commercial-policy

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

commercial-policy

Purpose

用途

Design the rules of engagement that govern discounting off list price — the artifact that Deal Desk and AEs operate under. Three deterministic tools:

```
discount_matrix_builder.py
```
— builds a 4-dimensional matrix (ARR band × term length × payment terms × strategic value tier), each cell carrying an approved discount band backed by current win-rate + NRR data, plus an approver tier (AE / Manager / Director / VP / CFO).
```
exception_router.py
```
— when an asks-for-discount lands outside the matrix, routes it through the named approver chain, attaches required compensating commitments (multi-year prepay + named expansion path + reference commitment + MSA tightening), produces machine-readable audit-trail metadata, and flags precedent risk if 3+ similar exceptions have landed in the trailing quarter.
```
policy_linter.py
```
— lints the matrix for governance defects: approver inversion, band inversion, margin-floor violation, coverage gaps, cliff edges, undefined strategic tiers, inconsistent margin floors, thin data backing.

The output is the policy itself (matrix + exception flow + lint report), not a per-deal application of it.

设计管控标价折扣的参与规则——即Deal Desk和AEs执行工作所依据的规则文档。包含三个确定性工具：

```
discount_matrix_builder.py
```
——构建一个四维矩阵（ARR区间 × 合同期限 × 付款条款 × 战略价值层级），每个单元格包含基于当前赢单率+NRR数据的获批折扣区间，以及对应的审批层级（AE / 经理 / 总监 / 副总裁 / CFO）。
```
exception_router.py
```
——当申请的折扣超出矩阵范围时，将其转入指定的审批链，附加必需的补偿承诺（多年预付+明确的拓展路径+推荐承诺+MSA条款收紧），生成机器可读的审计跟踪元数据，若过去一季度出现3次及以上类似异常申请，会标记先例风险。
```
policy_linter.py
```
——检查矩阵中的管控缺陷：审批层级倒置、区间倒置、违反最低利润率、覆盖空白、断崖式规则、未定义战略层级、最低利润率不一致、数据支撑不足。

输出结果是政策本身（矩阵+异常流程+检查报告），而非针对单笔交易的政策应用。

When to use

使用场景

A new Head of Commercial or Head of Deal Desk is writing the company's first formal commercial policy
The existing matrix is older than 6 months and discount drift is showing in margin reviews
Reps are citing "Maria approved 28% on Acme last quarter" as precedent and you need to break the precedent loop
Q-over-Q exception count is rising and you suspect the matrix bands are mispriced
CFO has tightened the margin floor and the matrix needs to be rebuilt against the new constraint
A board / exec is asking "why do we discount this much?" and you need a data-backed defensible policy

Do NOT use this skill to:

Approve a specific deal — that's
```
commercial/skills/deal-desk
```
Set the pricing model + list price — that's
```
commercial/skills/pricing-strategist
```
Author a proposal / SOW / MSA prose — that's
```
business-growth/contract-and-proposal-writer
```
Make the strategic "when do we hire a VP Sales" call — that's
```
c-level-advisor/cro-advisor
```

新任商业负责人或Deal Desk主管正在制定公司首份正式商业政策
现有矩阵已使用超过6个月，且利润率审查中发现折扣偏差
销售代表以"Maria上季度批准了Acme公司28%的折扣"作为先例，你需要打破这种先例循环
季度间异常申请数量上升，你怀疑矩阵区间定价不合理
CFO收紧了最低利润率要求，需要基于新约束重建矩阵
董事会/高管询问"我们为何给出如此高的折扣？"，你需要一份有数据支撑、可辩护的政策

请勿将此技能用于：

审批特定交易——该场景属于
```
commercial/skills/deal-desk
```
设置定价模型+标价——该场景属于
```
commercial/skills/pricing-strategist
```
撰写提案/SOW/MSA文案——该场景属于
```
business-growth/contract-and-proposal-writer
```
做出"何时招聘销售副总裁"这类战略决策——该场景属于
```
c-level-advisor/cro-advisor
```

Workflow

工作流程

Audit current discount distribution. Pull the last 4 quarters of closed-won + closed-lost deals from CRM. Fill
```
assets/policy_design_template.md
```
(~20 minutes). Capture:
```
arr
```
,
```
discount_pct
```
,
```
term_months
```
,
```
payment_terms_days
```
,
```
strategic_value
```
,
```
win_lost
```
,
```
nrr_12mo
```
per deal.
Design the data-backed matrix. Run
```
scripts/discount_matrix_builder.py --input policy_intake.json --profile {saas|enterprise-software|api|marketplace|services}
```
. Output is a 4-dimensional matrix with approved discount band + approver tier + margin floor + observed win-rate + observed NRR per cell. Cells with
```
n < 5
```
observed deals are flagged
```
THIN
```
.
Design the exception flow. Run
```
scripts/exception_router.py --sample
```
to see the structure. For each severity band of exception (0-5 pts over, 5-10, 10-20, 20+), the router enforces required compensating commitments. Codify the flow in your policy doc; the router becomes the operational implementation.
Lint the matrix. Run
```
scripts/policy_linter.py --input matrix.json
```
. Get a ranked findings report — BLOCKER / MAJOR / MINOR — across 10 lint rules. Resolve every BLOCKER before publishing the matrix to AEs.
Publish + quarterly review. Publish the matrix as a versioned artifact. Re-run the builder and the linter every quarter against the new 4-quarter rolling deal corpus. Cells where observed NRR <
```
target_nrr
```
are flagged for review.

审计当前折扣分布。从CRM中提取过去4个季度的已赢单+已丢单数据。填写
```
assets/policy_design_template.md
```
（约20分钟）。记录每笔交易的：
```
arr
```
、
```
discount_pct
```
、
```
term_months
```
、
```
payment_terms_days
```
、
```
strategic_value
```
、
```
win_lost
```
、
```
nrr_12mo
```
。
设计基于数据的矩阵。运行
```
scripts/discount_matrix_builder.py --input policy_intake.json --profile {saas|enterprise-software|api|marketplace|services}
```
。输出为四维矩阵，包含每个单元格的获批折扣区间+审批层级+最低利润率+观测赢单率+观测NRR。观测交易数
```
n < 5
```
的单元格会标记为
```
THIN
```
。
设计异常流程。运行
```
scripts/exception_router.py --sample
```
查看结构。对于每个严重程度的异常申请（超出0-5个百分点、5-10个百分点、10-20个百分点、20个百分点以上），路由工具会强制执行必需的补偿承诺。在政策文档中固化该流程，路由工具将作为实际执行的实现方案。
检查矩阵。运行
```
scripts/policy_linter.py --input matrix.json
```
。获取分级发现报告——阻塞项（BLOCKER）/主要问题（MAJOR）/次要问题（MINOR），涵盖10条检查规则。在向AEs发布矩阵前，需解决所有阻塞项。
发布+季度审查。将矩阵作为版本化文档发布。每季度基于过去4个季度的滚动交易数据，重新运行构建工具和检查工具。观测NRR <
```
target_nrr
```
的单元格会标记为需审查。

Scripts

脚本

Script	Purpose	Industry profiles
`scripts/discount_matrix_builder.py`	4-dim data-backed matrix with approver tiers + margin floors	saas, enterprise-software, api, marketplace, services
`scripts/exception_router.py`	Routes exception requests with compensating commitments + audit trail	n/a (matrix-driven)
`scripts/policy_linter.py`	10-rule lint pass over the matrix	n/a (deterministic across profiles)

All three: stdlib-only,

--help

--sample

--input <json>

--output {markdown,json}

脚本	用途	行业配置文件
`scripts/discount_matrix_builder.py`	生成带审批层级+最低利润率的四维数据驱动矩阵	saas, enterprise-software, api, marketplace, services
`scripts/exception_router.py`	处理异常申请路由，附加补偿承诺+审计跟踪	n/a（基于矩阵驱动）
`scripts/policy_linter.py`	对矩阵执行10条规则的检查	n/a（跨配置文件通用）

所有脚本：仅依赖标准库，支持

--help

、

--sample

、

--input <json>

、

--output {markdown,json}

参数。

References

参考资料

```
references/discount_governance_canon.md
```
— Discount governance evidence base: OpenView Partners benchmarks, David Skok (For Entrepreneurs) discount math, Tomasz Tunguz on discount distribution, Bessemer State of the Cloud, KeyBanc Capital Markets SaaS Survey, Bridge Group AE-compensation research, RevOps Co-op playbooks, Forrester deal-desk research. 8 sources.
```
references/policy_design_canon.md
```
— Policy-as-artifact design: SaaStr (Jason Lemkin), Winning by Design (Jacco van der Kooij) on commercial discipline, Forrester deal-desk maturity research, MIT Sloan on incentive-system gaming, McKinsey on commercial-policy effectiveness, Bain Pricing Power, Salesforce CPQ implementation guides. 7 sources.
```
references/policy_anti_patterns.md
```
— 8 named anti-patterns with sourced studies + countermeasures + lint-rule mapping: precedent-sets-policy, no-data-backing, no-compensating-commitments, approver/margin misalignment, no audit trail, cliff edges, undefined "strategic value", no quarterly review. 8 sources.

```
references/discount_governance_canon.md
```
——折扣管控证据库：OpenView Partners基准数据、David Skok（For Entrepreneurs）的折扣计算方法、Tomasz Tunguz关于折扣分布的研究、Bessemer云状态报告、KeyBanc Capital Markets SaaS调研、Bridge Group AE薪酬研究、RevOps Co-op操作手册、Forrester交易台研究。共8个来源。
```
references/policy_design_canon.md
```
——政策文档设计：SaaStr（Jason Lemkin）、Winning by Design（Jacco van der Kooij）关于商业纪律的内容、Forrester交易台成熟度研究、MIT Sloan关于激励系统漏洞的研究、McKinsey关于商业政策有效性的研究、Bain《定价力》、Salesforce CPQ实施指南。共7个来源。
```
references/policy_anti_patterns.md
```
——8种已命名的反模式，包含来源研究+应对措施+检查规则映射：先例替代政策、无数据支撑、无补偿承诺、审批/利润率不一致、无审计跟踪、断崖式规则、"战略价值"未定义、无季度审查。共8个来源。

Assumptions

假设前提

The skill assumes the pricing model and list price already exist (set via
```
commercial/skills/pricing-strategist
```
). Commercial-policy governs discounts off list — it does not set list.
The CFO owns the
```
min_margin_pct
```
constraint (margin floor). The CRO / Head of Deal Desk owns the
```
max_discount_pct_without_exception
```
constraint (band cap). The skill keeps these inputs separate by design (per Bain Pricing Power — mixing accountability is the most common cause of policy drift).
Industry profiles bake in customary band widths. Companies with idiosyncratic economics should pass overrides via the input JSON.
The matrix is data-backed but not data-driven: the band is set by the constraints + profile; observed data is annotation that tells you whether the cell is performing. If observed NRR < target, that's a signal to review the band, not to keep discounting deeper.
"Strategic value" tiers (
```
logo
```
,
```
expansion
```
,
```
lighthouse
```
) are useful only if defined with concrete tests. The lint rule L06 enforces this.
This is a policy-design skill, not a deal-approval skill. It never says "approve" — it produces the matrix + exception flow that deal-desk then applies.

本技能假设定价模型和标价已确定（由
```
commercial/skills/pricing-strategist
```
设置）。商业政策管控标价折扣——不负责设定标价。
CFO拥有
```
min_margin_pct
```
约束（最低利润率）。CRO/Deal Desk主管拥有
```
max_discount_pct_without_exception
```
约束（区间上限）。本技能设计上将这些输入分开（依据Bain《定价力》——权责混淆是政策偏差最常见的原因）。
行业配置文件内置了常规的区间宽度。具有特殊经济模式的公司可通过输入JSON传递覆盖参数。
矩阵基于数据但不由数据驱动：区间由约束+配置文件设定；观测数据用于标注单元格是否表现达标。若观测NRR < 目标值，这是审查区间的信号，而非进一步加大折扣的理由。
"战略价值"层级（
```
logo
```
、
```
expansion
```
、
```
lighthouse
```
）只有在定义了具体验证标准时才有用。检查规则L06强制执行此要求。
这是政策设计技能，而非交易审批技能。它从不直接"批准"交易——而是生成矩阵+异常流程，供deal-desk技能后续应用。

Anti-patterns

反模式

Setting discount bands without data backing. "VP Sales argued for it in a Slack thread" is not data backing. If you can't show win-rate and NRR for the band, the band is rhetoric. (Caught by
```
data_backing
```
per cell + lint L08.)
Letting precedent set policy. "Maria approved 28% on Acme last quarter" is not a band — it's an exception that didn't break the policy.
```
exception_router.py
```
flags 3+ similar exceptions as a signal that the matrix is wrong, not the deal. (Anti-pattern AP-1.)
Approving exceptions without compensating commitments. Discount-for-nothing is a leak (Winning by Design). Every exception severity band requires non-negotiable commitments. (
```
exception_router.COMPENSATING_LIBRARY
```
.)
Cliff edges at round-number ARR thresholds. A hard $100K threshold produces deal-size gaming within 2 quarters (MIT Sloan agency theory). Smooth the gradient. (Lint L05.)
"Strategic value" as an undefined catch-all. If "strategic" is undefined, within a quarter 60% of deals will be flagged strategic and the matrix is dead. Define with concrete tests. (Lint L06.)
No quarterly review. Markets shift; matrices unchanged for 12 months are mispriced. Re-run the builder and linter every quarter. (Anti-pattern AP-8.)
Mixing CFO and CRO accountabilities. CFO owns the margin floor; CRO owns the band cap. Same accountable owner = predictable drift toward whatever they're compensated on (Bain Pricing Power).
Skipping the lint pass before publishing. BLOCKER findings (approver inversion, margin-floor violation, inverted bands) make the policy unsignable. Lint is the gate, not the after-action review.

无数据支撑设定折扣区间。"销售副总裁在Slack线程中主张该区间"不属于数据支撑。若无法展示该区间的赢单率和NRR，该区间只是空谈。（由每个单元格的
```
data_backing
```
检查+规则L08捕获。）
让先例替代政策。"Maria上季度批准了Acme公司28%的折扣"不是区间——而是未违反政策的异常申请。
```
exception_router.py
```
会将3次及以上类似异常申请标记为矩阵存在问题的信号，而非交易存在问题。（反模式AP-1。）
无补偿承诺批准异常申请。无回报的折扣是利润流失（Winning by Design）。每个严重程度的异常申请都需要不可协商的承诺。（
```
exception_router.COMPENSATING_LIBRARY
```
。）
在整数ARR阈值处设置断崖式规则。硬性的10万美元阈值会在2个季度内引发交易规模操纵（MIT Sloan代理理论）。应平滑梯度。（检查规则L05。）
"战略价值"作为未定义的万能项。若"战略"未定义，一个季度内60%的交易都会被标记为战略交易，矩阵将失效。需定义具体验证标准。（检查规则L06。）
无季度审查。市场在变化；12个月未更新的矩阵定价不合理。每季度重新运行构建工具和检查工具。（反模式AP-8。）
混淆CFO和CRO的权责。CFO负责最低利润率；CRO负责区间上限。同一负责人会导致政策向其考核指标倾斜（Bain《定价力》）。
发布前跳过检查步骤。阻塞项（审批层级倒置、违反最低利润率、区间倒置）会导致政策无法生效。检查是发布的前置关卡，而非事后回顾。

Distinct from

与其他技能的区别

Sibling	Scope	Difference
`commercial/skills/deal-desk`	Applies the policy to one deal at a time	Commercial-policy designs the policy itself. Deal-desk consumes the matrix; commercial-policy produces it.
`commercial/skills/pricing-strategist`	Sets pricing model (per-seat / usage / value / tiered) + list price	Commercial-policy governs discounts off list. Pricing-strategist sets the menu; commercial-policy governs the menu's discount discipline.
`c-level-advisor/cro-advisor`	Strategic CRO judgment ("when do we hire VP Sales?", "is our motion product-led or sales-led?")	Strategic, not operational. Commercial-policy is the artifact CRO commissions; it isn't CRO judgment itself.
`c-level-advisor/cfo-advisor`	Margin floor + unit-economics judgment	The CFO supplies `min_margin_pct` to commercial-policy as an input. Commercial-policy operationalizes the CFO's constraint as per-cell margin floors.
`business-growth/contract-and-proposal-writer`	Authors proposal/SOW/MSA prose	Commercial-policy emits structured matrix + audit-trail JSON, not customer-facing prose.

关联技能	范围	差异
`commercial/skills/deal-desk`	将政策应用于单笔交易	本商业政策技能设计政策本身。Deal-desk技能使用矩阵；商业政策技能生成矩阵。
`commercial/skills/pricing-strategist`	设置定价模型（按席位/使用量/价值/分层）+ 标价	商业政策技能管控标价折扣。Pricing-strategist技能制定定价菜单；商业政策技能管控菜单的折扣纪律。
`c-level-advisor/cro-advisor`	战略CRO决策（"何时招聘销售副总裁？"、"我们的模式是产品驱动还是销售驱动？"）	战略层面，非运营层面。商业政策是CRO委托生成的文档；而非CRO决策本身。
`c-level-advisor/cfo-advisor`	最低利润率+单位经济决策	CFO向商业政策技能提供 `min_margin_pct` 作为输入。商业政策技能将CFO的约束落地为每个单元格的最低利润率。
`business-growth/contract-and-proposal-writer`	撰写提案/SOW/MSA文案	商业政策技能输出结构化矩阵+审计跟踪JSON，而非面向客户的文案。

Forcing-question library (Matt Pocock grill discipline)

强制问题库（Matt Pocock grill discipline）

Walked one at a time by

/cs:grill-commercial

or the Commercial orchestrator before the skill runs. Recommended answer + canon citation per question. Never bundled.

"What's your observed discount distribution across the last 4 quarters — and is the median inside or outside your current matrix?" Recommended: pull the corpus before designing any band. If the observed median is outside the matrix, the matrix is rhetoric. Canon: OpenView SaaS Benchmarks; RevOps Co-op playbooks. Anti-pattern AP-2.
"What's the win-rate AND the 12-month NRR for deals at your current 'max discount' band?" Recommended: both, not one. A band with high win-rate but low NRR is buying logos with leaky-bucket retention. Tunguz benchmarks: top-NRR-quartile companies discount 6 pts less than bottom quartile. Canon: Tomasz Tunguz; Bessemer State of the Cloud.
"Who at the company owns the margin floor, AND who owns the discount-band cap — are those the same person?" Recommended: CFO owns floor; CRO/Head of Deal Desk owns cap. Same owner = drift toward what they're compensated on. Canon: Bain Pricing Power — separation of accountability is the structural fix. Anti-pattern AP-4.
"How is 'strategic value' defined in your current policy — with concrete tests, or with adjectives?" Recommended: concrete tests. "Top-20 named account in 2026 target list" is a test; "important customer" is not. Canon: SaaStr (Lemkin); Forrester deal-desk research. Lint rule L06. Anti-pattern AP-7.
"For exceptions above your matrix max, what compensating commitments are required — and are they in writing before the approver signs?" Recommended: minimum multi-year prepay + named expansion path; deeper exceptions require reference commitment + MSA tightening + executive sponsor. Canon: Winning by Design (van der Kooij); McKinsey B2B pricing studies. Anti-pattern AP-3.
"Has the same kind of exception been approved 3+ times in the trailing quarter — and if so, is the matrix wrong?" Recommended: 3+ similar exceptions means the band is mispriced. Rebuild the matrix; don't keep approving exceptions. Canon: OpenView discount drift studies;
```
exception_router._precedent_risk
```
. Anti-pattern AP-1.
"When was the last time you re-ran the matrix against the previous 4 quarters of data?" Recommended: quarterly. Annual review is too slow; the disciplined cohort revises quarterly. Canon: OpenView benchmarks; RevOps Co-op. Anti-pattern AP-8.
"For every exception in the last quarter, is there a machine-readable audit-trail record — or is the approval in Slack and email?" Recommended: structured record in CPQ or equivalent. Slack/email approvals don't survive year-2 renewal negotiations. Canon: Salesforce CPQ best practices; Forrester deal-desk maturity research. Anti-pattern AP-5.

Walk depth-first. Lock 1-4 before opening 5-8. After all 8 are answered, invoke

discount_matrix_builder.py

→

policy_linter.py

→

exception_router.py --sample

in sequence to produce the policy artifact.

在技能运行前，由

/cs:grill-commercial

或商业编排器逐个提问。每个问题配有推荐答案+参考资料引用。请勿批量提问。

"过去4个季度的观测折扣分布如何——中位数是否在当前矩阵范围内？" 推荐答案：在设计任何区间前先提取交易数据。若观测中位数超出矩阵范围，说明矩阵只是空谈。参考资料：OpenView SaaS基准数据；RevOps Co-op操作手册。反模式AP-2。
"当前‘最高折扣’区间的交易赢单率和12个月NRR分别是多少？" 推荐答案：两者都要提供，不能只给其一。赢单率高但NRR低的区间相当于用高流失率换取客户。Tunguz基准数据：NRR前四分之一的公司比后四分之一的公司折扣低6个百分点。参考资料：Tomasz Tunguz；Bessemer云状态报告。
"公司中谁负责最低利润率，谁负责折扣区间上限——是同一人吗？" 推荐答案：CFO负责最低利润率；CRO/Deal Desk主管负责区间上限。同一负责人会导致政策向其考核指标倾斜。参考资料：Bain《定价力》——权责分离是结构性解决方案。反模式AP-4。
"当前政策中‘战略价值’如何定义——是具体的验证标准，还是形容词？" 推荐答案：具体验证标准。"2026年目标客户列表中的前20名指定客户"是验证标准；"重要客户"不是。参考资料：SaaStr（Lemkin）；Forrester交易台研究。检查规则L06。反模式AP-7。
"对于超出矩阵上限的异常申请，需要哪些补偿承诺——且这些承诺是否在审批人签字前以书面形式确认？" 推荐答案：至少包含多年预付+明确的拓展路径；更严重的异常申请需要推荐承诺+MSA条款收紧+高管 Sponsor。参考资料：Winning by Design（van der Kooij）；McKinsey B2B定价研究。反模式AP-3。
"过去一季度是否有3次及以上同类异常申请获批——如果有，是不是矩阵存在问题？" 推荐答案：3次及以上同类异常说明区间定价不合理。应重建矩阵，而非继续批准异常申请。参考资料：OpenView折扣偏差研究；
```
exception_router._precedent_risk
```
。反模式AP-1。
"上次基于过去4个季度数据重新生成矩阵是什么时候？" 推荐答案：每季度一次。年度审查太慢；规范的团队每季度修订一次。参考资料：OpenView基准数据；RevOps Co-op。反模式AP-8。
"过去一季度的每笔异常申请，是否都有机器可读的审计跟踪记录——还是审批仅在Slack和邮件中进行？" 推荐答案：在CPQ或类似系统中生成结构化记录。Slack/邮件审批无法支撑第二年的续约谈判。参考资料：Salesforce CPQ最佳实践；Forrester交易台成熟度研究。反模式AP-5。

按深度优先顺序提问。在回答完1-4题后再提问5-8题。所有8题回答完毕后，依次调用

discount_matrix_builder.py

→

policy_linter.py

→

exception_router.py --sample

生成政策文档。

Quick examples

快速示例

bash

undefined

bash

undefined

Design the matrix

设计矩阵

python3 scripts/discount_matrix_builder.py --sample python3 scripts/discount_matrix_builder.py --input policy_intake.json --profile saas --output json > matrix.json

Lint the matrix

检查矩阵

python3 scripts/policy_linter.py --sample python3 scripts/policy_linter.py --input matrix.json

Walk the exception flow

查看异常流程

python3 scripts/exception_router.py --sample python3 scripts/exception_router.py --input request.json --output json


The sample matrix lints to **FAIL** with 4 BLOCKERs + 6 MAJORs + 2 MINORs — by design, to exercise every rule path. A real policy intake should lint to PASS or PASS_WITH_WARNINGS. The sample exception (42% on a $320K logo deal) routes to AE → Sales Manager → Director → VP Sales with 3 required compensating commitments (multi-year 36mo, prepay, named expansion path).

python3 scripts/exception_router.py --sample python3 scripts/exception_router.py --input request.json --output json


示例矩阵检查结果为**FAIL**，包含4个阻塞项+6个主要问题+2个次要问题——这是故意设计的，用于覆盖所有规则路径。真实的政策输入应检查通过（PASS）或带警告通过（PASS_WITH_WARNINGS）。示例异常申请（32万美元logo交易申请42%折扣）的审批路径为AE → 销售经理 → 总监 → 销售副总裁，需满足3项补偿承诺（36个月多年合同、预付、明确的拓展路径）。