Calibration: aligning expectations before the question is asked校准：在发问之前把预期对齐

1. The Problem: Most Disagreements Are Not About Price

Complete and aligned are two different things. Aligned materials are often shorter.

When a round falls apart, the post-mortem is usually one sentence: we could not agree on price. Convenient, but it compresses a whole process into a single instant.

Run the clock backwards. Before the price was spoken, the two sides had already exchanged a great deal: a deck, several meetings, a set of numbers, a sentence about how much this round is meant to be. If any one of those things meant different things to the two sides, then the moment a price appears, all of them settle their accounts at once. On the surface two numbers fail to meet. Underneath, a dozen small mismatches have been amplified by one number.

The Glacier Institute makes this step a milestone of its own:

「D01 事实底座搭建 → D09 专业材料开发 → D18 市场预期校准 → D26 领投机构组织 → D35 深度尽调支持 → D43 投委会过会协同 → D52 核心条款谈判 → D60 资金交割收尾」 D01 build the fact base → D09 develop professional materials → D18 calibrate market expectations → D26 organise the lead investor → D35 support deep due diligence → D43 coordinate the investment committee → D52 negotiate core terms → D60 close and settle. — Glacier Institute website, /institute.html [1]

D18 comes before D26. Calibration is finished before the lead investor is met, not patched up afterwards. The published working paper GI-WP-2026-P2 puts the reason in one sentence:

「校准要在发问之前做，不是在被拒之后做——发出去的预期改一次，要付一次解释的钱。」 Calibrate before you ask, not after you are turned down. Every time an expectation that has gone out is revised, you pay for an explanation. — Glacier Institute Working Paper GI-WP-2026-P2 [2]

That sentence is where this paper starts. P2 places the milestone within the sixty days, in one section. This paper is about what happens inside that section: calibration is an action with a checklist, not an attitude.

A companion paper in the same series, GI-WP-2026-P8, argues an adjacent point: irreversible actions should go one beat slower than reversible ones, and should be taken by whoever bears the consequence. This paper follows it by saying what happens during that beat. The two do not overlap: P8 is about when to press; this paper is about which twelve items to align before pressing.

This paper answers two questions founders ask. First: what actually has to be prepared before a round opens. Second: the materials are complete, so why does the room still take them apart. Both have the same answer: complete and aligned are two different things. Materials can be complete without being aligned, and aligned materials are often shorter. The Glacier Institute's phrasing:

「标准是自洽，不是齐全。每一个数字、每一张图，都要经得起反过来查。这一层的活，说白了就是举证。」 The standard is internal consistency, not completeness. Every number and every chart must survive being checked in reverse. The work at this layer is, plainly, producing evidence. — Glacier Institute website, /institute.html [1]

Section 10 answers both questions directly. The route: Section 2 defines calibration and separates it from expectations management and from negotiation; Section 3 gives four mechanisms; Section 4 brings evidence from one public template; Section 5 sets out the Glacier Institute's published practice; Section 6 gives a checklist usable the next day; Section 7 states the claim as a proposition that can be refuted; Sections 8 and 9 mark the boundaries.

---

2. Definition: What Calibration Is, and What It Is Not

"Valuation is one notch apart" usually calls for aligning premises, not price.

2.1 An operational definition

Calibration, in this paper, is this action: before making any request of an outside party, put every premise that both sides will rely on side by side on one table, and confirm that they refer to the same thing.

Three points follow. First, it happens before the ask. What happens after the ask is called explaining, and it costs differently. Second, it handles premises, not conclusions. Price is a conclusion; the choice of peer companies is a premise. Segment assignment is a premise; a valuation range is a conclusion. Third, it must be item by item. An expectation is not one thing; it is a dozen things, and any one of them left unaligned can make the whole look like a disagreement about price.

2.2 It is not "managing expectations"

In ordinary usage, managing expectations means talking the number down so that it can be met. That is a useful move, but it is not calibration. The Glacier Institute draws the line directly:

「硬度可以留在判断里，但是不能留在报价里。管理预期不是让步，是把一笔做不成的交易换成一笔做得完的。先能做完，再谈好坏。」 Firmness may stay in the judgment; it must not stay in the quote. Managing expectations is not conceding; it is trading a deal that cannot be completed for one that can. First make it completable, then argue about how good it is. — Glacier Institute website, /institute.html [1]

Read through this paper's lens: "firmness may stay in the judgment" says calibration does not require you to give up your view; "must not stay in the quote" says a view that has not been aligned should not be turned straight into a number. Calibration works on the stretch between a judgment and a number, not on the judgment.

2.3 It is not negotiation either

Negotiation presupposes that both sides know what they are arguing about. Calibration is what makes them know. In order, calibration precedes negotiation; in kind, negotiation allocates value while calibration removes ambiguity. Put calibration inside a negotiation and an awkward thing happens: one side thinks it is clarifying while the other thinks it is being pushed down on price. The Glacier Institute has a more general phrasing for that mismatch:

「答非所问，比答错更贵。」 Answering the wrong question costs more than answering wrongly. — Glacier Institute website, /institute.html [1]

A wrong answer at least means both sides were on the same question. Answering the wrong question means even the question was not shared, and that is hard to notice in the room. The original continues:

「为难之处在于，双方都以为在认真对话。」 The difficulty is that both sides believe they are having a serious conversation. — Glacier Institute website, /institute.html [1]

2.4 Two kinds of disagreement

Table 1 sorts disagreements in a round into two kinds. One criterion only: after the premises have been laid out item by item, is the disagreement still there?

表 1 / Exhibit 1　Two kinds of disagreement in a financing round.

Source and notes: The criterion is whether the disagreement survives laying out the premises item by item. This paper's classification; no transaction record appears.

How it looks	What it actually is	Criterion	How to handle
Valuation is one notch apart	Possibly a mismatch in segment assignment	Are both sides talking about the same segment's exam paper	Align the segment first, then discuss price
Valuation is one notch apart	Possibly a mismatch in peer companies	Do both sides have the same firms in mind as peers	Align the peer list first, then discuss multiples
Valuation is one notch apart	A real disagreement about price	Still one notch apart after every premise is aligned	Negotiate
"The data is not enough"	Possibly a mismatch in time window	Are both sides using the same cut-off date	Align cut-off dates and statistical scope
"Too much dilution"	Possibly a mismatch on the option pool	Is the pool counted pre-money or post-money	Align the arithmetic, then look at the number
"Let us keep watching"	Possibly a mismatch on who else is at the table	Who does the other side think is still looking	Align the list and the tempo
"This is not what you said last time"	A mismatch in phrasing	Are the two statements the same version	Align versions, mark the update date

The first three rows look identical on the surface: valuation is one notch apart. Their handling is entirely different, and only the third belongs in a negotiation. Treating the first two as the third means bargaining without a shared premise, and every concession made there endorses a frame that has not yet been agreed.

---

3. Mechanism: Why Skipping Calibration Guarantees Disagreement

The benchmark you offer becomes the ruler used on you later.

Section 2 was classification. This section is why the classification is needed — four mechanisms, each showing that an unaligned premise does not sit quietly; it actively drags the conclusion.

3.1 A number, once spoken, starts pulling other people's judgment

Tversky and Kahneman [7] showed that once people are exposed to an initial value, subsequent judgments drift toward it, even when they know the value was generated at random. They called it anchoring and adjustment: people start from the anchor and adjust, and the adjustment is typically insufficient.

The effect is not a laboratory artefact. Northcraft and Neale [8] took real-estate professionals to an actual property, gave them different listing prices, and asked for independent appraisals: the appraisals shifted systematically with the listing price, and the professionals generally denied being influenced. Epley and Gilovich [9] separated two routes — self-generated anchors, which are actively adjusted, and externally provided anchors, which work through priming — and neither is removed by telling yourself to be careful.

The implication for a round is direct: a number spoken before it has been aligned does not wait around to be corrected. It has entered the other side's judgment, and every later explanation must drag back from it. The Glacier Institute puts it in two sentences:

「所以硬拉估值意义不大。价格能被拉高一次，但是拉不住第二次。」 So forcing the valuation up means little. A price can be pulled up once; it cannot be held up a second time.

「这张牌不争，估值的锚就被别人先钉住了。」 If you do not contest this card, someone else nails the valuation anchor first. — Glacier Institute website, /institute.html [1]

The first says an anchor you give cannot be withdrawn; the second says that if you give none, someone else will. Together they explain why calibration can be neither skipped nor postponed: skip it and someone else drives the nail; postpone it and you drive it yourself into a spot you have not verified.

3.2 Whoever offers first fixes the result

Negotiation research gives the sharper form. Galinsky and Mussweiler [10] found that in dyadic bargaining the party who makes the first offer significantly shapes the settlement — the first offer works as an anchor. They also found two things that weaken it: focusing on one's own alternative, or taking the other side's perspective and reservation point. Ritov [11] obtained results in the same direction in simulated competitive market negotiation.

Two points carry into practice. First, the power of a first offer does not come from being right; it comes from being first. Second, both of the things that weaken it are forms of preparation: the alternative has to be worked out before sitting down, and the other side's position has to be researched before speaking.

So this paper argues neither "never offer first" nor "always offer first". It argues: the precondition for offering first is that calibration is finished. Offering first without it means using a number you are not yet sure of to pin the other side's judgment — and you will be the one pulling that nail back out.

3.3 Why words do not work and actions do

Information economics supplies the other half. Akerlof [4] showed that when the seller knows more about quality than the buyer, adverse selection drives good goods out of the market. Spence [5] gave the remedy: the seller can take an action that is cheaper for high-quality types, and use it to separate; the informational value of that action comes from its cost structure, not from its content. Leland and Pyle [6] carried the same logic into financing: how much equity an entrepreneur retains is itself a signal of project quality, because bearing that risk is costlier for a weaker project.

This explains something practitioners see constantly: the same content carries entirely different weight when it sits in a deck versus when it has been verified by a third party. Calibration is not saying things better; it is converting statements into verifiable form. This is why the Glacier Institute writes evidence-production into its method:

「像企业考古学家：重建业务、技术、客户、交易历史与关键风险，让所有判断建立在同一组事实之上。」 Like corporate archaeologists: reconstruct the business, the technology, the customers, the transaction history and the key risks, so that every judgment rests on one and the same set of facts. — Glacier Institute website, /institute.html [1]

"One and the same set of facts" is the smallest unit of calibration. It does not ask the two sides to reach the same conclusion; it asks them to work from the same facts.

3.4 Walking expectations down in advance has a price

Accounting research on how listed companies manage market expectations of their own results supplies evidence from a different direction.

Skinner [12] found that firms prefer to disclose bad news voluntarily rather than let it surface on the earnings date. Matsumoto [13] showed that firms guide expectations in order to avoid negative surprises. Bartov, Givoly and Hayn [14] quantified the market-reaction side: firms that meet or beat analyst expectations earn a premium after controlling for the level of earnings, and the premium also holds for firms that guided expectations down first and then met them. Healy and Palepu [15] place this class of behaviour, in their review, among the governance tools for information asymmetry.

That literature is not the primary market, but one finding carries over: the market punishes surprise, not bad numbers. A low number stated in advance costs less than a high expectation that suddenly fails. The Glacier Institute's version sounds more like ordinary speech:

「上来把自己对标成天花板，对方随即按天花板的尺子来量你。之后每一步真实进展，都变成「不及预期」。把预期留在真实进展之下，不是客气，是留余量的习惯。谦虚在这里不吃亏，但是它需要沉住气。」 Benchmark yourself against the ceiling at the outset, and the other side will immediately measure you with the ceiling's ruler. Every real step of progress after that becomes "below expectations". Keeping expectations below real progress is not politeness; it is the habit of leaving margin. Modesty does not cost you here, but it does require patience. — Glacier Institute website, /institute.html [1]

"Measure you with the ceiling's ruler" is exactly the anchoring of 3.1: the benchmark you offer becomes the ruler used on you later. So peer companies must be on the calibration checklist, and near the top of it.

3.5 The contract layer: a sentence eventually becomes a number

Venture-contracting research shows where all these premises end up. Kaplan and Strömberg [16], reading contracts one by one, found that they allocate cash-flow rights, voting rights, board seats and liquidation rights separately, and make them contingent on performance. Gornall and Strebulaev [17] show that once the protective terms given to later investors are counted, the gap between headline valuation and fair value is large — which is to say an isolated valuation figure is not a quantity that can simply be compared.

For calibration the implication is: every unaligned premise has a matching blank in the term sheet. The blank must be filled with a number, and at that moment ambiguity stops being ambiguity and becomes a fixed value both sides must live by. Section 4 counts those blanks in a public template.

---

4. Evidence: How Many Blanks a Public Template Leaves

Where the template leaves a blank is exactly where calibration is needed.

4.1 Source and method

The document used here is the NVCA model term sheet, July 2020 edition [18]. It is among the most widely used templates for US Series A financings, is publicly downloadable, and every figure in this section can be reproduced by anyone.

Method: the original .docx (37,889 bytes) was fetched with a standard download tool; word/document.xml was read per the Office Open XML format; after stripping markup, 124 lines of plain text remain. Clause headings were then identified line by line (short lines ending in a colon), and within each clause's body we counted bracketed blanks containing underscores (of the form [______], $[_____], [__]%). The script and the plain-text extract are archived in the working-papers directory listed in the writing notes; line numbers are reproducible.

The file contains 45 clause headings in total. This paper takes only the six that bear directly on expectations, quotes each verbatim, and says which sentence each blank converts. The "blanks" column in Table 2 was counted by this paper, not quoted from anyone.

4.2 Six clauses that bear directly on expectations

表 2 / Exhibit 2　The six clauses in the NVCA model term sheet (July 2020 edition [18]) that bear directly on expectations.

Source and notes: "Blanks" is the count, made by this paper, of bracketed slots in that clause's body which must be filled with a number or a name; "Line" refers to the 124-line plain-text extract described in 4.1. The quoted column is verbatim; square brackets are the template's own blanks. The last column is this paper's reading.

Clause (original heading)	Line	Blanks	The parts left blank in the original (verbatim)	Which sentence this blank turns into a number
Pre-Money Valuation	L21–22	3	"a fully-diluted pre-money valuation of $[_____]"; "an [unallocated and uncommitted] employee option pool representing [__]% of the fully-diluted post-money capitalization"; "a fully-diluted post-money valuation of $[______]"	"We are worth roughly this much this round" → one pre-money figure, one pool percentage, one post-money figure. The pool sits inside the pre-money sentence, which assigns its dilution to the existing holders
Amount Raised	L19–20	2	"$[________], [including $[________] from the conversion of SAFEs/principal [and interest] on bridge notes]"	"We want to raise some money" → a total, and how much of it is old convertible instruments converting in
Dividends	L24–27	2 (plus 3 alternatives)	"$[_____] per share" (Alternative 2); "an annual [__]% cumulative dividend" (Alternative 3)	"Is there a dividend" → a choice among three structures, plus a rate. Choose the cumulative one and it is a claim that accrues year after year
Liquidation Preference	L28–33	5 (plus 3 alternatives)	"[__ times] the Original Purchase Price" (once in each of Alternatives 1, 2 and 3); "an aggregate of [_____] times the Original Purchase Price" (Alternative 3); "holders of [___]% of the Series A Preferred"	"How do we split on exit" → a multiple, a cap multiple, and a threshold percentage
Founders' Stock	L115–116	2	"Buyback right/vesting for [__]% for first [12 months] after Closing; thereafter, right lapses in equal [monthly] increments over following [__] months"	"The founders are in it for the long run, obviously" → a percentage and a number of months
No-Shop/Confidentiality	L119–120	2	"for a period of [______] days from the date these terms are accepted"; "acceptable to [_________], as lead Investor"	"This round we will talk to you" → a number of days, and a named lead

4.3 What the table shows

First, the template supplies not one default value. Every slot in Table 2 is empty. The template does not say "typically so many days" or "typically so many points". These are not industry constants; they are negotiated each round — and the precondition for negotiating them is that both sides understand the sentence behind the number in the same way. Where the template leaves a blank is exactly where calibration is needed.

Second, the option-pool blank is hidden inside the valuation sentence. The verbatim text in the first row of Table 2 places the "unallocated and uncommitted employee option pool" inside the definition of the pre-money valuation. The same sentence also carries the post-money figure. In other words, the term "pre-money valuation" itself requires first settling whether the pool is included and how large it is. Two parties each quote a pre-money number on their own understanding; the numbers may look barely apart while their meanings differ by an entire pool. This is the archetypal case of a premise mismatch wearing the face of a price disagreement.

Third, the template itself states which clause binds from signature. The introductory paragraph of the 2020 edition reads:

"In consideration of the time and expense devoted and to be devoted by the Investors with respect to this investment, the No Shop/Confidentiality provisions of this Term Sheet shall be binding obligations of the Company whether or not the financing is consummated. No other legally binding obligations will be created until definitive agreements are executed and delivered by all parties." — NVCA Model Term Sheet, July 2020 edition [18]

In the whole term sheet, only the no-shop and confidentiality provisions take effect from the day of acceptance; everything else waits for definitive agreements. So the number of days in the last row of Table 2 is the earliest clock in the document — once it is filled in, half the window for calibration has already closed.

What this paper does not do should be stated: Table 2 contains no market distribution of days, percentages or multiples. Such statistics require transaction samples with process records; public templates do not contain them, and this paper does not invent them. Section 9 lists this as a gap.

---

5. The Glacier Institute's Practice: Facts First, Then Let the Market Speak

Content must be uniform; sequence may be designed.

What follows is the Glacier Institute's published practice. The original sentences are taken from the website and from published working papers without rewriting; what this paper adds is a re-ordering by calibration.

5.1 The first step is not to speak, but to find out what they want

「第一步不是开口，是先弄明白对方要什么。」 The first step is not to speak, but to work out what the other side wants.

「嘴上说的和心里要的常常不是一回事。有人要的是单位时间的效率，有人要的是估值上一个台阶，有人只要现金落袋。三种人问出来的问题往往一模一样，但是要的东西差着一个量级。」 What people say and what they want are often not the same. Some want efficiency per unit of time, some want the valuation to move up a step, some simply want cash in hand. All three ask questions that look identical, yet what they want differs by an order of magnitude. — Glacier Institute website, /institute.html [1]

"Questions that look identical" is the real-world version of Table 1: same appearance, different class. The first piece of calibration work is to separate things that look alike, and the way to separate them is not guessing but asking and checking.

5.2 The fact base comes before the materials

「事实底座还原是第一件事：专利、订单、营收质量、团队结构、潜在风险，一项一项翻干净。材料 (deck) 里含糊过去的一句话，到尽职调查（尽调，due diligence）阶段要花十倍的时间解释。改一页纸，那时已经变成改一张时间表。跳过这一步，翻译就容易变成创作。」 Restoring the fact base comes first: patents, orders, revenue quality, team structure, latent risks — turned over one by one until clean. A sentence glossed over in the deck takes ten times as long to explain by the due-diligence stage. Changing one page has by then become changing the whole schedule. Skip this step and translation easily turns into invention. — Glacier Institute website, /institute.html [1]

That gives calibration's cost curve: the same vague line costs one sentence to fix at the fact-base stage and one schedule to fix at due diligence. The eight milestones put the fact base at D01, the materials at D09, and calibration at D18 — the order is itself a cost ranking.

5.3 One set of facts, released in windows

「同一组事实分窗口同步给不同相关方，谁先看、谁后看，是节奏。定价由市场完成。」 One and the same set of facts, synchronised to different parties in windows; who sees it first and who later is tempo. Pricing is completed by the market. — Glacier Institute website, /institute.html [1]

This separates two things that are often conflated: content must be uniform; sequence may be designed. Calibration does not mean everyone sees everything at once; it means different people see different slices of the same set of facts, rather than different facts. That matches the signalling logic of 3.3 — verifiability comes from the identity of the facts, not from skill in the telling.

5.4 Internal alignment counts too

「四查同舟。这一条我们看得最重。老股东、联合创始人、核心团队的预期，要在起飞前摆到同一张桌上。」 Check four: same boat. We weigh this one most heavily. The expectations of existing shareholders, co-founders and the core team must be on one table before launch.

「同舟不能有人中途想回去。老股东与联合创始人的预期，要在起飞前对齐，不是在半路。」 In the same boat, nobody can want to turn back halfway. The expectations of existing shareholders and co-founders are aligned before launch, not en route. — Glacier Institute website, /institute.html [1]

This is the line most often missed on a calibration checklist. However finely external expectations are prepared, anyone inside holding a different set of numbers will eventually put them on a table — usually at the most expensive moment.

5.5 After calibration, the market still prices

「先在市场上校准，再持续修正齿形——错一次，修一次。」 Calibrate against the market first, then keep filing the key's teeth — wrong once, corrected once. — Glacier Institute website, /institute.html [1]

This guards against a misreading this paper might invite: that once calibration is done, you may set your own price. You may not. Calibration makes the market's answer interpretable; it does not answer for the market. Another line on the same page puts it harder:

「我们不强扭共识，只收双方已经熟透的那颗果子。熟透的一句话就落地，青的说破嘴也拧不下来。熟没熟，谁说了算？市场，不是我们。」 We do not force consensus; we pick only the fruit that is already ripe for both sides. A ripe one lands in a sentence; an unripe one will not come off however much you talk. Who decides ripeness? The market, not us. — Glacier Institute website, /institute.html [1]

5.6 Who decides

「要不要接一份条款清单（term sheet）、要不要调整预期、要不要换掉整条航道，最后拍板的永远是企业。我们的活，是把信息、选项和后果完整地摆在船长面前。摆完就退后一步。」 Whether to accept a term sheet, whether to adjust expectations, whether to change the whole course — the final call always belongs to the company. Our work is to lay the information, the options and the consequences in front of the captain, in full. Then step back. — Glacier Institute website, /institute.html [1]

"Whether to adjust expectations" appears in the same sentence as "whether to accept a term sheet", which means that in this framing adjusting expectations is a call the principal makes, not an errand to be handled. That is consistent with GI-WP-2026-P8 and is not expanded here.

---

6. The Tool: A Calibration Checklist

The first six rows are about computing correctly; the last six about staying consistent. What goes wrong is usually in the last six.

The first five sections compress into one table. Table 3 lists twelve items to align before the ask, each with three cells: what is aligned, what happens if it is not, and which materials do the aligning. The order runs from cheapest-to-align-early downward, roughly following the cost curve in 5.2.

The way to use it is not to tick boxes but to ask one question of each row: on this item, are both sides talking about the same thing? The rows you cannot answer are the most expensive part of this round.

表 3 / Exhibit 4　The calibration checklist.

Source and notes: Twelve items to align before the ask, ordered by how much cheaper it is to align them early. This paper's compilation; the term-sheet facts are from Table 2 and [18], and milestone codes follow the Glacier Institute's published eight-milestone statement [1][2]. The table contains no transaction record and no figure of the Institute's own.

No.	What is aligned	What happens if it is not	Which materials do the aligning
1	The fact base: patents, orders, revenue quality, team structure, known risks, each turned over until checkable	A vague sentence costs ten times as long to explain at due diligence [1], and surfaces at the most expensive moment	Source documents and third-party verifiable records; every figure carries a cut-off date and a scope
2	The choice of peers: which companies we are placed beside	The benchmark you offer is the ruler used on you [1]; benchmark too high and every real step becomes "below expectations"	A one-page peer note: who was chosen, why these, and which firms are not peers and why
3	Segment assignment: which segment this company stands in, which exam paper it should answer	Answering the previous segment's paper on this segment's question, with both sides believing they are in a serious conversation [1][3]	The segment framework and the reason for this company's assignment; the previous segment's answered questions listed separately
4	Time window: the cut-off date and statistical scope of every figure	Two sides compare numbers from different cut-offs and it looks like a disagreement about performance	A cut-off table: one row per key figure, with date, scope and last update
5	Dilution and option pool: pre-money or post-money, how large, sufficient until when	The template puts the pool inside the pre-money sentence [18]; each side quotes on its own reading and the meanings differ by a pool	A cap table computed both ways, pre-money and post-money, shown side by side
6	Round size and structure: how much is raised, how much of it is old convertible instruments converting in	The template requires the total and the converting portion to be filled separately [18]; blur them and dilution cannot be computed	A sources-and-uses breakdown; conversion terms of old instruments listed separately
7	Milestones: which specific things this money buys, and the acceptance criterion for each	Vague milestones leave nothing to evidence in the next round	A milestone table: event, acceptance criterion, expected date, resources attached
8	The participant list: who is in this round, who has seen it, who is synchronised	The other side assumes others are (or are not) looking, and reads the tempo wrongly	The list and a tempo note; fix the version before it goes out
9	Which figures may be public: which numbers may be said, to what level of detail, in what form	A number spoken before it is aligned does not wait to be corrected [7][8]	A disclosure tier table: one row per figure — public, restricted, not to be circulated
10	Internal expectations: the price and timing expectations of existing shareholders, co-founders and the core team	Anyone holding a different set of numbers will produce them at the most expensive moment [1]	One internal alignment meeting; write down each party's floor rather than confirming it verbally
11	Who says it: which statements the principal makes, which may be delegated	The same sentence carries different weight from different speakers, and costs differently to retract	A division-of-labour table with two columns, split by whether a misstatement can be withdrawn
12	How a misstatement is corrected: who corrects it, within what time, in what form	Corrections do not propagate on their own; old versions keep being quoted	One correction path: a single current-state table plus an update time; do not rely on chasing the message stream

The first six rows are "facts and arithmetic"; the last six are "who says it, how far, and what happens when it is wrong". Most teams spend their effort on the first six, while what actually goes wrong tends to sit in the last six — the first six are about computing correctly, the last six about staying consistent.

One more word on row 2. Choosing peers looks like a technical task; it is in fact the most expensive cell in this paper. It simultaneously determines segment assignment (row 3), which figures may be public (row 9) and internal expectations (row 10). Get it wrong once and the other eleven rows tilt with it. That is the subject of the next working paper in this series.

---

7. A Proposition That Can Be Refuted

Calibration has to hold both directions. Only pushing expectations down is not calibration.

The claim of this paper can be written as a proposition, together with the evidence that would refute it.

Proposition. If, before a formal request goes out, the two sides confirm the twelve items of Table 3 one by one, then in the process that follows, the share of price-related disagreements attributable to "different premises" should be significantly lower than where no such confirmation was made; and disagreements should be discovered earlier and more dispersed in time, rather than concentrated in the one meeting after a price is named.

Corollary one (template level). Where a general-purpose financing template requires a number, it should also state the definition the number attaches to, rather than leaving only a blank. Table 2 shows the NVCA 2020 edition writing the option-pool definition into the same sentence as the pre-money valuation [18], consistent with the corollary. If some general-purpose template wrote "pre-money valuation" as a bare blank without a definition, and no systematic dispute followed, corollary one is weakened.

Corollary two (process level). In financing samples with process records, the group that confirmed premises item by item should show fewer "premise mismatches first surfacing after a price was named". If the two groups do not differ on this measure, or the confirming group shows more, the claim does not hold.

Corollary three (direction). Calibration should reduce deviation in both directions. If item-by-item confirmation is observed to reduce only downward revisions while upward revisions stay flat or rise, then the mechanism at work is not "aligning premises" but "talking expectations down" — which Section 2.2 explicitly excludes — and the proposition must be rewritten.

What would refute the proposition: financing samples with process records, and such records are mostly within confidentiality scope. What this paper can do is state the proposition clearly so that whoever has the samples can test it. Corollary one has been tested once against a public template and was not refuted; corollaries two and three have not been tested by anyone.

---

8. Boundaries and Counter-Examples

An unverified premise should be written "pending", not filled in with a number.

A checklist supported only by favourable cases does not deserve trust. Four situations follow where this method is known to fail, or must be read in reverse.

Counter-example one: calibration is not conceding. The most common misuse is to execute "align expectations" as "lower expectations until the other side accepts". The accounting literature cited in 3.4 invites that misreading — firms guide expectations down and then meet them. But that literature measures surprise, not level: the premium comes from the absence of a negative surprise [13][14], not from a low number. The Glacier Institute's phrasing closes this boundary: 「硬度可以留在判断里，但是不能留在报价里」「管理预期不是让步」[1] — firmness may stay in the judgment but not in the quote; managing expectations is not conceding. The test is corollary three of Section 7: if only downward revisions fall, it was not calibration.

Counter-example two: alignment is not agreement. Calibration requires the same premises, not the same conclusion. If a notch of difference remains after every premise is aligned, that is a genuine disagreement about price (row three of Table 1) — and a valuable one: it means both sides answered the same question differently, which is precisely where market pricing belongs. The Glacier Institute's phrasing is 「定价由市场完成」[1] — pricing is completed by the market. Pushing calibration to "both sides must think alike" crosses the line.

Counter-example three: calibration itself creates anchors. This risk comes from the method itself. Row 2 of Table 3 asks you to settle peer companies first, row 3 to settle segment assignment first — and once spoken, both become the ruler used on you later. The anchoring research of 3.1 admits no exception here: an anchor you volunteer is just as hard to withdraw [7][9]. So the correct use of the checklist is not "put every premise on the table" but put each one on the table only once it is verifiable. An unverified premise is listed as "pending", written explicitly as its own cell; do not fill in a number in order to look prepared.

The other face of the same rule: the Glacier Institute builds "facts first, claims after" into its own structure [1]. Facts may go first; claims wait until the facts stand.

Counter-example four: some expectations can only be calibrated by the market. Several of the twelve items simply have no reference point at a very early stage. The published working paper GI-WP-2026-P1, discussing its own scope, notes the same situation: when a company has no price band yet to be calibrated against, this step loses its reference [3]. Forcing calibration then produces a set of premises you made up, which is worse than not doing it — every later conversation stands on a fictional coordinate. The right move is to mark those cells explicitly as "not yet determined" rather than filling in a figure that merely looks clear. The Glacier Institute's original:

「先在市场上校准，再持续修正齿形——错一次，修一次。」 Calibrate against the market first, then keep filing the key's teeth — wrong once, corrected once. — Glacier Institute website, /institute.html [1]

"Wrong once, corrected once" concedes that calibration iterates. The checklist here is a starting point, not an answer.

---

9. The Scope Within Which This Checklist Holds

This paper recommends no values. What is on the checklist are questions, not answers.

One, subject matter. This paper addresses preparation within an equity financing process, from building the fact base to sending out the formal list. Mergers, debt financing and public offerings also involve expectation alignment, but the items differ, and this paper does not extrapolate. The data in Table 2 come from a general-purpose US template; clause names and structures cannot be transplanted to other jurisdictions. What the table is used to show is the shape of "how a sentence becomes a number", not the clauses themselves.

Two, the operational definition of calibration. See 2.1: before the ask, put every premise both sides will rely on side by side, and confirm they mean the same thing. It does not require identical conclusions (counter-example two), does not equal lowering expectations (counter-example one), and cannot substitute for market pricing (counter-example four).

Three, strength of evidence. The anchoring and negotiation research [7]–[11], information economics [4]–[6], the expectations literature [12]–[15] and venture-contracting research [16][17] are peer-reviewed; they support the mechanism. Table 2 was counted by this paper from a public template, the clause text is unaltered, and the "blanks" and "which sentence this blank turns into a number" columns are this paper's count and reading, not the original authors' conclusions. Tables 1 and 3 come from the Glacier Institute's published statements and this paper's compilation, and have not been independently tested.

Four, limits of transfer. The expectations literature in 3.4 studies listed companies in public markets, where analyst expectations provide a continuously observable quantity. The primary market has no such quantity. This paper takes only one directional finding from it — the market punishes surprise, not bad numbers — and does not claim its quantitative results carry over.

Five, what it cannot do. This paper gives no recommended value for any price, range, number of days, percentage or multiple; it does not judge what any round should be priced at; and it is not investment advice. Table 3 is a list of questions, not a list of answers.

Six, an honest gap. What this paper has not done: use traceable transaction samples to measure the share of disagreements attributed to price that were in fact unaligned premises. That needs samples with process records, and such records are mostly within confidentiality scope. Until that work is done, the classification here is a falsifiable working hypothesis; Section 7 has already written it in testable form.

---

10. Conclusion

Disagreements in a round all look like disagreements about price. Lay the premises out item by item and most of them disappear during the laying out; what remains is what belongs in a negotiation.

Anchoring research says a number once spoken drags later judgment, and professionals are not exempt. Negotiation research says whoever offers first fixes the result, and both of the things that weaken that effect are forms of preparation. Information economics says verifiable actions beat statements, because their informational value comes from their cost structure. The expectations literature says the market punishes surprise, not bad numbers. Contracting research says every unaligned premise ends as a blank in a clause that must be filled. Counted clause by clause, one public template leaves sixteen blanks across the six clauses that bear directly on expectations, and not one default value; one of those blanks, the option pool, sits inside the definition of the pre-money valuation.

Back to the two questions of Section 1.

What actually has to be prepared before a round opens. Not a complete set of materials, but the twelve items of Table 3 brought to the point where both sides are talking about the same thing. The order follows cost: fact base, peers, segment assignment and time window come first, because if they are wrong the other eight tilt with them. Items that cannot be settled are written explicitly as "pending"; do not put a number there to hold the place.

Why a complete deck still gets taken apart in the room. Because complete and aligned are two different things. Completeness is about one's own side; alignment is about both sides being identical. The questions that take a deck apart are usually not where the material is missing, but where the material exists and the two sides read it differently — most typically the pre-money valuation: two numbers that look barely apart, differing by an entire option pool.

Three sentences summarise the paper:

1. Most disagreements in a round are not about price; they are expectations that were never aligned before the ask.
2. The benchmark you offer becomes the ruler used on you later; an unverified premise should be written "pending".
3. Calibrate before you ask. Every time an expectation that has gone out is revised, you pay for an explanation.

---

Eleven papers are not eleven topics. They are eleven links on one chain: segment first, then sequence, then take apart the one move in each link that most often goes wrong.

Read nextGI-WP-2026-P8　One-way doors: irreversible actions in the primary market, and who presses them校准是不可逆动作之前那一拍要做的事。P9 是 P8 的前置。

FOR FOUNDERS

For founders这一轮怎么打、要多久、我们接不接，一次说清。进去看 →

转人工 · ASK A HUMAN把这个问题交给赵羚茜（Gogo Zhao，投资人关系负责人）安排——她会把它交到对的人手里。 联系庚辛

庚辛人生（Certainty of Money） · 庚辛研究院通讯论文 · GitHub · PDF 全文 · 可引用，CC BY 4.0。

References

Format: author (year). Title. Journal/publisher, volume(issue), pages. DOI or accessible link. Grouped by source tier: primary material is the Glacier Institute's published external statements; secondary material is peer-reviewed literature and public template documents. Every entry below was checked on 2026-09-22 by a single script issuing a real request; no unverified entry is listed. Item-by-item results appear in the writing notes.

Primary material (the Glacier Institute's published statements; all quoted sentences come from [1]–[3] unaltered)

[1] Glacier Institute (2026). Glacier Institute website (including the "four-layer vortex", "finding the key", "five pre-launch checks" and "60-day benchmark cadence" sections and the essays). Glacier Capital official website. https://glacier.mba/institute.html (accessed 2026-09-22)

[2] Glacier Institute (2026). Sixty days, eight checkpoints: the tempo of a fundraise. Glacier Institute Working Paper GI-WP-2026-P2. https://doi.org/10.5281/zenodo.22869123 (landing page https://glacier.mba/research/GI-WP-2026-P2.html, accessed 2026-09-22)

[3] Glacier Institute (2026). The $0–100bn valuation band: segment structure of the primary market. Glacier Institute Working Paper GI-WP-2026-P1. https://doi.org/10.5281/zenodo.22869076 (landing page https://glacier.mba/research/GI-WP-2026-P1.html, accessed 2026-09-22)

Secondary literature (information economics, anchoring and negotiation, expectations management, venture contracting)

[4] Akerlof, G. A. (1970). The Market for "Lemons": Quality Uncertainty and the Market Mechanism. The Quarterly Journal of Economics, 84(3), 488–500. https://doi.org/10.2307/1879431

[5] Spence, M. (1973). Job Market Signaling. The Quarterly Journal of Economics, 87(3), 355–374. https://doi.org/10.2307/1882010

[6] Leland, H. E., & Pyle, D. H. (1977). Informational Asymmetries, Financial Structure, and Financial Intermediation. The Journal of Finance, 32(2), 371–387. https://doi.org/10.1111/j.1540-6261.1977.tb03277.x

[7] Tversky, A., & Kahneman, D. (1974). Judgment under Uncertainty: Heuristics and Biases. Science, 185(4157), 1124–1131. https://doi.org/10.1126/science.185.4157.1124

[8] Northcraft, G. B., & Neale, M. A. (1987). Experts, Amateurs, and Real Estate: An Anchoring-and-Adjustment Perspective on Property Pricing Decisions. Organizational Behavior and Human Decision Processes, 39(1), 84–97. https://doi.org/10.1016/0749-5978(87)90046-X

[9] Epley, N., & Gilovich, T. (2006). The Anchoring-and-Adjustment Heuristic: Why the Adjustments Are Insufficient. Psychological Science, 17(4), 311–318. https://doi.org/10.1111/j.1467-9280.2006.01704.x

[10] Galinsky, A. D., & Mussweiler, T. (2001). First Offers as Anchors: The Role of Perspective-Taking and Negotiator Focus. Journal of Personality and Social Psychology, 81(4), 657–669. https://doi.org/10.1037/0022-3514.81.4.657

[11] Ritov, I. (1996). Anchoring in Simulated Competitive Market Negotiation. Organizational Behavior and Human Decision Processes, 67(1), 16–25. https://doi.org/10.1006/obhd.1996.0062

[12] Skinner, D. J. (1994). Why Firms Voluntarily Disclose Bad News. Journal of Accounting Research, 32(1), 38–60. https://doi.org/10.2307/2491386

[13] Matsumoto, D. A. (2002). Management's Incentives to Avoid Negative Earnings Surprises. The Accounting Review, 77(3), 483–514. https://doi.org/10.2308/accr.2002.77.3.483

[14] Bartov, E., Givoly, D., & Hayn, C. (2002). The Rewards to Meeting or Beating Earnings Expectations. Journal of Accounting and Economics, 33(2), 173–204. https://doi.org/10.1016/S0165-4101(02)00045-9

[15] Healy, P. M., & Palepu, K. G. (2001). Information Asymmetry, Corporate Disclosure, and the Capital Markets: A Review of the Empirical Disclosure Literature. Journal of Accounting and Economics, 31(1–3), 405–440. https://doi.org/10.1016/S0165-4101(01)00018-0

[16] Kaplan, S. N., & Strömberg, P. (2003). Financial Contracting Theory Meets the Real World: An Empirical Analysis of Venture Capital Contracts. Review of Economic Studies, 70(2), 281–315. https://doi.org/10.1111/1467-937X.00245

[17] Gornall, W., & Strebulaev, I. A. (2020). Squaring Venture Capital Valuations with Reality. Journal of Financial Economics, 135(1), 120–143. https://doi.org/10.1016/j.jfineco.2018.04.015

Public template document (source of Table 2)

[18] National Venture Capital Association (2020). NVCA Model Term Sheet (Series A Preferred Stock Financing), July 2020 edition. https://nvca.org/wp-content/uploads/2020/07/NVCA-2020-Term-Sheet.docx (index page https://nvca.org/model-legal-documents/; accessed 2026-09-22; line numbers in Table 2 refer to the 124-line plain-text extract described in 4.1)

Verification note: [1]–[3] and [18] were fetched over HTTP on 2026-09-22 and found accessible; for [4]–[17], each DOI was resolved against doi.org and its title retrieved from Crossref, matching the entries above. Verification used the check() function in _仓储/_build/refcheck.py; item-by-item results are in Section 4 of the writing notes.

JEL Classification: D82 (Asymmetric and Private Information · Mechanism Design), D83 (Search · Learning · Information and Knowledge · Communication · Belief · Unawareness), D91 (Role and Effects of Psychological, Emotional, Social, and Cognitive Factors on Decision Making), G24 (Investment Banking · Venture Capital · Brokerage · Ratings and Ratings Agencies), G32 (Financing Policy · Financial Risk and Risk Management · Capital and Ownership Structure · Value of Firms · Goodwill)

Appendix: Source Passages Quoted from the Glacier Institute Website

All website quotations are from [1] https://glacier.mba/institute.html (fetched 2026-09-22; plain text after stripping comments, scripts, styles and tags, 6,184 lines) and [2] https://glacier.mba/research/GI-WP-2026-P2.html. The Chinese originals are given in the body, each followed by an English rendering in italics; the renderings are this paper's and are not the Institute's official English text. Line numbers are listed item by item in Section 1 of the writing notes. The passages are:

• The eight-milestone statement (「D01 事实底座搭建 → … → D60 资金交割收尾」)
• "Translation" (「第一步不是开口，是先弄明白对方要什么」「嘴上说的和心里要的常常不是一回事」「硬度可以留在判断里，但是不能留在报价里」「管理预期不是让步」「事实底座还原是第一件事」「我们不强扭共识，只收双方已经熟透的那颗果子」)
• "The four-layer vortex" (「标准是自洽，不是齐全」「同一组事实分窗口同步给不同相关方」)
• "Classify first, then price" (「答非所问，比答错更贵」「为难之处在于，双方都以为在认真对话」「所以硬拉估值意义不大」「这张牌不争，估值的锚就被别人先钉住了」)
• "Finding the key" (「像企业考古学家」「先在市场上校准，再持续修正齿形」)
• "Low posture" (「上来把自己对标成天花板，对方随即按天花板的尺子来量你」)
• "Five pre-launch checks" (「四查同舟」「同舟不能有人中途想回去」)
• "Full stewardship" (「要不要接一份条款清单（term sheet）、要不要调整预期、要不要换掉整条航道，最后拍板的永远是企业」)
• Working Paper GI-WP-2026-P2 (「校准要在发问之前做，不是在被拒之后做——发出去的预期改一次，要付一次解释的钱」)

---

Glacier Institute 庚辛研究院. Calibration: aligning expectations before the question is asked[R]. Glacier Institute Working Paper GI-WP-2026-P9-EN. 2026.

Glacier Institute. (2026). Calibration: aligning expectations before the question is asked (Glacier Institute Working Paper No. GI-WP-2026-P9-EN). https://glacier.mba/research/GI-WP-2026-P9-EN.html

@techreport{glacier2026p9en,
  author      = {{Glacier Institute}},
  title       = {{Calibration: aligning expectations before the question is asked}},
  institution = {Glacier Institute},
  type        = {Working Paper},
  number      = {GI-WP-2026-P9-EN},
  year        = {2026},
  url         = {https://glacier.mba/research/GI-WP-2026-P9-EN.html}
}