negotiation room and starts eating into real margin. The person who signs off should be someone who can read the deal’s profitability, not just its optics. That is a finance role, not a sales role. A sales manager is measured on closed revenue; asking them to be the margin guardian is asking them to argue against their own bonus.
- Above 50% → VP. At half price, this is no longer a discount. It is a strategic decision: are we buying a logo, entering a market, rescuing a relationship? Decisions like that belong to someone with the authority to own the consequence, which is why the escalation jumps to the VP rather than one more rung up the finance ladder.
And everything below 30%? It flows through without ceremony. That is not looseness. That is the design saying, clearly, “this range is your professional judgment, and we trust you with it.” The approvals that remain are rare enough that approvers actually read them.
Your numbers will differ. A software business with 85% gross margin can tolerate thresholds a hardware reseller cannot. The method is the same: find where the real financial pain begins, and put the gate there, guarded by the person who feels that specific pain.
Reps should predict the path before they submit
Here is a test I now apply to every approval design: can a rep, looking at their quote, tell you exactly who will approve it before they click submit?
If the answer is no, the matrix has failed, even if the routing logic is flawless. An unpredictable approval process feels arbitrary, and people do not respect arbitrary systems. They fear them, then they resent them, then they game them.
A legible matrix does the opposite. When a rep knows “under 30% is mine, over 30% means the Finance Director will read this,” two useful things happen. First, they self-regulate: many will land at 28% instead of 32% and simply close the deal. Second, when they do go above the line, they come prepared, because they know a specific, named person will be reading their justification. The threshold teaches, quietly, before anyone submits anything.
So publish the matrix. Put it in onboarding, pin it in the sales channel, keep it to numbers a person can hold in their head. Two or three tiers is a matrix. Nine tiers with regional exceptions is a puzzle, and nobody respects a puzzle.
Automate the routing, and pull in context
Once the thresholds are designed, the machinery should be invisible. The rep submits; the system reads the discount and routes the quote to the right approver with the context they need to decide quickly. No spreadsheets, no forwarding emails, no asking around about who signs off this month.
In the TechnoStore build I paired the approval matrix with Expression Sets handling account-context-aware product qualification: declarative logic that evaluates the account’s context to determine what the customer actually qualifies for. That matters to approvals more than it first appears, because it moves a whole category of questions out of the approval queue entirely. The system already knows what this account qualifies for, so the approver is never burning attention on “should this customer even be offered this?” The human checkpoint stays focused on the one question only a human can answer: is this margin sacrifice worth it?
That is the pattern I would hand any beginner: let declarative logic answer every question that has a rule, so the approval step only carries questions that need judgment.
The cultural part nobody configures
Now the uncomfortable truth. You can build a technically perfect matrix and still fail, because an approval process reps route around is worse than having none at all.
Worse, because it adds all the friction while providing none of the protection, and it teaches your team that official process is theater. Watch for the symptoms: discounts clustering suspiciously at 29.9%, deals split into two quotes to stay under a threshold, “special pricing” negotiated in side conversations and pasted in afterward. Each one is a rep telling you the system fights them instead of helping them.
When you see that, resist the reflex to add enforcement. Ask why the path around is more attractive than the path through. Usually the answer is one of three things: approvals are too slow, thresholds sit below what the business routinely accepts anyway, or approvers reject without explanation so reps stop asking. All three are design problems, and all three are fixable.
The healthy end state is a matrix reps treat as a colleague. Fast on the deals that are fine, thoughtful on the deals that are risky, predictable everywhere. I spent twenty years teaching before Salesforce, and the classroom rule holds here perfectly: rules people understand and consider fair barely need enforcing. Rules that feel arbitrary get tested every single day.
Your next step
If you have not read it yet, Discounting in CPQ covers the underlying mechanics of schedules, manual discounts, and approvals. To see where reps experience all of this, read The CPQ Quote Line Editor, and for the rule engine that can automate pricing logic before approvals ever fire, see CPQ Price Rules. The full CPQ series is here when you are ready to go further.