A working deal desk is four things written down: what triggers a review, who approves what, where the record lives, and how fast you promise to answer. Everything else is commentary. This guide is the rulebook itself, the discount bands, the contract review path, the CRM approval-workflow features that make it hold, and the numbers that tell you it is working.
If you are still deciding whether you need the function at all, that is a different question and we answered it separately in do you need a deal desk at $5M ARR. This page assumes the answer was yes and you are building the thing.
The approval matrix is the whole rulebook
Start with one page that says: at this discount depth and this deal size, this person approves. Not a process map, not a Lucidchart. One page.
The published matrices agree on the shape and disagree on the numbers, which is the right way round. Pulse RevOps describes a three-tier mid-market standard of AE or manager up to 15%, RevOps plus finance to 25%, and CRO plus CFO above that (Pulse RevOps). A more granular published version routes 20 to 30% to a regional VP plus the deal desk lead on an eight business hour SLA, and anything past 30% to the CRO, CFO and sometimes general counsel on a 24 business hour SLA with a written strategic rationale (Pulse RevOps). Others express the same thing in dollar impact rather than percent: manager below a $10K margin hit, VP below $50K, CFO above.
Glencoyne's version is the one worth copying structurally, because it uses two axes rather than one. Discount depth runs in bands, shallow, moderate, deep, against deal size drawn from your own deal histogram rather than a round number, with each cell carrying an approval tier and exactly one accountable approver per threshold (Glencoyne). Building the size axis from your real distribution is the part people skip, and it is why copied matrices misfire: a $50K threshold means something different in a business whose median deal is $18K.
Two rules about the matrix itself. Three to four tiers maximum, because reps stop reading anything longer. And the matrix has to cover more than discount depth, which is the failure we see most often.
What the matrix has to cover besides discount
Headline price is the leak everyone watches. The expensive ones are quieter.
- Payment terms. Net 30 against net 90 against annual prepaid. A payment term change is a financing decision and finance has a real stake in it.
- Contract length and ramp structure. Multi-year discounts stack in ways people do not model. Published 2026 benchmarks put the monthly-to-annual discount median at 15 to 20%, with a top quartile of 20 to 25%, then a further 5 to 8% beyond annual for a two-year term, 12 to 15% for three years, and 18 to 25% for a five-year strategic contract (GrowthSpree). Approve the annual discount and the term discount independently and you can hand over a third of list without anyone deciding to.
- Cancellation rights and outs. A twelve-month deal with a 90-day out is not a twelve-month deal. Price it as what it is.
- Custom SLAs and liability terms. A buyer rewriting your uptime commitment or liability cap needs legal eyes before you agree, not after.
- Free pilot extensions. The most commonly ungoverned concession, because it does not look like a discount.
It is also worth separating the two kinds of discount in the rulebook, because they deserve different approvers. DealHub's framing is the clean one: strategic discounts advance a long-term goal such as market entry, a marquee logo, or a competitive displacement and should require executive approval, while tactical discounts handle routine competitive pressure and stay inside the standard bands (DealHub). Mixing them means either every routine deal goes to an executive, or every strategic concession gets waved through on a tactical rule.
Who owns which part
The most useful published statement of the split, and the one that prevents the two common failure modes, is this: RevOps owns the process and the SLA, finance owns pricing thresholds and margin rules, legal owns non-standard terms, and sales leadership owns the escalation path for exceptions (DealHub).
Read it as a warning about reporting lines. A desk owned by sales approves everything. A desk owned by finance rejects everything. The function is independent judgment on margin, which means it sits between them and reports to neither.
What runs through the desk, and what must not
The fastest way to build a deal desk badly is to route everything through it. Then it becomes the committee everyone feared and reps start working around it.
Notice what is not on the list. A standard discount inside the rep's band. A clean annual contract at list price. A renewal with no changes. Those should never see the desk. If your triggers are catching half of all deals, the thresholds are wrong and you are taxing the whole pipeline to govern the fraction that carries real risk. The design goal that most modern write-ups converge on is that the large majority of deals auto-approve and only edge cases escalate (HelloGrowthCRM).
What deal desk and approval workflow features to look for in a CRM
You do not need a deal desk platform to run this. You need your CRM to do five specific things. This is the checklist we use when someone asks what to look for, and it is deliberately short.
| Feature | What it has to do | Why it matters |
|---|---|---|
| Intake on the deal record | A small set of fields on the opportunity or deal itself, not a form in a separate tool | If finding the request is half the cycle time, the desk is already slow |
| Conditional routing | Route on discount depth, deal size and term together, not on one variable | A single-variable rule cannot express a two-axis matrix |
| Approval logged back to the record | The decision, the approver and the timestamp written onto the deal | No audit trail means no cycle-time data and no way to learn from patterns |
| Parallel approvals | Finance and legal reviewing at the same time where both are needed | Serial chains are where the days go |
| Post-close field carry | Approved non-standard terms copied to renewal and handover records | Otherwise CS learns what was promised at the kickoff call |
Two of those are worth expanding. Parallel rather than serial approval is the single highest-leverage configuration choice, and the published guidance is explicit that approvals should run in parallel wherever possible with exactly one accountable approver per threshold (Glencoyne). And the context the approver sees matters as much as the routing: when a contract enters review the desk should automatically surface the deal record alongside it, customer tier, exceptions already approved, the non-standard terms the rep flagged, and the clause-level risk policies that apply to that deal type (Scopic).
Both HubSpot and Salesforce do all five natively. The intake is deal properties or a custom object, the routing is a native workflow or an approval process, and the Slack integration is where the approver actually answers. This is configuration, not a purchase, and it is the same sales process automation plumbing you already use for stage gating.
If reps genuinely cannot build a valid quote by hand because the catalogue is too complex, that is a CPQ conversation rather than a deal desk one, and we wrote a separate guide on when CPQ is actually worth it. CPQ stops bad quotes being built. The desk governs the exceptions to your pricing. You can need one without the other.
The SLA is half the value
A deal desk with no turnaround commitment is just a new place for deals to wait. Put a number on it and publish it.
Four business hours for routine exceptions and same-day for anything with a hard close date is a reasonable starting point. The published matrices attach an SLA to each tier rather than one number to everything, eight business hours at the middle tier and 24 at the deepest (Pulse RevOps), which is the better pattern because it lets you promise something fast on the cases that are actually routine.
A turnaround promise on routine exceptions, published and measured. This is what separates a desk reps use from a desk they route around. If you cannot commit to a window and hit it, fix the process before you announce the function.
Reps will trust a desk that answers fast and work around one that does not. They always do. The correct framing, and it is not just a slogan, is the desk as an enablement function rather than a policing one: reps should see it as the fastest route to an approved deal, not the biggest obstacle (DealHub).
What to measure
Five numbers. Resist adding more.
Approval cycle time, median and 90th percentile. From intake submitted to decision logged. Track both, and treat the 90th percentile as the real number, because the tail is where reps lose patience and bypass the process.
Desk capture rate. The share of deals over the threshold that actually went through the desk. Below 95% and reps are routing around you. Fix the SLA, not the reps. A slow desk is the usual cause.
Discount depth, by rep and by segment. Total discount against list, then split three ways: inside the rep's band, between band and floor, and below floor. The below-floor number is the one finance will ask about. Track it by rep, region, deal size, segment and product line so you can see patterns rather than incidents (DealHub).
Price realisation against the benchmark. Realised revenue as a share of list. The published average across the $5M to $50M B2B SaaS band is 84% of list price, and the same source attributes 40 to 60% of total discount volume to habitual discounting, applied reflexively because the rep expects a negotiation or because the approval path is unstructured enough that discounting is the easiest route (TechGrowth Insights). That second figure is the one a deal desk can actually move, because habit responds to a written rule in a way that competitive pressure does not.
Win rate, desked against non-desked. Compare deals that touched the desk with comparable-ACV deals that did not. If desked deals win less often, the desk is killing deals that should have closed. If they win more but at materially deeper discounts, reps have learned that escalating gets a price break, and you have a discipline problem rather than a process one.
One caution on benchmarking the function itself. The figures you will see everywhere, 25 to 40% shorter cycles and 15 to 20% better productivity attributed to PwC, trace back to a consultancy describing its own client experience rather than to a published study. We set out that trail in the companion guide. Measure your own before and after instead.
The five ways deal desks fail
Single point of failure. The desk is one person, who then takes leave. Name a backup approver in the matrix and write the runbook.
No floor price. Tiers but no absolute floor, so reps push to the next tier every quarter until tier three looks like tier one used to. Set an absolute floor per product, reviewed annually by the pricing committee rather than by sales leadership alone.
Approvals in DMs. No audit trail, no metrics, no learning. Enforce intake on the deal record and decline approvals that arrive any other way. Painful for a quarter, normal after that.
The desk becomes the bottleneck. Cycle time creeps from hours to a week and reps bypass it. Watch the 90th percentile, not the median, and add capacity when the tail passes 72 hours.
Nothing captured for after the close. Non-standard terms live in someone's memory. At renewal the customer says they were promised an exit at 60 days and nobody can prove otherwise. Every approval and every non-standard term goes on the deal record.
What happens after the deal closes
The part everyone forgets. The desk does not stop at signature, and this is where a well-run one earns trust outside sales.
- Finance handoff. Non-standard payment terms, ramp schedules and milestone billing have to land in the billing system correctly. The desk owns the translation from sales terms to billing terms.
- Customer success handoff. A promised custom SLA, a dedicated CSM, a training credit: if it does not reach the CS system, CS finds out at the kickoff call and the relationship starts with an apology.
- Renewal context. What was discounted, and why. Strategic logo, competitive replacement, multi-year commitment. The AE inheriting the account twelve months later needs the reason, not just the rate.
Build this as a closed-won workflow that copies the deal desk fields into renewal tracking, notifies CS, and stamps the non-standard terms onto the customer record. It is a contained piece of automation work and it runs indefinitely.
Approvals turning into scavenger hunts?
Book a free 30-minute audit and we will show you the matrix and the CRM configuration we would put in first, mapped to your own deal distribution.
Book an audit →How we approach this at Ziel Lab
Most of our CRM and RevOps engagements that start as a deal desk project are really a governance project. The rules do not exist, so we write them: the trigger list, the two-axis matrix built from the client's own deal histogram rather than a template, the floor prices, and the SLA per tier.
Then it goes into the CRM they already pay for. In HubSpot that is deal properties for requested discount, approval status and approver, native workflows routing on tier, Slack for the answer, and a closed-won workflow carrying the approved terms to renewal and CS. In Salesforce it is the equivalent approval processes on the opportunity. Two to four weeks, and the team runs it themselves afterwards.
Where routing genuinely outgrows native workflows, multi-product approvals, ramp modelling, segment-based routing, we layer AI automation on top rather than replacing the CRM. The order is always rules first, then a human owner, then automation. Automate before the rulebook exists and you have automated the confusion.
Frequently asked questions
What are the discount approval bands most B2B teams use?
Three tiers is the common shape: the rep or their manager up to roughly 15%, RevOps plus finance to around 25%, and the CRO plus CFO above that, with a written rationale required at the deepest tier. Some teams express the same thing in margin dollars rather than percent. The better practice is two axes, discount depth against deal size drawn from your own deal histogram, with exactly one accountable approver per cell.
What should the deal desk govern besides discount?
Payment terms, contract length and ramp structure, cancellation rights, custom SLAs and liability terms, and free pilot extensions. Term discounts are the quiet one: published 2026 benchmarks put the monthly-to-annual discount median at 15 to 20% with a further 12 to 15% for a three-year commitment, so approving those two independently can hand over a third of list without anyone deciding to.
What deal desk and approval workflow features should we look for in a CRM?
Five: intake fields on the deal record itself, conditional routing on discount depth and deal size and term together, the decision and approver written back to the record, parallel rather than serial approvals where more than one function is needed, and post-close carry of approved terms into renewal and handover records. HubSpot and Salesforce both do all five natively. If a vendor is selling you a sixth thing, check whether it is really CPQ or contract lifecycle management wearing a deal desk label.
Who should own the deal desk?
RevOps owns the process and the SLA, finance owns pricing thresholds and margin rules, legal owns non-standard terms, and sales leadership owns the escalation path. The desk itself should report to RevOps, the COO or the CRO. Reporting into sales means every exception gets approved. Reporting into finance means every exception gets rejected.
How fast should a deal desk answer?
Attach an SLA to each tier rather than one number to everything. Four business hours for routine exceptions, eight for the middle tier, 24 for anything needing executive sign-off, and same-day for anything with a hard close date. Publish the promise and measure compliance against it, because that is the number reps actually feel.
What should we measure?
Approval cycle time at median and 90th percentile, desk capture rate, discount depth split by rep and segment and by band against floor, price realisation against the 84% benchmark for $5M to $50M SaaS, and win rate on desked against non-desked deals once the basics are stable. Five numbers is enough. Watch the 90th percentile rather than the median, because the tail is what drives reps to route around you.