A deal desk is the cross-functional group that reviews and approves non-standard deals: discounts beyond a threshold, unusual payment terms, custom contract language, bespoke bundles. It usually draws on sales operations, finance and legal, and its approvals set the price, the terms and therefore the basis that commission is calculated on.
Someone has to say yes to the 18% discount. In a small company that someone is the VP of Sales, answering on a phone call, using judgement built from the last forty deals. That works until volume or complexity outruns one person's attention, at which point the answers start disagreeing with each other, the same customer gets two different prices, and a contract goes out with a payment schedule nobody in finance has seen. A deal desk is the response: one queue, a documented approval matrix, and a defined group that owns the answer.
What it approves lands directly on top of the commission calculation, which is why the two functions belong in the same conversation rather than in different quarters. A discount changes revenue and changes margin by a larger proportion. A payment schedule moves the month a billing-triggered plan pays in. A multi-year commitment with a ramped first year changes the annual contract value a plan measures against. Custom termination language changes what a clawback clause can reach. Every one of those is a deal desk decision that the rep will feel on a statement, and if the approved terms live only in a contract document while the pay run reads standard fields, the payout is calculated against terms that were overridden.
The failure mode worth watching for is the desk becoming a queue rather than a decision. Reps submit, wait, chase, and learn to submit everything at the highest discount they might need, because a second trip through the queue costs three days. The desk then sees a distribution of requests that has been shaped by its own latency rather than by what the deals require. Published turnaround times, a clear threshold below which no approval is needed at all, and a standing answer for the three most common requests remove most of that, and none of them require anybody to be more disciplined than they already are.
A rep requests 18% off list on a $200,000 deal. The desk approves 12%. The rep is paid 10% of gross margin, and the cost of delivering this configuration is $110,000.
On a revenue plan paying 3%, the same approval moves the rep by $360 rather than $1,200. Margin plans make the desk's judgement matter more than three times as much to the person selling, which is a decent argument for telling reps what the desk decided and why.
The desk approves the 12% discount and it goes into the contract. The CRM opportunity still carries the list amount, because updating it was nobody's named job, and the pay run reads the CRM. The rep is paid on $200,000 of a deal sold for $176,000, and the error is found during an annual reconciliation eleven months later, by which point recovering it means a clawback on a deal the rep did everything right on. The approval has to write back to the field the calculation reads, and a check that flags any closed-won deal whose value differs from its approved terms costs very little to run.
The long form lives in the guides: Commission structures.
Commish pays deal desk the way your plan describes it, shows the arithmetic on every line, and traces each payment back to the deal that earned it.