Deal desk

A rep once escalated a 38% discount to me at 4pm on the last day of the quarter with a note that said “customer needs it signed today.” No multi-year, no logo rights, no reason the number was 38 and not 25. I approved it, because the alternative was blowing the quarter on principle. That deal taught me the desk’s real job is not saying no at quarter-end. By 4pm on the last day I have no leverage left. The job is making sure the 38% never arrives unstructured in the first place.

The deal desk is the checkpoint that governs non-standard pricing, discounts, and terms before a quote goes out. In 2025 and 2026 it is also the seam where three systems meet: CPQ that builds the quote, contract lifecycle that papers it, and billing that invoices it. That is the lead-to-cash stack everyone is racing to unify. Most teams reach for a replatform to fix it. Nearly every CPQ complaint I have dug into turned out to be a process problem wearing a tooling costume, and a new tool automates whatever process you already have, including the broken one.

Draw the standard line before anything else

The most important thing a deal desk defines is what never reaches it. If every quote routes through approval, the desk becomes the bottleneck reps route around, and you get the 4pm escalation with no context. So I define a self-serve envelope reps close inside without ever touching me: discount within a pre-approved band, standard terms and contract length, standard config, no non-standard redlines. Everything inside that box closes clean. Everything outside trips a routed approval on a matrix everyone can see before they build the quote.

The matrix is the artifact. Publish it, tape it to the wall, wire it into CPQ. A rep should know the cost of every point they give away before they type it into a quote:

Discount band	Approver	What happens
0-15%	Rep, self-serve	Closes with no desk touch
16-25%	Sales manager	Manager sees it, desk logs it
26-40%	VP Sales + Deal Desk	Desk structures the give-get
Over 40%	CRO + Finance	Margin case required at intake

The margin the desk actually defends

Discounting feels free to a rep chasing quota, because it comes off a number they do not own. It comes straight off gross margin, which someone does own, and the math is worse than it looks. A point of discount comes off the top of a number that already had cost baked in, so it costs more than a point of margin.

Every discount over the self-serve band gets a give-get. Never approve a point without getting something back that the desk can value. Here is the exchange rate I use, so a counter is a structured trade instead of a plea:

The give (rep wants)	The get (desk requires)	Why it holds margin
5 extra discount points	Multi-year commit, 2yr minimum	Locks revenue, cuts churn risk
10+ points	Upfront annual payment	Improves cash, cuts collection cost
Logo / case-study rights	Counts toward the give	Marketing value offsets margin give
Deep one-time concession	Expansion commit or usage floor	Buys back margin on the back end

Fix the process before you buy new CPQ

The loudest 2025-26 pattern is teams unifying CPQ, contract lifecycle, and billing into one lead-to-cash stack. Good direction, wrong first move for most orgs. Replatforming CPQ on top of an undefined process gives you the same mess, faster and more expensive to unwind. Before I let anyone scope a new tool, I make the current process answer four questions in writing:

	Buy CPQ first	Fix process first
Approval thresholds	Configured in a tool nobody agreed on	Written and enforced, then wired into CPQ
Give-get rule	Enforced by rep goodwill	Required at intake, valued by the desk
Approved terms	Re-keyed into the contract by hand	Locked so CPQ output = contract = invoice
Outcome	The old mess, faster and pricier	A working process the tool just accelerates
When CPQ decision comes	First, hoping the tool invents a process	Last, automating a process that already works

When those answers are written and enforced, the CPQ decision gets simple, because I am automating a process that already works instead of hoping a tool invents one. The give-get and the threshold matrix belong in config, not in a rep’s head. Here is the shape of the approval rule I wire in, so routing is deterministic and every point over the band demands a structured trade:

# deal-desk approval matrix (CPQ config)
approval_rules:
  - band: [0, 0.15]        # 0-15% discount
    approver: none          # self-serve, closes clean
    require: [standard_terms, standard_config]
  - band: [0.1501, 0.25]
    approver: sales_manager
    log: true
  - band: [0.2501, 0.40]
    approver: [vp_sales, deal_desk]
    routing: parallel        # never sequential
    require_giveget: true    # multi-year | prepay | logo | expansion
  - band: [0.4001, 1.0]
    approver: [cro, finance]
    require_margin_case: true
    stage_gate: intake        # not at signature

term_lock:
  # what the desk approved is what gets contracted and billed
  source: approved_quote
  propagate_to: [clm_contract, billing_invoice]
  on_drift: block_and_alert

The two rules that save the most pain are routing: parallel and term_lock. Sequential approval is how a deal loses a week: each approver waits for the last. Route them in parallel and a complex deal clears in 24 hours instead of 120. The term_lock rule closes the leak where a rep re-keys terms into the contract and quietly changes the discount the desk approved.

Parallel routingSequential (avoid)

View as table

Stage	Value
Simple exception	4 hrs
Standard complex deal	24 hrs

Here’s how I’d build it: the request lifecycle

Six steps, in order. The first and last are the ones teams skip, and they are the ones that decide whether the desk holds margin or just rubber-stamps it. Structured intake kills the “customer needs it” note; the feedback loop is what makes next quarter’s matrix smarter than this one’s.

1. 1

   1. Structured intake

   Amount, discount, terms, and the give-get the rep is offering, in fields, not free text. A "customer needs it" note is a rejected intake. If the give-get is blank on a deal over 25%, the request does not route.

2. 2

   2. Auto-route by the matrix

   To every required approver in parallel, never one after another. The matrix decides who, the CPQ config decides how. No rep guesses who to email.

3. 3

   3. Desk checks margin and config

   Run the discount against gross margin and verify the CPQ configuration for correctness before it goes out. This is where the 11-point drop gets caught, not after signature.

4. 4

   4. Approve or counter within SLA

   A counter is a structured alternative that hits the customer goal at better terms: a smaller discount plus a services attach, or the same discount for a multi-year commit. Escalate on SLA breach, do not let it sit.

5. 5

   5. Lock approved terms into CPQ

   What was approved is what gets contracted and billed. The term_lock rule blocks drift between quote, contract, and invoice. This is the leak that undoes the margin you just fought for.

6. 6

   6. Log discount, rationale, cycle time

   Feed every approved exception back to pricing. If a discount band gets requested every week, the matrix is wrong and the standard price should move. The log is how the envelope tunes itself.

Target the majority of deals, 70 to 80%, closing inside the self-serve envelope. If most deals need the desk, the envelope is too tight and I widen it. If discounts creep every quarter, the give-get rule is not being enforced. The desk that governs discounts well feeds the operating cadence too: fewer unstructured exceptions means fewer surprises in the weekly forecast inspection. For that side, see the operating cadence guide.

The 38% at 4pm was not a pricing failure. It was a process failure that surfaced as a pricing emergency. Publish the matrix, require the give-get at intake, lock the approved terms through to billing, and the emergency never forms. The desk stops being the person who says no at quarter-end and becomes the process that made no unnecessary.