event-budget

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Event Budget

活动预算

You model the money for one edition of a technical event. Produce five things:

The cost lines, and which of them scale with heads.
The revenue total, and where it comes from.
The cushion.
The dates cash moves.
The cut order when revenue falls short.

Consume prices - ticket, sponsor, venue, vendor - never derive them. Hand back every number as either a dated quote or a labelled placeholder.

Treat every ranking below as a default, not a law. Re-rank against what the interview turns up: a recurring event with last edition's actuals, an organizer with no accounting capacity, a fiscal host already taking a share off the top, a format with no ticket-revenue line at all. Each overturns a default rung.

No currency amount appears anywhere in this file. The real figures - published P&Ls, revenue splits, unit costs, fee ranges - live in the references with their event, year, and source, carried across as ratios, never copied as amounts.

你为一个版本的技术活动进行资金建模。产出五样东西：

成本线，以及哪些随人数变化。
收入总额，以及来源。
缓冲。
资金流动的日期。
当收入不足时的削减顺序。

消费价格——门票、赞助、场地、供应商——绝不推导。交回每个数字时要么是带日期的报价，要么是有标签的占位符。

把下面的每个排序视为默认值，而不是法则。根据访谈发现重新排序：一个有上届实际数据的周期性活动、一个没有会计能力的组织者、一个已经在顶部抽成的财政托管方、一个根本没有门票收入线的活动形式。每一项都可能推翻默认的一级。

本文件中不出现任何货币金额。 真实数字——已公布的损益表、收入拆分、单位成本、费用区间——保存在参考资料中，附带活动、年份和来源，以比率方式引用，绝不作为金额复制。

Ticketed or free-by-construction

有门票或天生免费

The axis that changes the budget work is whether the format has a ticket-revenue line at all, never who pays for the attendee. The same catering invoice arrives whether the attendee expensed the ticket or paid it themselves, and sponsors are a business buyer in both cases.

Ticketed - a genuine two-line revenue mix. Sponsorship share is a lever the team sets and can miss. Once sponsorship is booked, break-even is an attendee count at a price, and the downgrade ladder exists because the ticket side is uncertain.
Free-by-construction - typically a hackathon, sometimes a conference. Sponsorship is close to the whole funding source, not by target but because the format offers no priced alternative. Break-even collapses to a fundraising target, and the only lever left is total spend per head.

Name which one the model assumes in the first line of the output. Keep the two cost taxonomies apart too: they differ in shape and must never be blended (see references/cost-taxonomies-and-sourced-benchmarks.md).

决定预算工作形态的轴心是这种形式是否有一条门票收入线，而不是谁为参会者买单。无论参会者是报销门票还是自付，同一张餐饮发票都会到来；而且在这两种情况下，赞助商都是企业买家。

有门票——真正的两行收入组合。赞助占比是团队可以设定也可能无法达成的杠杆。一旦赞助被预订，盈亏平衡就是某个价格下的参会人数；降级阶梯存在是因为门票这侧存在不确定性。
天生免费——通常是黑客松，有时是会议。赞助几乎就是全部资金来源，不是因为有目标，而是因为这种形式没有其他可定价的选择。盈亏平衡坍缩为一个筹款目标，剩下的唯一杠杆是每人总支出。

在输出的第一行说明模型假设的是哪一种。分开两种成本分类法：它们形态不同，绝不能混用（见 references/cost-taxonomies-and-sourced-benchmarks.md）。

What arrives already decided

作为输入已经确定的内容

Take these as inputs. Re-deriving any of them produces a second, competing number nobody reconciles.

Sibling	Owns	You do
`event-ticket-pricing`	The ticket price ladder	Hand it the ticket target and the paid capacity after comps; it returns the actual sold total for variance
`event-sponsor-pricing`	Tier splits	Own the sponsorship-share target; it splits into tiers
`event-venue-sourcing`	Real quotes, terms	Hold the line-item slot, fixed/per-head character, deposit timing
`event-vendor-sourcing`	Every other supplier quote	Same split: it quotes, you shape
`event-production`	The Media line, pre-split by tier	Take it as handed
`event-no-show-management`	The expected-show-up number	Budget from it, never the registration count; the overbooking buffer is the gap
`event-team-structure`	Which entity holds the money	Model what it costs and how fast it releases funds
`event-risk-management`	The risk case for a reserve, go/no-go placement	You pick the contingency number
`event-growth-strategy` / `event-portfolio-strategy`	Whether the edition grows, portfolio split	Re-run this model at whatever scale they set

Hand post-edition variance to

event-debrief

, never re-modelling it, and hand a cost line creeping across editions to

event-continuous-improvement

. Plan and track one edition only.

把这些当作输入。重新推导其中任何一项都会产生第二个竞争数字，没有人会去对账。

同系列Skill	负责	你要做
`event-ticket-pricing`	门票价格阶梯	把门票目标和扣减赠票后的付费容量交给它；它返回实际售出总数用于差异分析
`event-sponsor-pricing`	层级拆分	负责赞助占比目标；它将目标拆分成层级
`event-venue-sourcing`	真实报价、条款	持有成本行插槽、固定/人均属性、押金时间
`event-vendor-sourcing`	所有其他供应商报价	同样拆分：它报价，你塑形
`event-production`	Media 行，已按层级预拆分	原样接收
`event-no-show-management`	预期到场人数	按它做预算，绝不用报名人数；超订缓冲就是两者差距
`event-team-structure`	哪个实体持有资金	建模其成本以及释放资金的速度
`event-risk-management`	储备的风险依据、go/no-go 位置	你选择应急数字
`event-growth-strategy` / `event-portfolio-strategy`	本届是否增长、组合拆分	以它们设定的任何规模重跑本模型

把活动后的差异交给

event-debrief

，绝不重新建模；把跨届不断攀升的成本线交给

event-continuous-improvement

。只规划和跟踪一届。

Interview

访谈

Ask one at a time, multiple-choice where possible. Questions 6-8 exist because the menus below diverge on time-to-effect, durability and effort - their defaults cannot be picked for the user.

What is the format, and does it charge for attendance at all?
What is the expected show-up number, who produced it, and is it a show-up figure or a registration count?
Which cost lines already have a real dated quote? Which are still guesses?
Which entity holds the money, and what comes off the top (fiscal-host share, processing fees, parent contribution)?
What is already contractually committed, and on what date does each commitment stop being recoverable?
By what date must sponsor and ticket money be in the account, not invoiced? A hard date promotes postures that book early.
One-off edition, or one year of a recurring event? A recurring event reuses last edition's costed tiers, making the ladder near-free the second time.
Effort ceiling: who holds a standing tracking job through the sale window, and is there accounting capacity at all?
Who approves spend, and does any expenditure need a second approver who isn't the spender?
Does the audience typically attend on employer training budgets, and does the program already clear those budgets?
Does any counterparty - venue, vendor - offer a genuine discount for paying early?

一次问一个，尽可能用选择题。问题 6-8 存在是因为下面的选项在见效时间、持久性和投入上不同——它们的默认值不能替用户选择。

活动形式是什么？是否向参会收费？
预期到场人数是多少？谁产生的？是到场数还是报名数？
哪些成本线已经有真实的带日期报价？哪些仍是猜测？
哪个实体持有资金？顶部会抽走什么（财政托管方分成、处理费、上级/母组织出资）？
哪些已经合同承诺？每个承诺到哪一天为止不可收回？
赞助和门票资金必须在什么日期前到账，而不仅仅是已开票？硬性截止日期有助于采取提早预订的姿态。
一次性活动，还是周期性活动的一年？周期性活动会复用上届已定价的层级，使阶梯在第二次时几乎免费。
投入上限：谁在销售窗口期承担持续跟踪工作？是否具备会计能力？
谁批准支出？是否有任何支出需要非支出者的第二批准人？
观众通常是否靠雇主的培训预算参会？活动议程是否已经符合这些预算要求？
是否有任何交易对手——场地、供应商——为提前付款提供真正的折扣？

Workflow

工作流程

Run the interview. Fix the format, scale, and which funding-model decisions are already made elsewhere.
List cost lines from the format's own taxonomy - conference and hackathon differ in shape, never blend them: references/cost-taxonomies-and-sourced-benchmarks.md.
Cut every line into fixed and per-head, naming each per-head driver. Both the split and the assignment of each named line to a side are this skill's own construction, carried over from general financial modelling rather than measured against event data. Take the head count from Q2's expected-show-up figure, never the registration count: budgeting off registrations is the single most common way an event over-orders. Keep the resulting cost per head on hand; it answers a "too expensive" pricing complaint with arithmetic instead of an opinion.
Fill each line with a real dated quote or an explicit placeholder. Never invent a price to balance the model - write "no quote yet" and carry it as a range.
Deduct what comes off the top first: fiscal-host share, processing fees on both rails, required parent contribution. Fee ranges: references/cash-flow-fiscal-hosting-and-tax.md.
Set the revenue mix (menu below), then hand the targets out: the sponsorship target to
```
event-sponsor-pricing
```
; the ticket target (expenses minus sponsorship share) plus paid capacity after comps to
```
event-ticket-pricing
```
. Count comps from agreements actually held, never from a ratio.
Cost the Good/Better/Best scope-tier ladder - three fully-costed versions (Good runs at all, Better adds expected niceties, Best has real headroom), break-even computed per tier. On free-by-construction, each tier is a fundraising target instead.
Size the contingency using the three-tool stack below.
Build the cash-flow schedule: every outflow on the date it leaves, every inflow on the date it lands, then read the lowest cash point. Hand checkpoint dates to
```
event-planning-timeline
```
and
```
event-risk-management
```
.
Write the tax and jurisdiction question rather than answering it. Route it to the fiscal host, a local accountant, or counsel, then record the answer as a budget line.
Track live against the plan, validating with the user section by section.

If your harness has persistent memory, record:

The line taxonomy, with each line's fixed/per-head character.
The quotes, with their dates.
The three costed tiers.
The contingency figure, and how it was reached.
The cash-flow schedule, with its lowest point.
After the edition, actual against planned, per line.

The next edition starts from that record instead of a blank spreadsheet. When rewriting a recurring per-head consumable line from what was actually used, record its calibration basis too: the headcount and staff/volunteer/speaker composition it was priced against. A later edition scaling that line then knows what baseline it is adjusting, not just the raw number.

进行访谈。确定形式、规模，以及哪些资金模式决策已在其他地方做出。
按该形式自身的分类法列出成本线——会议和黑客松形态不同，绝不混用：references/cost-taxonomies-and-sourced-benchmarks.md。
将每条线拆分为固定和人均，并注明每个人均线的驱动因素。 拆分本身以及每条具名线分配到哪一侧，是本 skill 自身的构造，来自通用财务建模而非针对活动数据测量。人数取问题 2 的预期到场数，绝不用报名数：按报名数做预算是活动超量订购最常见的原因。保留由此得到的人均成本；它用算术而非观点回应“太贵”的定价抱怨。
每条线填入真实带日期的报价或明确的占位符。绝不为了平衡模型而编造价格——写“暂无报价”并将其作为区间保留。
先扣除顶部抽成：财政托管方分成、两条支付通道上的处理费、必须的上级/母组织出资。费用区间：references/cash-flow-fiscal-hosting-and-tax.md。
设定收入组合（见下方菜单），然后把目标分派出去：赞助目标交给
```
event-sponsor-pricing
```
；门票目标（支出减去赞助份额）加上扣除赠票后的付费容量交给
```
event-ticket-pricing
```
。赠票数按实际达成的协议计算，绝不按比例估算。
为 Good/Better/Best 范围层级阶梯定价——三个完整定价的版本（Good 基本能办，Better 增加预期亮点，Best 有真正的余量），每个层级分别计算盈亏平衡点。对天生免费的活动，每个层级改为筹款目标。
用下方的三工具叠层确定应急储备规模。
构建现金流时间表：每笔流出按其离开日期、每笔流入按其到账日期，然后读取最低现金点。将检查点日期交给
```
event-planning-timeline
```
和
```
event-risk-management
```
。
把税务和司法管辖区问题写出来，而不是回答它。将其转给财政托管方、当地会计师或法律顾问，然后把答案记录为预算行。
对照计划实时跟踪，逐节与用户确认。

如果你的运行环境有持久记忆，请记录：

成本线分类法，以及每条线的固定/人均属性。
报价及其日期。
三个已定价层级。
应急储备数字，以及它是如何得出的。
现金流时间表及其最低点。
活动结束后，每条线实际与计划的对比。

下一届从该记录开始，而不是从空白电子表格开始。当根据实际使用量重写周期性的人均消耗品成本线时，也记录其校准基础：定价时所依据的人数和员工/志愿者/演讲者构成。这样以后某届调整该线的规模时，知道它在调整什么基线，而不仅仅是原始数字。

Revenue-mix posture

收入组合姿态

Which shape the funding takes. This posture is largely inherited rather than chosen here: the format and the upstream funding-model decision usually fix it. Name which one arrived, and what it implies for the model. Ranking (default, not a law - re-rank against Q1, Q6, Q7):

effort (modelling work plus the standing revenue management each posture demands):
```
registration-led > sponsor-led > sponsorship-by-construction
```
value (budget covered without the decision to run hanging on money that arrives late):
```
sponsor-led > sponsorship-by-construction > registration-led
```
efficiency:
```
sponsor-led > registration-led
```
Sponsor-led (default for a community conference) - sponsorship carries the bulk, registrations cover per-attendee cost. Decouples "will we run at all" from the ticket curve, which makes the downgrade ladder workable.
Registration-led - registration carries the bulk, sponsorship marginal. Tops effort - a sales pace managed across the whole window, money lands latest. Promotion condition, keyed to Q10: the audience attends on employer training budgets and the program already clears those budgets.
Sponsorship-by-construction - free-to-attend, no ticket line, share near-total by format not target. Deliberately absent from the efficiency line: the format chooses this posture, not this skill. Cost is concentration - one revenue source, no second lever.

What this order starves: registration-led, the only posture whose revenue scales with what the audience pays rather than with sponsor budgets. Follow the default long enough and the sponsor pool becomes the event's ceiling. A community that outgrows that pool has no second lever, and this order gives no warning until the pool is exhausted. Promote registration-led on the condition above.

Delete, do not demote: sponsor-led and registration-led, on a free-to-attend event - from this menu and the axis lines. Ranking postures that assume a ticket line invites a ticket price to reappear later as the fix for a shortfall - a funding-model reversal dressed as a budget adjustment.

Three published actuals disagree almost completely - see references/cost-taxonomies-and-sourced-benchmarks.md for the range. Read them as postures that exist, never a norm to hit.

资金采取哪种形态。这种姿态很大程度上是继承而来，而不是在这里选择的：活动形式和上游的资金模式决策通常已经确定它。说明到达的是哪一种，以及它对本模型意味着什么。排序（默认值，不是法则——根据问题 1、6、7 重新排序）：

投入（建模工作加上每种姿态所需的持续收入管理）：
```
registration-led > sponsor-led > sponsorship-by-construction
```
价值（预算被覆盖，而运行决策不依赖于迟迟才到的资金）：
```
sponsor-led > sponsorship-by-construction > registration-led
```
效率：
```
sponsor-led > registration-led
```
Sponsor-led（赞助主导）（社区会议默认）——赞助承担大头，注册费覆盖人均成本。将“我们到底办不办”与门票曲线解耦，这使降级阶梯可行。
Registration-led（注册费主导）——注册费承担大头，赞助是边际。投入最高——在整个窗口期管理销售节奏，资金最晚到账。提升条件，对应问题 10：观众靠雇主培训预算参会，且活动议程已经符合这些预算。
Sponsorship-by-construction（构造上即赞助主导）——免费参加，没有门票线，赞助占比接近全部，由形式决定而非目标决定。刻意不出现在效率行中：选择这种姿态的是活动形式，不是本 skill。代价是集中度——单一收入来源，没有第二个杠杆。

这个顺序饿死的是： registration-led——唯一收入随观众支付而非赞助预算规模变化的姿态。默认值跟随太久，赞助池就会成为活动的天花板。一个超出该池子的社区没有第二个杠杆，而这个顺序在池子耗尽之前不会给出任何警告。在上述条件下提升 registration-led。

删除，而非降级：对于免费参加的活动，从本菜单和轴心行中删除 sponsor-led 和 registration-led。 为假设存在门票线的姿态排序，会诱使门票价格在之后重新出现，作为弥补缺口的“修复”——这实际上是资金模式逆转，却伪装成预算调整。

三份已公布的实际数据几乎完全不一致——区间见 references/cost-taxonomies-and-sourced-benchmarks.md。把它们解读为存在的姿态，绝不是要达成的规范。

Contingency provision: a stack, not a ranking

应急准备：一个叠层，不是排序

This section deliberately does not rank its options.

The three tools compose: a real budget runs all three at once. Ranking them would imply picking one, the opposite of how the stack holds together.

Even if they were competing, a ranking would rest on nothing:

Value and effort run in the same order across all three - floor lowest on both, ladder highest on both.
Every pair therefore splits the two axes: each option wins one and loses the other.
No dominance relation exists anywhere in the menu for an efficiency line to honour.

An unfalsifiable ranking is not a passed check.

A floor - an absolute minimum cash reserve held regardless of budget size, covering the small certain surprises that always arrive (figure and source in references/cost-taxonomies-and-sourced-benchmarks.md). Cheapest of the three: one number, decided once.
A pad on the estimates - a percentage added to expense estimates before summing, protecting against line-item underestimation rather than revenue shortfall. The one figure available for it is a convention carried over from startup financial modelling, not measured against event data. Label it that way every time it is quoted.
A downgrade ladder - the Good/Better/Best tiers with dated checkpoints, driven by revenue actually received. The only one of the three that absorbs a whole-event shortfall rather than a line-item miss, and the most expensive: three fully-priced versions plus checkpoint dates threaded into the plan. On a recurring event (Q7) the tiers are reusable, cutting cost sharply from the second edition on.

Neither the flat floor nor the borrowed pad is an event-industry convention, and this skill will not present either as one. Say that to the user rather than producing a percentage that sounds authoritative.

本节刻意不对其选项排序。

这三个工具是组合使用的：一份真正的预算会同时运行三者。对它们排序意味着只能选一个，这与叠层的组合方式相反。

即使它们是竞争的，排序也毫无依据：

价值和投入在三个选项上以_相同_顺序排列——下限在两者上都最低，阶梯在两者上都最高。
因此每一对都在两个轴向上分裂：每个选项赢一个，输另一个。
菜单中不存在任何支配关系可供效率行遵循。

一个不可证伪的排序不是通过检查。

下限（floor）——无论预算规模如何都持有的绝对最低现金储备，用于覆盖那些总会到来的小而确定的意外（数字和来源见 references/cost-taxonomies-and-sourced-benchmarks.md）。三者中最便宜：一个数字，决定一次。
对估算加缓冲（pad）——在求和之前给支出估算加一个百分比，防止单项低估而非收入缺口。可用的唯一数字是从创业财务建模沿袭而来的惯例，而非根据活动数据测量。每次引用时都要这样标注。
降级阶梯（downgrade ladder）——带日期检查点的 Good/Better/Best 层级，由实际收到的收入驱动。三者中唯一能吸收_整个活动_缺口而非单项偏差的，也是最贵的：三个完整定价版本加上嵌入计划的检查点日期。对于周期性活动（问题 7），层级可复用，从第二届起大幅降低成本。

固定下限和借来的缓冲都不是活动行业惯例，本 skill 也不会把任何一个说成惯例。 把这一点告诉用户，而不是给出一个听起来很权威的百分比。

The cut order lives inside the ladder

削减顺序存在于阶梯之内

The cut order is the ladder's content - the tiers are this sequence, pre-priced and pre-decided. Not a second menu: presenting it separately would ship the same decision twice. Not ranked either: both readings of it (damage avoided per cut, how late the cut can be made) produce the identical order, so a ranking line would only restate it.

Each rung rests on a sourced mechanic: the scope-tier ladder behind the tiers, the attrition exposure behind the fixed lines. The order they sit in is this skill's own construction, not an observed convention.

Discretionary extras first - the evening event, swag upgrades, sponsored add-ons. This is the Best tier's own content, and cutting it protects the core experience entirely.
Per-head variable quantities second - catering covers, badge and lanyard counts - only where the guarantee hasn't locked yet. Past the lock date this rung is gone, which is why the cut order must be decided against the cash-flow schedule, not in the abstract.
Fixed structural lines last - the venue tier, the core Media line. These commit earliest and can cost more to cut than to keep. Attrition thresholds and food-and-beverage minimums are owed on what was committed, not on what turned up, so a downgrade can produce a penalty larger than the saving.

削减顺序_就是_阶梯的内容——层级就是这一序列，已预先定价和预先决定。不是第二个菜单：单独呈现它会把同一个决定出货两次。也不是排序：对它的两种解读（每次削减避免的损害、削减可以多晚进行）产生相同的顺序，因此排序行只会重述它。

每一级都建立在有来源的机制上：层级背后是范围阶梯，固定线背后是损耗风险敞口。它们所处的顺序是本 skill 自身的构造，不是观察到的惯例。

可自由决定的额外项目优先削减——晚间活动、周边升级、赞助附加项。这是 Best 层级自身的内容，削减它完全保护核心体验。
人均可变数量其次——餐饮份数、胸牌和挂绳数量——仅在保底数量尚未锁定时。一旦过了锁定日期，这一级就消失了，这就是为什么削减顺序必须对照现金流时间表来决定，而不是抽象地决定。
固定结构性成本线最后——场地层级、核心 Media 行。这些承诺最早，削减的代价可能比保留更大。损耗门槛和餐饮最低消费按已承诺的数量承担责任，而不是按实际到场数量，因此降级可能产生比节省更大的罚款。

Outflow commitment posture

支出承诺姿态

How the budget commits money against revenue not yet banked - the liquidity question, distinct from

samber/dev-event-organizer-skills@event-planning-timeline

's choice of buffer mechanic. That skill owns the calendar; this one decides how far the discipline goes and hands the dates back. Ranking (default, not a law - re-rank against Q5, Q6, Q8):

effort (tracking work, counterparty conversations, standing attention through the sale window):
```
staged against a tracked signal > latest safe date > earliest date allowed
```
value (money still in hand when a revenue line misses, and quantities committed against a real number):
```
staged against a tracked signal == latest safe date > earliest date allowed
```
compliance cost (the review each posture triggers and the reversibility it costs at the moment of payment):
```
earliest date allowed > latest safe date > staged against a tracked signal
```

efficiency:

latest safe date > staged against a tracked signal > earliest date allowed

Dominance check: 3 pairs, zero strict-dominance relations. Two mechanisms block it:

The two pairs against earliest-date-allowed fail on effort: keeping more money in hand costs strictly more tracking and counterparty attention, even though it is cheaper on reversibility.
Staged against latest-safe ties on value, each cheaper on a different cost axis (staged spends less reversibility, latest-safe less attention), so neither is at least equal on both.

Staged and latest-safe tie on value: both leave the same money uncommitted on the same date. They differ only in whether commitment arrives as one decision or several, which changes when you learn, not how much you keep. Learning is a demand read

samber/dev-event-organizer-skills@event-market-fit

already owns, so it is not counted twice here.

Compliance cost is real here, expressed as reversibility, never money. Paying before a stated cancellation date spends recoverability you were under no obligation to spend, and it does not come back.

A room block or catering guarantee committed at full size carries a usage threshold owed regardless of turnout. A small block grown in steps keeps exposure proportional to what is confirmed - see references/cash-flow-fiscal-hosting-and-tax.md for the staging tactic in practice.

Latest safe date - the default and the efficiency winner: identify, per commitment, the last date it can be made without losing the thing, and hold until then. The asymmetry is the whole argument - adding capacity late is cheap and fast, unwinding a committed order is expensive or impossible.
Staged against a tracked signal - the starved option: ties top on value, tops effort, so efficiency never leads with it. Promotion condition (Q5 + Q8): the commitment carries a threshold owed regardless of turnout and someone can hold a standing tracking job through the sale window. Both required; staging without a tracker is just a smaller order.
Earliest date allowed - the do-nothing baseline, kept in the menu only so the reader can see what it costs. It tops compliance cost and buys nothing back.

Delete, do not demote: earliest date allowed, whenever Q11 says no counterparty offers a genuine early-payment discount - from this menu and the axis lines. With nothing on the other side of the reversibility it spends, it is a pure loss that, parked at the bottom, reappears as "we may as well pay it, it's budgeted."

预算如何针对尚未到账的收入承诺花钱——这是流动性问题，区别于

samber/dev-event-organizer-skills@event-planning-timeline

对缓冲机制的选择。那个 skill 拥有日历；本 skill 决定纪律走多远并把日期交回。排序（默认值，不是法则——根据问题 5、6、8 重新排序）：

投入（跟踪工作、交易对手沟通、销售窗口期的持续关注）：
```
staged against a tracked signal > latest safe date > earliest date allowed
```
价值（某条收入线未达标时手中仍有的资金，以及针对真实数字承诺的数量）：
```
staged against a tracked signal == latest safe date > earliest date allowed
```
合规成本（每种姿态触发的审查以及在付款时刻失去的可逆性）：
```
earliest date allowed > latest safe date > staged against a tracked signal
```

效率：

latest safe date > staged against a tracked signal > earliest date allowed

主导性检查：3 对，零严格主导关系。 两个机制阻止了它：

与 earliest-date-allowed 相关的两对在投入上失败：手中保留更多资金会花费更多的跟踪和交易对手关注，即使它在可逆性上更便宜。
Staged 与 latest-safe 在价值上打平，各自在某个成本轴上更便宜（staged 花费更少可逆性，latest-safe 花费更少关注），因此两者都不在两个方面上至少持平。

Staged 和 latest-safe 在价值上打平：两者在同一日期保留同样多的资金未承诺。它们只区别于承诺是以一个决定还是多个决定到来，这改变_你何时知道_，而不是你保留多少。学习是

samber/dev-event-organizer-skills@event-market-fit

已经拥有的需求读取，所以这里不重复计算。

合规成本在这里是真实的，以可逆性表达，绝不是金钱。在规定的取消日期之前付款，会花费你本无义务花费的可追回性，而且它不会回来。

以全量承诺的房间预订或餐饮保底，无论实际上座人数多少，都带有必须承担的用量门槛。小规模预订再逐步扩大，使风险敞口与已确认量成比例——实际的分阶段策略见 references/cash-flow-fiscal-hosting-and-tax.md。

Latest safe date（最晚安全日期）——默认且效率最优：针对每项承诺，找出在不失去该事物的前提下可以做出的最晚日期，然后等到那时。不对称性就是全部论据——晚添加容量便宜且快速，而撤销已承诺订单则昂贵或不可能。
Staged against a tracked signal（按跟踪信号分阶段）——被饿死的选项：价值并列第一，投入最高，所以效率绝不会以它领先。提升条件（问题 5 + 8）：承诺带有无论实际上座人数都需承担的阈值且有人能在销售窗口期持续跟踪。两者都必须；没有跟踪者的分阶段只是更小的订单。
Earliest date allowed（允许的最早日期）——不作为基线，保留在菜单中只是为了让读者看到它的代价。它合规成本最高，却什么也买不回来。

删除，而非降级：当问题 11 表明没有交易对手提供真正的提前付款折扣时，从菜单和轴心行中删除 earliest date allowed。 它所花费的可逆性对面没有任何回报，因此是纯粹损失；把它放在底部，它会以“反正做了预算，不如就付了”重新出现。

What this skill will never state

本 skill 绝不断言的内容

A tax rate, threshold, form name, or registration requirement, as though it were general. Ticket and sponsorship revenue follow different, jurisdiction-specific rules - see references/cash-flow-fiscal-hosting-and-tax.md for contrasting jurisdiction examples. The answer comes from the fiscal host, a local accountant, or counsel.
A contingency percentage presented as an industry convention. Size it from the three tools in the stack above and label the result self-set.
A venue or vendor price benchmark. No published benchmark exists for either; three real quotes on one identical requirements sheet is the substitute, owned by
```
samber/dev-event-organizer-skills@event-venue-sourcing
```
and
```
samber/dev-event-organizer-skills@event-vendor-sourcing
```
.
A named accounting platform, ticketing platform or fiscal host as a recommendation. Fee structures appear in the references as published facts about a market, not as picks.

税率、门槛、表格名称或注册要求，仿佛它们是通用的。 门票和赞助收入遵循不同的、因司法管辖区而异的规则——对比司法管辖区示例见 references/cash-flow-fiscal-hosting-and-tax.md。答案来自财政托管方、当地会计师或法律顾问。
把应急百分比说成行业惯例。 用上方叠层中的三个工具确定规模，并把结果标注为自行设定。
场地或供应商价格基准。 两者都不存在已发布的基准；在同一份需求单上取得三个真实报价是替代方案，由
```
samber/dev-event-organizer-skills@event-venue-sourcing
```
和
```
samber/dev-event-organizer-skills@event-vendor-sourcing
```
负责。
推荐具名的会计平台、票务平台或财政托管方。 费用结构在参考资料中以关于市场的已发布事实出现，而不是作为推荐选项。

Failure modes

失败模式

Per-head costs budgeted off the registration count. The registration count is who was accepted; the expected-show-up number is who eats. Catering ordered against the first is over-ordered by construction.
A budget that is profitable and still fails. Deposits leave before sponsor invoices are paid - see references/budget-failure-case-studies.md for two cases of exactly this shape. Read the lowest cash point, not the bottom line.
Announcing a date before there is a working way to process money. Settle who holds the money, and how long it takes to get back out, before the date goes public - an organizing guide states this failure directly in references/cost-taxonomies-and-sourced-benchmarks.md.
One budget instead of three tiers. A single budget has one break-even and no move except cancelling. The ladder exists so a revenue miss has a response that is not a cancellation.
Treating the money-handling share as a rounding error. A fiscal host's cut and processing fees come off everything passing through, on both rails, before any line is funded.
Cutting a fixed structural line first. Attrition thresholds and food-and-beverage minimums are owed on the commitment, so the penalty can exceed the saving. Follow the cut order.
Quoting a contingency percentage, or a tax rule, as a standard. Neither has an event-industry standard behind it. Saying so costs nothing; inventing either gets discovered by an auditor or a tax authority.
Reading one published event's revenue split as a norm. The three published actuals disagree almost completely. They bound the space of real postures; none of them is a target.
Re-deriving a ticket or sponsor price to make the model balance. That produces a second number competing with the prices
```
event-ticket-pricing
```
and
```
event-sponsor-pricing
```
already set. Change the target and hand it back instead.

人均成本按报名数做预算。 报名数是“谁被录取”，预期到场数是“谁吃饭”。按前者订购餐饮必然超订。
一份盈利却仍然失败的预算。 押金在赞助商发票支付之前就已付出——见 references/budget-failure-case-studies.md 中两个完全这种形态的案例。要读最低现金点，而不是盈亏底线。
在还没有可用的收款处理方式之前就公布日期。 在日期公开之前确定谁持有资金，以及资金需要多久才能再出来——一本组织指南在 references/cost-taxonomies-and-sourced-benchmarks.md 中直接陈述了这一失败。
只有一份预算而不是三个层级。 单份预算只有一个盈亏平衡点，除了取消之外没有动作。阶梯存在是为了让收入未达标有一个不是取消的应对。
把资金处理分成当作舍入误差。 财政托管方的抽成和处理费会在任何一行获得资金之前，从两条通道流经的一切中扣除。
先削减固定结构性成本线。 损耗门槛和餐饮最低消费按承诺承担，因此罚款可能超过节省。遵循削减顺序。
把应急百分比或税务规则作为标准引用。 两者背后都没有活动行业标准。说明这一点不花任何代价；编造任何一个都会被审计师或税务机关发现。
把某个已公布活动的收入拆分当作规范。 三份已公布的实际数据几乎完全不一致。它们界定了真实姿态的空间；其中没有任何一个是目标。
重新推导门票或赞助价格来让模型平衡。 这会产生第二个数字，与
```
event-ticket-pricing
```
和
```
event-sponsor-pricing
```
已经设定的价格相竞争。改变目标并把它交回去。

Measurement

衡量

This skill sets one binary gate for itself, plus the self-set observables below. Pick two or three of them, and write down before the first commitment what each would change.

Gate (binary, checkable): at the first go/no-go date, every cost line carries either a dated quote or an explicit "no quote yet" label, and the cash-flow schedule has a computed lowest point. A budget with silent placeholders has not been built yet, whatever its total says. Iterate until this passes.

Variance per line, planned against actual (self-set): the number
```
samber/dev-event-organizer-skills@event-debrief
```
will ask for. Per line, never as one total - a total nets an underestimate against an overestimate and teaches nothing.
Lowest projected cash point against the lowest actual cash point (self-set): if the actual trough is far below the projection, the inflow dates were optimistic, which is a different fix from the costs being wrong.
Contingency drawn against contingency held (self-set): drawing none for several editions is evidence the floor is oversized; drawing all of it is evidence the pad is undersized.
Which tier the event actually ran at, against which was planned (self-set): a ladder that never downgrades and never upgrades is not being used as a ladder.

本 skill 为自己设定一个二元门禁，加上下面的自行设定可观测项。选择其中两三项，并在第一次承诺之前写下每项会改变什么。

门禁（二元、可检查）： 在第一个 go/no-go 日期，每条成本线要么带有带日期的报价，要么带有明确的“暂无报价”标签，并且现金流时间表已有计算出的最低点。一份带着静默占位符的预算无论总额如何都还没有构建完成。迭代直到通过。

每条线计划与实际的差异（自行设定）：
```
samber/dev-event-organizer-skills@event-debrief
```
会要的数字。按线计算，绝不要算成一个总数——总数会把低估和高估相互抵消，学不到任何东西。
预计最低现金点与实际最低现金点对比（自行设定）：如果实际低谷远低于预测，那么流入日期过于乐观，这与成本出错是不同性质的修复。
已动用应急与已持有应急对比（自行设定）：连续几届一次都没动用，说明下限过大；全部动用，说明缓冲过小。
活动实际运行在哪个层级，与计划运行层级对比（自行设定）：一个从不降级也从不升级的阶梯，并没有被当作阶梯使用。

Invocation examples

调用示例

"We are planning a 250-person community conference and have a venue quote and nothing else. Build me the budget."
"Sponsorship is at 60% of target with eight weeks to go - what do we cut, and in what order?"
"Our hackathon is free to attend. How do I work out what we need to raise, and how much cushion?"
"The venue wants a deposit in March but sponsor invoices are not due until June. Are we going to run out of money?"

Expected output: a budget model containing:

The format and revenue posture, named in the first line.
The cost taxonomy, split into fixed and per-head, with each per-head line's driver.
Every line, labelled with a dated quote or an explicit placeholder.
The top-of-budget deductions.
The sponsorship and ticket targets, with their arithmetic.
Three costed tiers, with a break-even each.
The contingency figure, with the three tools stated separately.
A dated cash-flow schedule, with its lowest point marked.
The open tax question, written as a question.

Deliver it section by section for validation, with every number labelled published or self-set.

“我们正在计划一场 250 人的社区会议，只有一份场地报价，其他什么都没有。帮我建预算。”
“距离目标还有八周，赞助只完成了 60%——我们削减什么？按什么顺序？”
“我们的黑客松免费参加。我如何算出我们需要筹多少钱，以及多少缓冲？”
“场地三月份就要押金，但赞助商发票六月份才到期。我们会不会资金耗尽？”

预期输出：一个包含以下内容的预算模型：

形式与收入姿态，在第一行命名。
成本分类法，拆分为固定和人均，并注明每个人均线的驱动因素。
每一行都带有带日期的报价或明确占位符的标签。
预算顶部扣除。
赞助和门票目标及其算术。
三个已定价层级，各有盈亏平衡点。
应急数字，三个工具分别说明。
带日期的现金流时间表，并标出最低点。
未决税务问题，写成问题形式。

逐节交付以供验证，每个数字都标注为已发布或自行设定。

References

参考资料

references/cost-taxonomies-and-sourced-benchmarks.md - the two cost taxonomies with their fixed/per-head cut, the published itemized P&Ls and revenue splits with their events, years and sources, the per-person unit benchmarks, the comp and scholarship allocation figures with their caveats, and how to carry any of it across by ratio.
references/cash-flow-fiscal-hosting-and-tax.md - deposit and attrition mechanics, sponsorship payment schedules, fiscal-host fee structures with their dates and one recent structural change, fund-access timing, and the jurisdictional tax contrasts as examples routed to counsel.
references/budget-failure-case-studies.md - five real events read for their financial mechanism rather than their headline cause, and what each one implies for a line in this model.