What's the right architecture for discount governance when a company spans both sales-led enterprise and PLG SMB motion — should they operate entirely separate approval chains or integrate them in 2027?

The company that routed a $39 upgrade through its enterprise deal desk

Picture an enterprise SaaS business with a mature sales motion — five approval tiers, a three-person deal desk, average contract value in the low six figures, blended realized discount somewhere in the mid-twenties. It launches a self-serve tier to finally capture the SMB segment its reps have been disqualifying for years. Signups start arriving. Within a few weeks, self-serve customers begin asking the obvious question: can I get a better price if I pay annually, or if I buy ten seats instead of three?

The company already has governance. It has a deal desk, an approval matrix, a CPQ with routing rules. So the instinct is entirely reasonable: route the self-serve discount requests through the machinery that already exists. Nobody wants two systems. Finance likes one set of thresholds. The CRO likes one dashboard.

What happens next is mechanical and predictable. A customer wanting fifteen percent off a $39-per-month plan lands in the same approval queue as a $480,000 three-year renewal that closes on the last day of the quarter. The deal desk, which was staffed to handle perhaps forty to eighty enterprise quote reviews a week, is suddenly fielding hundreds of micro-approvals. Either it treats them with equal seriousness — in which case the enterprise turnaround time that used to be four hours stretches into days, and reps start escalating to the CRO to jump the line — or it triages, approving the small ones on sight without real review. That second outcome is worse than it looks: the company now has an undocumented two-tier system with none of the design benefits of a deliberate one and none of the visibility either. Nobody chose it, nobody wrote it down, and nobody can explain the rules to a new hire.

Meanwhile the self-serve motion is dying at the checkout page. A prospect who was going to enter a credit card in ninety seconds is now told a human will get back to them. The entire economic premise of the PLG motion — that acquisition and monetization happen without a human touching the transaction — has been quietly repealed by a governance decision that nobody framed as a strategy change.

Now run the mirror image. A PLG-native developer tools company, capped at a twenty percent promo ceiling with no human in the loop anywhere, starts landing inbound six-figure deals from customers who self-served, hit a ceiling, and called in. The founder eyeballs the first few in Slack. Trained by PLG economics to treat twenty percent as an enormous concession, the founder holds firm at fifteen on a deal where the buyer's procurement team would have signed at list plus a two-year term — and simultaneously gives a flat discount with no term commitment attached, because the PLG mental model has no concept of trading price for duration. The company leaves money on the table and gets nothing structural in return. Then the next rep, hired from an enterprise background, discovers there is no ceiling at all in the enterprise path because nobody built one, and closes at thirty-four percent because it was available.

Both companies are making the same category error from opposite directions. They are treating discount governance as one thing that must either be shared or not shared. It was never one thing.

Splitting governance into its four parts, and why two are shared and two are not

The separate-versus-integrated debate dissolves the moment you stop treating discount governance as a monolith. It is four distinct sub-systems, and they have different natural homes:

The policy. What discounts legitimately exist, what they cap at, what the margin floor is, what reasons justify a concession. This is a statement about the company's pricing power and margin structure. It cannot differ by motion without the company literally telling the market two different stories about what its product is worth.

The authority model. Who or what may grant a discount and up to what limit. In enterprise this is a human escalation ladder. In PLG it is a rule compiled into a promo engine. These are not variations on a theme — they are structurally different mechanisms, because one has a human in the transaction and the other does not, by design.

The instrumentation. How each grant gets recorded, routed, and made visible. Enterprise needs an SLA-governed queue with an approval trail. PLG needs a promo registry and a redemption log. Again, different mechanisms for different transaction shapes.

The feedback loop. How the company reads the aggregate and adjusts. This is inherently company-level — the board asks about blended margin, not motion-specific margin, and the whole point of governance is catching drift before it compounds.

Lay those four against the question and the answer is not a compromise. The policy and the feedback loop are shared by nature. The authority model and the instrumentation are separate by nature. Most people arguing about whether to integrate are actually only arguing about the authority model and the instrumentation — the two things that should obviously be separate — while implicitly assuming the policy and the feedback loop are already shared, when in most hybrid companies they are not.

The useful metaphor is a shared spine with two motion-specific limbs. The spine holds everything that determines whether the company is coherent and whether finance can trust the margin: the canonical price list, the margin floor logic, the discount policy object, the reconciliation layer, and the governance cadence. The limbs are the two enforcement paths, each tuned to its own economics.

The decision rule for what goes where is a single question. *Does this affect whether the company tells the market a consistent story, or whether finance can trust the margin number?* If yes, it belongs in the spine. If it is purely about how a discount gets approved and applied inside one motion, it belongs in that motion's limb. List price is spine. Margin floor is spine. Promo code expiration logic is a PLG limb. The VP-approval Slack workflow is an enterprise limb.

There is a second precision worth having: discount governance is a margin-protection system, not a deal-approval system. When a deal desk believes its job is to approve or reject deals, it becomes a bottleneck and an adversary, and reps route around it. When it understands its job is to protect margin while enabling the deal, it starts reshaping deals instead of blocking them — turning a flat thirty percent request into twenty-two percent plus a three-year term plus an expansion commitment. Design the architecture around margin protection with deal enablement as a co-equal constraint, not the reverse.

How the mechanism actually works: one policy object, two enforcement paths

The keystone artifact is the discount policy object — a structured, version-controlled, machine-readable definition of the rules that lives in exactly one place and is referenced by both motions. It is not a PDF. It is not a slide deck. It is not tribal knowledge in the deal desk's heads. It is a data object, ideally in the CPQ configuration layer or the billing platform's config, containing per product and per plan:

• List price and its effective date, plus currency variants
• Hard margin floor and soft margin floor, both expressed in gross-margin percentage terms rather than discount-percentage terms
• PLG discount cap — the absolute maximum any stack of self-serve discounts may reach
• Enterprise approval tier table — the discount-depth bands and the authority each maps to
• Approved discount reason picklist — a closed list, with any reason-specific caps
• Stacking rules — which discounts may combine, which are mutually exclusive, and the order of application
• Promo registry pointers — every active PLG promo, its owner, cap, and expiration

Both the promo engine and the CPQ read from this object. When finance updates the margin floor because cost-to-serve changed, it changes in one place and both motions inherit it. When a new discount reason is approved, it is added once.

The floor deserves particular attention because it is the shared constitution. It must be expressed in gross margin terms, not discount percentage, because thirty percent off a high-margin product and thirty percent off a low-margin one are entirely different decisions — a percentage cap lets the low-margin product go underwater while needlessly constraining the high-margin one. Compute it off fully-loaded cost to serve: infrastructure, support, payment processing, and a fair allocation of customer success. Cost-to-serve genuinely differs between motions — PLG carries heavy payment-processing drag across thousands of low-touch accounts, enterprise carries white-glove CS and sometimes dedicated infrastructure — and the floor logic must account for that even though the floor itself is shared.

In practice you want two levels. The soft floor protects target margin and can be crossed by a VP or deal desk with documented justification. The hard floor protects against selling at a loss and only the CFO crosses it. Both motions inherit both. The PLG limb rarely approaches the floor because its discounts are small and capped — but the floor still catches promo-stacking accidents where three individually legitimate promos combine into an effective price nobody intended. The enterprise limb approaches the floor regularly on competitive and strategic deals, which is exactly why it must be a system-enforced number rather than a guideline in a document.

The PLG limb honors the core promise: no human touches the transaction. Every discount is therefore pre-approved by construction — it exists as a rule, not as a request. There are four legitimate mechanisms and you should be suspicious of any fifth. Plan-level structural discounts (annual prepay versus monthly-equivalent, programmatic volume tiers). Time-boxed promotional codes, each with a hard expiration, a hard usage cap, and a hard ceiling. Lifecycle-triggered offers, such as an automated retention offer to a churn-risk account. Partner or marketplace discounts governed by pre-set contractual terms.

What the PLG limb must not have is a back channel. The moment a self-serve customer can email support and get a one-off "let me see what I can do," you have created an ungoverned, unreported, infinitely abusable enterprise-style negotiation inside your PLG motion, and it spreads by word of mouth faster than you can contain it. The correct answer to "can I get a discount?" in PLG is always a rule: yes, annual prepay saves you this much — or no. If the PLG motion needs deeper discounts to convert, that is a pricing or packaging problem, not a discounting problem, and the fix is the price, not a discount valve.

The enterprise limb is built around human judgment, tiered by authority and dollar consequence. The canonical structure is a five-tier ladder: rep self-approval for small routine concessions, front-line manager for the next band, deal desk for the band where discounting starts to genuinely matter, VP or CRO approaching the soft floor, CFO at or beyond the hard floor. The deal desk tier is where structured trade-offs enter — it does not merely approve or deny, it negotiates the shape of the deal, trading discount depth for term length, payment terms, expansion commitments, or reference rights.

The hard architectural commitment: the two motions share the policy object and the reporting layer, but they must not share an approval queue. The PLG limb's queue is essentially nonexistent — exceptions are policy-object decisions made once, not per-transaction requests. The enterprise limb's queue is a real, SLA-governed workflow. The routing logic lives in the CPQ for enterprise and the promo engine for PLG, and both read the same policy object to know where the tier boundaries sit. When finance moves the deal-desk threshold, it changes in the object and both routing systems inherit it. The outcomes then flow into the warehouse, where they finally meet. They meet in the data, not in the workflow. That is the whole trick.

Real numbers: thresholds, escalation rates, and what healthy looks like

Specific numbers depend on your gross margin structure, your competitive position, and your average deal size — but the shape of a healthy system is consistent, and the ranges below are the ones practitioners actually converge on. Treat them as starting calibration to be tuned against your own reconciliation data, not as universal truths.

Enterprise tier bands. A common ladder runs roughly: rep self-approval to around ten to fifteen percent, front-line manager to about twenty, deal desk in the twenty-to-thirty range, VP or CRO from thirty to roughly forty, CFO above that or anywhere the hard floor is threatened. The specific numbers matter far less than the principle that each tier of authority corresponds to a real increment of business consequence, and that the bands are wide enough that most deals never escalate past tier one or two — otherwise the desk drowns and the tiers become rubber stamps.

Escalation distribution. A healthy enterprise limb sees roughly sixty to seventy-five percent of deals close inside rep self-serve authority, fifteen to twenty-five percent touching a manager or deal desk, under ten percent reaching a VP, and CFO exceptions in the low single digits. If forty percent of your deals escalate, your tier-one threshold is set too low and you are manufacturing friction that reps will route around. If two percent escalate, your threshold is too high and you are not governing anything — the discounting is all happening below the line where anyone looks.

PLG cap. Set the absolute ceiling for any stack of self-serve discounts deliberately low — commonly in the fifteen-to-twenty-five percent range, with the annual-prepay structural incentive typically sitting around fifteen to twenty percent off monthly-equivalent. The reason for the low cap is arithmetic: a discount applied at plan level is instantly multiplied across every self-serve account touching that plan. In enterprise, a ten percent concession is one negotiated trade on one deal, fully reversible at renewal. In PLG, ten percent at plan level is a permanent-feeling repricing of an entire segment, because rolling it back reads to customers as a price increase.

Realized discount targets. PLG blended realized discount should generally sit well below ten percent — if it is drifting toward the cap, promos are stacking or zombies are alive. Enterprise blended realized discount in most B2B SaaS lands somewhere in the high teens to high twenties depending on segment and competitive intensity. The number itself is less diagnostic than its trend: a stable twenty-four percent is a functioning system; a twenty-one percent that has been climbing 150 basis points per quarter for four quarters is a system failing in slow motion.

Deal desk SLA. Under four business hours for standard requests, under twenty-four hours for non-standard. Miss those consistently and reps stop using the desk — they pad requests to avoid a second round trip, or they escalate socially to a VP who will say yes faster than the desk will say yes with conditions.

Exception rate maturity. By the second year of a functioning system, the share of deals needing above-rep approval should have settled under about fifteen percent. Early on it will be much higher, because thresholds are wrong and reps have not internalized what closes without a concession.

Beyond output metrics, measure the governance system's own health, because a system producing good numbers by accident today will produce bad ones tomorrow. Track escalation rate and its trend. Track approval cycle time against SLA and whether it is degrading. Track approval rate by tier — if every tier approves ninety-eight percent of what reaches it, the tiers are decorative; healthy tiers reject or reshape a meaningful minority. Track structured-reason completeness: anything materially below full capture means the reconciliation layer is partly blind. Track policy-object change velocity — zero changes per quarter means the policy is ossifying, constant changes mean it was never right. Track promo registry hygiene: active promo count, average promo age, and the count of expired-but-still-redeemable codes. Track cross-motion arbitrage incidents detected — zero forever might mean none exist, or might mean nobody is looking.

The reason taxonomy is what turns these numbers from measurements into diagnoses. "Blended discount is nineteen percent" is a number. "Nineteen percent, of which eleven points bought multi-year and volume commitments, five points are competitive displacement against two named competitors, and three points are uncategorized rep concessions" is a diagnosis — it tells you the three uncategorized points are the actual leak and the five competitive points are a product and pricing signal. Keep the picklist closed and short, eight to twelve reasons, so reps do not have to think hard and the categories stay meaningful. Free-text reason fields are where governance goes to die; they are uncategorizable and therefore unanalyzable.

Trade-offs: what you give up with each architecture

Neither pure option is stupid. Each buys something real and pays for it somewhere specific.

Full integration — one policy, one queue, one threshold set — buys genuine simplicity. One system to build, one to train on, one dashboard, one owner, no reconciliation work. For a company where one motion is under five percent of revenue, this is often correct: the overhead of a second limb exceeds the value. The price is the mechanical failure described above. A unified queue means enterprise SLAs collapse under SMB volume, or the desk silently triages. A unified threshold means a number that is a rounding error in a six-figure deal becomes the default expectation across tens of thousands of self-serve accounts. And there is a cultural cost in both directions: routing PLG discounts through humans teaches the monetization team that discounts are negotiable, reintroducing exactly the cost structure PLG existed to escape; forcing enterprise reps into rigid no-exception promo logic removes the negotiation surface a skilled AE uses to trade discount for term, expansion, or a faster close.

Full separation — independent promo engine, independent deal desk, no shared anything — buys maximum local velocity. Each motion moves at its own speed with no cross-motion coordination tax. It fails more slowly but just as completely, in four predictable cracks.

*Floor drift.* Independently set margin floors will diverge. Within a year you have a PLG plan whose promo-stacked effective price sits below the floor the enterprise team treats as sacred, or the reverse. Nobody notices until finance runs a cost-to-serve analysis and finds a whole segment underwater.

*Arbitrage.* Procurement departments are extremely sophisticated. They will discover that the self-serve price for a given seat count is lower than the negotiated enterprise price, or that buying self-serve and then calling sales for a true-up games both systems. Two governance systems that cannot see each other cannot detect this.

*Reporting incoherence.* The board asks whether blended realized discount is getting worse, and RevOps cannot answer, because the two systems define discount differently, calculate against different baselines, and never reconcile.

*Trust damage.* A customer who started self-serve and grew into an enterprise conversation gets quoted numbers that contradict what they were already paying. The relationship starts on a credibility deficit.

The hybrid is not a compromise between two bad options. It takes the coherence benefit from integration and the velocity benefit from separation while discarding both failure modes. What it costs is real build work: the reconciliation layer has to exist, both motions must define discount against the same baseline, and someone must own the policy object with actual change control. That is a quarter of unglamorous RevOps work at minimum, and at an acquisition-merged company it can be considerably more.

Tooling trade-offs sit inside each limb. The enterprise limb runs on the CRM plus a CPQ. Salesforce CPQ is the incumbent and integrates natively but is heavy and expensive to configure; DealHub, Subskribe, and similar newer platforms have taken share specifically because they configure faster and handle usage-based and hybrid pricing better — which matters disproportionately for a hybrid company, because enterprise deals increasingly include usage components that are literally the PLG product. The CPQ's four governance jobs: block any quote below the hard floor, route above-threshold discounts to the right tier, capture a structured reason as controlled data, and write realized discount plus the full approval trail back to the CRM and onward to the warehouse. One configuration detail matters more than the rest: approval must precede quote finalization. A quote sent and then approved retroactively is not governance, it is theater. And the approver must see margin impact at the moment of decision — "rep requests twenty-eight percent" is a guess; "twenty-eight percent takes this deal from seventy-one to fifty-four percent gross margin, competitive displacement against a named competitor" is a decision.

The PLG limb runs on a billing platform — Stripe Billing, Chargebee, Recurly, or a usage-metering layer like Metronome or Orb — plus a promotions engine. The governance requirement is that the promo engine be structurally incapable of issuing a discount that violates the PLG cap or the hard floor, because there is no human to catch it. If your billing platform cannot enforce a hard cap and floor natively, that is a genuine gap, and the standard fix is a thin governance service sitting between the promo engine and checkout that validates every discount against the policy object before it can apply.

Pitfalls: the six ways hybrid discount governance actually breaks

Hybrid companies fail in a small number of recognizable patterns. Naming them lets a team self-diagnose in an afternoon.

The leaky tier-one threshold. The rep self-serve limit was set when average deal size was a third of what it is now, and nobody revisited it. The majority of discounting happens with zero oversight and the desk only sees rare extremes. The symptom is deceptive: a low escalation rate that reads as healthy paired with a blended discount that keeps creeping. *Fix:* treat the tier-one threshold as a number that must be re-derived every time average deal size shifts materially, and check it as a standing agenda item, not on intuition.

The rubber-stamp deal desk. The desk reports to the CRO, or its scorecard is purely deal velocity, so it approves essentially everything. Symptoms: high throughput, near-total approval rate, almost no deals structurally reshaped. *Fix:* this is an org-design problem, not a process problem. The deal desk must report into RevOps or Finance, not into Sales. The reasoning is structural rather than about trusting individuals: a desk whose comp and career incentives are aligned with closing deals will, in aggregate and over time, approve discounts, because every individual approval helps the number it is measured on. A desk measured on margin integrity, SLA adherence, and deal-structure quality becomes a real governor. This does not make it adversarial — the best desks are deeply partnered with reps and make them better negotiators — but the partnership works *because* the incentives are independent.

PLG promo sprawl. Nobody owns the promo registry, promos never expire, and the effective self-serve discount runs far above the nominal cap through stacking and zombies — the launch code from two years ago that is somehow still redeemable, the "temporary" partner discount that became permanent through neglect. The symptom is realized PLG discount drifting above target with no single decision that caused it. *Fix:* a named promo owner and a quarterly sweep that kills anything expired, unowned, or purposeless. Zombie promos are the PLG equivalent of enterprise discount creep and erode margin just as silently.

The reconciliation gap. The two motions never pipe into a common schema, so blended discount is genuinely unanswerable and the quarterly review has no data to act on. The symptom is leadership debating discounting with anecdotes and quarter-end war stories. *Fix:* build the warehouse layer even if it is rough. Visibility before control — you cannot govern what you cannot see, and the reconciliation layer is cheap relative to its diagnostic value. Critically, force both motions to calculate discount against the same baseline, list price from the canonical object. PLG often calculates against monthly-equivalent of an annual plan while enterprise calculates against undiscounted list, and the two never reconcile until someone forces a common denominator.

Incentive inversion. Comp pays identically for a discounted deal and a full-price one, so the entire field is economically motivated to discount. Symptoms: discounting concentrated at quarter-end, reps volunteering concessions before the customer asks. *Fix:* discount-aware accelerators and decelerators — full-price and lightly-discounted deals pay a marginally higher rate or count more toward accelerators, heavily discounted deals count slightly less. Do not comp reps on margin directly; it is too noisy and too far from their control. On the PLG side the parallel failure is measuring the growth team purely on conversion rate, which guarantees they lobby to loosen the cap, because discounting always lifts conversion short-term. Measure them on net revenue and blended margin so a conversion lift bought with a margin giveaway shows up as a wash in their own numbers, and they self-police.

The immutable-policy trap. The policy object exists but has no change-control cadence, so it ossifies, drifts from market reality, and reps route around it. The symptom is a documented policy everyone privately ignores. *Fix:* a real quarterly governance review with the authority to actually change the object — RevOps as owner, Finance, the CRO or VP Sales, and the head of growth. Standing agenda: realized discount by motion versus target, trailing four-quarter trend, top discount reasons and whether they are still valid, exception-rate analysis, zombie-promo sweep results, detected cross-motion arbitrage, and proposed policy changes. The output is a small set of change-control tickets. This cadence is also where you catch discount creep — the slow, individually reasonable expansion where each deal's exception is justified but the aggregate is a steady bleed. It is invisible at the deal level and only legible in the trailing trend.

Two pitfalls sit outside that list and deserve their own treatment.

The hand-off zone. The messiest real case is the customer who does not stay in one motion. A self-serve account grows, hits a ceiling, and gets assigned a rep. Or an enterprise account spins up self-serve seats for a new team. These transitions leak the most margin and trust, because the customer's price history now spans two systems that, if the spine is wrong, disagree about what they were paying. The single worst version: a rep quotes standard enterprise list-minus-twenty-five to a customer already paying list-minus-thirty on self-serve through annual prepay. The customer experiences the "enterprise upgrade" as a price increase and the relationship starts underwater. The rule: the customer's existing realized price is the floor for the next conversation, regardless of which motion they came from — which requires the CPQ to read the reconciliation layer, which is exactly why that layer sits in the spine. The reverse case needs its own spine rule: self-serve seats added to an enterprise account inherit the negotiated contract rate for that product, not the public price, unless the contract explicitly carves out self-serve expansion. Get this wrong and procurement buys the minimum enterprise commitment then loads up on cheaper self-serve seats, gaming blended price below anything anyone approved.

Discounts that do not look like discounts. A twelve-month deal with the first three months free is a twenty-five percent discount that a naive system records as zero. Credits, rebates, ramp deals, and free-period structures are economically discounts but never appear as a discount percentage. The policy object and the CPQ must normalize all of them into one realized-discount and realized-margin number, or the reconciliation layer systematically understates real discounting. Two related contract-layer leaks: renewal uplift language — a discount without an explicit renewal step-back silently becomes permanent for the life of the account, one of the largest hidden margin leaks in enterprise SaaS, so the policy object should define default uplift behavior by reason (a multi-year commitment discount holds through the committed term; a competitive-displacement discount steps back toward list at renewal) — and most-favored-nation or price-protection clauses, which must escalate to the highest tier regardless of headline discount magnitude, because their blast radius spans an entire customer's future purchasing and can constrain your ability to raise prices at all.

Finally, a staging pitfall: do not build the enterprise limb's bureaucracy before enterprise revenue justifies it, and do not skip building it because PLG culture treats process as the enemy — until an enterprise motion that is suddenly forty percent of revenue is being governed by a promo-code mindset. Early stage, install exactly one discipline: capture a structured discount reason on every concession, because that habit is what makes every later layer possible. Growth stage, build the canonical price list and the policy object *before* the second motion's discounting habits calcify. Scale stage, bring the full architecture online. Under board time pressure, the sequence is always visibility first, then control.

Related questions

Should the PLG discount cap ever match the enterprise cap?

No. The blast radius differs by orders of magnitude — a plan-level PLG discount multiplies across every self-serve account instantly and feels irreversible, while an enterprise concession is one negotiated trade reversible at renewal. Set the PLG cap deliberately lower, typically well under half the enterprise ceiling.

Who should own the discount policy object?

RevOps owns it, with Finance as co-owner of the margin floor specifically. Neither sales nor growth should own it, because both have structural incentives to loosen it. The owner runs change control; the quarterly review is where changes get approved and logged.

What if our billing platform cannot enforce a hard discount cap?

Build a thin governance service between the promo engine and checkout that validates every discount against the policy object before it applies. In PLG there is no human backstop, so an unenforceable cap is not a cap — it is a suggestion that will be exceeded through stacking.

How do we detect cross-motion price arbitrage?

Query the reconciliation layer for configurations where the self-serve effective price for a given seat count undercuts a negotiated enterprise rate, or the reverse. Both require a common discount baseline. Make it a standing quarterly review item rather than an ad-hoc investigation after procurement finds it first.

Does a small company need all of this?

No. Pre-Series A with one dominant motion, a spreadsheet and a Slack channel are fine. Install only the structured-reason capture habit — that single discipline is what makes the policy object, the reconciliation layer, and every later governance layer buildable rather than a retrofit.

FAQ

Should enterprise and PLG discounts ever share an approval queue?

No. This is the one hard architectural commitment in the whole design. The two motions share the policy object and the reporting layer, but a shared queue means either enterprise SLAs collapse under SMB volume or the desk silently triages and you get an undocumented two-tier system with worse visibility than a deliberate one. The queues stay separate; the outcomes meet in the data warehouse.

Why should the margin floor be expressed in gross margin rather than discount percentage?

Because a fixed discount percentage means completely different things across products with different cost structures. Thirty percent off a high-margin product may still be comfortably profitable while thirty percent off a low-margin one is underwater. A percentage cap simultaneously lets the wrong product go below cost and needlessly constrains the right one. Compute the floor off fully-loaded cost to serve, including payment processing and a fair CS allocation.

Why can't the deal desk report to the CRO?

Because a deal desk whose comp, performance review, and career path are aligned with closing deals will approve discounts in aggregate over time — every individual approval helps the number it is measured on. That is a structural incentive, not a character question. Reporting into RevOps or Finance, with a scorecard covering margin protected and SLA adherence rather than deals cleared, is what turns a rubber stamp into a real governor.

What is the single first thing to build if we have nothing?

The reconciliation layer, even a rough version. Visibility precedes control — you cannot tune thresholds, catch discount creep, or detect arbitrage without a common view where both motions' discount outcomes reconcile against the same baseline. It is cheap relative to its diagnostic value, and it usually reveals the worst single leak within days.

How do we stop a one-time competitive discount becoming permanent?

Define default renewal-uplift behavior in the policy object, keyed to the discount reason. A multi-year commitment discount holds flat through the committed term because the customer paid for it with duration. A competitive-displacement discount should default to stepping back toward list at renewal, with the step-back written into the order form language rather than left to the renewal rep to negotiate from scratch.

How often should thresholds change?

Review them quarterly and change them when the data says to — typically a few times a year, not every quarter and not never. Zero changes over several quarters means the policy is ossifying and drifting from market reality. Constant changes mean the original calibration was wrong and reps cannot build reliable expectations. Any material shift in average deal size should trigger an immediate tier-one threshold re-derivation.

Sources

• Bessemer Venture Partners — State of the Cloud
• OpenView Partners — Product Led Growth resources
• a16z — Enterprise and SaaS content
• Salesforce CPQ product documentation
• Stripe Billing documentation — coupons and discounts
• Harvard Business Review — pricing and discounting research
• McKinsey — Growth, Marketing and Sales insights
• SaaS Capital — pricing and metrics research
• Chargebee — subscription billing documentation

Related on PULSE

• How to structure a deal desk that reps actually want to use
• Setting margin floors from fully-loaded cost to serve
• Comp plan design that supports rather than fights discount governance
• Building a CPQ approval matrix that scales with deal size
• Detecting and closing cross-motion price arbitrage
• When to add a sales-led motion on top of PLG