# Strategic Ambiguity in Entrepreneurial Promises In 2008, Tesla entered the electric vehicle market with a promise of "over 200 miles of range." This statement was not a precise number, but a lower bound, leaving significant room for final specifications. The firm ultimately delivered 244 miles for its initial Model S, leveraging this flexibility to iterate on battery technology while having already secured market interest. This example raises a fundamental question in entrepreneurship: What is the optimal variance of an entrepreneur's promise? It seems natural to think that an entrepreneur's primary communication goal should be maximum clarity (minimum variance), largely to solve the credibility problem: should a stakeholder believe the promise being made? From this perspective, a precise promise appears more credible. It avoids adding noise to an already challenging environment, signals the founder's clarity of thought, and demonstrates technical competence, particularly to those with an engineering mindset. By offering a concrete, verifiable target, a precise promise reduces the perceived risk for investors and customers, making it a "promise you can believe in." Yet, entrepreneurs systematically employ ambiguity in their communications. Recent empirical work suggests this is not accidental, but strategic. For instance, El-Zayaty et al. (2025) find that vague language can enhance resource acquisition when paired with strong founder credentials. This creates a puzzle: if precision is the answer to the credibility problem, why is ambiguity so prevalent and apparently effective? Existing theory provides limited guidance on when, why, and how much ambiguity is optimal. This paper develops a formal model to resolve this tension by analyzing the trade-off between mobilizing support with precise commitments and preserving adaptability through strategic ambiguity. We analyze a situation where an entrepreneur must make a public promise about a key quality attributeโ€”like battery rangeโ€”to a market of potential resource providers (e.g., customers). We model this promise as a strategic communication with two components: an expected quality level (mean, $\mu$) and a degree of variance ($\sigma^2$). We conceptualize this promise as a social prior: a publicly articulated distribution that serves as a focal point for coordination among all stakeholders, including the entrepreneur. Unlike a classical Bayesian prior that represents an agent's static, internal state of knowledge, this social prior is a strategically chosen variable, constructed to optimize the venture's success in the face of shared uncertainty. The promise is thus a tool for both persuasion and coordination. The core contribution of our paper is to endogenize the choice of promise variance ($\sigma^2$) by formalizing the trade-off between mobilization and adaptability. A precise promise (low variance) is beneficial for mobilization; it reduces perceived risk for stakeholders and makes the venture's claims appear more credible, thereby increasing the likelihood of securing resources. However, precision creates rigidity. It acts as a commitment that is costly to break. This tension aligns with prior work showing that while pivots are common, they can alienate stakeholders who bought into the venture's original claims, complicating the reorientation process (McDonald & Gao, 2019). If the entrepreneur discovers a better path or finds the initial goal is suboptimal, a precise prior promise generates significant rigidity costsโ€”reputational damage, loss of stakeholder trust, or the cost of being locked into an inferior design. An ambiguous promise (high variance) minimizes these rigidity costs, preserving the adaptability needed for learning and pivots, but it hinders initial resource mobilization. Our analysis demonstrates that the optimal level of promise variance is governed by the ratio of the venture's potential value to these endogenous rigidity costs. We derive a clear threshold rule: when the costs of being wrong are high relative to the value of securing initial support, an entrepreneur should remain deliberately ambiguous (maintain high variance). Precision must be "earned" as the venture de-risks and the value of commitment begins to outweigh the value of flexibility. This formalizes the intuition that ventures should transition from an exploratory, high-variance stance to a more defined, low-variance one as they mature. Our work is most closely related to two papers. Archibald et al. (2002) model startup survival by optimizing inventory decisions under demand uncertainty, but in their model, this uncertainty is an exogenous state of the world, not a chosen communication strategy. Closer to our work, El-Zayaty et al. (2025) provide empirical evidence that strategic ambiguity can succeed. Our model provides the theoretical counterpart: it prescribes when and how much variance should be used by framing it as a solution to an optimization problem. Our primary departure is to explicitly model promise variance as a strategic communication variable, distinct from an entrepreneur's private knowledge, to manage the fundamental tension between commitment and flexibility. The paper proceeds as follows. Section 2 develops the model, formalizing the costs and benefits of promise variance. Section 3 presents the main results, including the derivation of the optimal variance level. Section 4 discusses the implications for entrepreneurial strategy and communication. Section 5 concludes. ### References Archibald, G., Thomas, L. C., Betts, J. M., & Johnston, R. B. (2002). The Management of Start-Up Companies in the Service Sector. Management Science, 48(3), 357โ€“367. El-Zayaty, A., Ganco, M., & Khoshimov, B. (2025). Vague Language, Founding Team Human Capital, and Resource Acquisition. Organization Science. Forthcoming. McDonald, R., & Gao, C. (2019). Pivoting Isnโ€™t Enough? Managing Strategic Reorientation in New Ventures. Organization Science, 30(6), 1289โ€“1318. ---- # Introduction: Strategic Ambiguity in Entrepreneurial Promises --- ## Paragraph 1: Opening Hook (Concrete Example) 2013๋…„, Tesla๋Š” ์ „๊ธฐ์ฐจ ์‹œ์žฅ์— ์ง„์ž…ํ•˜๋ฉฐ "200๋งˆ์ผ ์ด์ƒ์˜ ์ฃผํ–‰๊ฑฐ๋ฆฌ"๋ฅผ ์•ฝ์†ํ–ˆ๋‹ค. ์ด ์•ฝ์†์€ ์˜๋„์ ์œผ๋กœ ๋ชจํ˜ธํ–ˆ๋‹คโ€”์ •ํ™•ํžˆ 200๋งˆ์ผ์ธ์ง€, 250๋งˆ์ผ์ธ์ง€, 300๋งˆ์ผ์ธ์ง€ ๋ช…์‹œํ•˜์ง€ ์•Š์•˜๋‹ค. ๊ฒฐ๊ณผ๋Š” ์–ด๋– ํ–ˆ๋Š”๊ฐ€? Tesla๋Š” ์ด ์œ ์—ฐํ•œ ์•ฝ์† ์•„๋ž˜์—์„œ ๋ฐฐํ„ฐ๋ฆฌ ๊ธฐ์ˆ ์„ ๋ฐ˜๋ณต์ ์œผ๋กœ ๊ฐœ์„ ํ–ˆ๊ณ , ์ตœ์ข…์ ์œผ๋กœ 244๋งˆ์ผ์„ ๋‹ฌ์„ฑํ•˜๋ฉฐ ์‹œ์žฅ์„ ์žฅ์•…ํ–ˆ๋‹ค. ์ด ํ•˜๋‚˜์˜ ์‚ฌ๋ก€๋Š” ์ฐฝ์—…๊ฐ€์˜ ์•ฝ์†(promise)์— ์žˆ์–ด ๊ทผ๋ณธ์ ์ธ ์งˆ๋ฌธ์„ ์ œ๊ธฐํ•œ๋‹ค: **์ฐฝ์—…๊ฐ€๋Š” ์ž์‹ ์˜ ์•ฝ์†์—์„œ ์–ผ๋งˆ๋‚˜ ์ •๋ฐ€ํ•ด์•ผ ํ•˜๋Š”๊ฐ€? ํ˜น์€ ๋‹ฌ๋ฆฌ ๋งํ•˜๋ฉด, ์•ฝ์†์˜ ๋ถ„์‚ฐ(variance)์„ ์–ผ๋งˆ๋‚˜ ํฌ๊ฒŒ ๊ฐ€์ ธ๊ฐ€์•ผ ํ•˜๋Š”๊ฐ€?** --- ## Paragraph 2: The Natural Null ("They Say") **์ƒ์‹์ ์œผ๋กœ๋Š” ์ฐฝ์—…๊ฐ€๊ฐ€ ๊ฐ€๋Šฅํ•œ ํ•œ ๋ช…ํ™•ํ•˜๊ณ  ์ •๋ฐ€ํ•˜๊ฒŒ ์†Œํ†ตํ•˜๋Š” ๊ฒƒ์ด ๊ฐ€์น˜ ์žˆ๋‹ค๊ณ  ์ƒ๊ฐํ•˜๋Š” ๊ฒƒ ๊ฐ™๋‹ค(It seems natural to think).** ์ •๋ฐ€์„ฑ์€ ์ด๋ฏธ ๋ถˆํ™•์‹คํ•œ ์„ธ๊ณ„์— ์žก์Œ์„ ์ถ”๊ฐ€ํ•˜์ง€ ์•Š์œผ๋ฉฐ, ์‚ฌ๊ณ ์˜ ๋ช…ํ™•์„ฑ์„ ์‹ ํ˜ธํ•˜๊ณ , ๊ธฐ์ˆ ์  ์—ญ๋Ÿ‰์„ ์ž…์ฆํ•˜๋Š” ๊ฒƒ์ฒ˜๋Ÿผ ๋ณด์ธ๋‹ค. ํˆฌ์ž์ž๋“ค์€ ๊ตฌ์ฒด์ ์ธ ์ˆ˜์น˜๋ฅผ ์„ ํ˜ธํ•˜๊ณ , ๊ณ ๊ฐ๋“ค์€ ๋ช…ํ™•ํ•œ ์‚ฌ์–‘์„ ์›ํ•œ๋‹ค๊ณ  ๊ฐ€์ •๋œ๋‹ค. **๊ทธ๋Ÿฌ๋‚˜ ์ฐฝ์—…๊ฐ€๋“ค์€ ์ฒด๊ณ„์ ์œผ๋กœ ์ž์‹ ์˜ ์•ฝ์†์— ๋ชจํ˜ธ์„ฑ์„ ์‚ฌ์šฉํ•˜๋ฉฐ, ์ตœ๊ทผ ์‹ค์ฆ ์—ฐ๊ตฌ๋“ค์€ ์ด๊ฒƒ์ด ์šฐ์—ฐ์ด ์•„๋‹ˆ๋ผ ์ „๋žต์ ์ผ ์ˆ˜ ์žˆ์Œ์„ ์‹œ์‚ฌํ•œ๋‹ค.** ์‹ค์ œ๋กœ, ์ ์ ˆํ•œ ์ธ์  ์ž๋ณธ ์‹ ํ˜ธ์™€ ๊ฒฐํ•ฉ๋  ๋•Œ ํŒ€ ์ปค๋ฎค๋‹ˆ์ผ€์ด์…˜์˜ ๋ชจํ˜ธํ•œ ์–ธ์–ด๊ฐ€ ์ž์› ํ™•๋ณด๋ฅผ ํ–ฅ์ƒ์‹œํ‚ฌ ์ˆ˜ ์žˆ๋‹ค๋Š” ๊ฒƒ์ด ๋ฐํ˜€์กŒ๋‹ค(El-Zayaty et al., 2025). ์„ฑ์žฅ ์•ผ๋ง์— ๋Œ€ํ•œ ์ง€๋‚˜์น˜๊ฒŒ ์ •๋ฐ€ํ•œ ์˜ˆ์ธก์€ ์˜คํžˆ๋ ค "cheap talk"๋กœ ๊ฐ„์ฃผ๋˜์–ด ์—ญํšจ๊ณผ๋ฅผ ๋‚ผ ์ˆ˜ ์žˆ๋‹ค(Kleinert, 2024). **ํ•˜์ง€๋งŒ ์ฐฝ์—…๊ฐ€๊ฐ€ ์–ธ์ œ, ์™œ, ๊ทธ๋ฆฌ๊ณ  ์•ฝ์†์˜ ๋ถ„์‚ฐ์„ ์–ผ๋งˆ๋‚˜ ํฌ๊ฒŒ ์„ค์ •ํ•ด์•ผ ํ•˜๋Š”์ง€์— ๋Œ€ํ•œ ์ด๋ก ์  ๊ธฐ๋ฐ˜์€ ์•„์ง ๋ฏธ๋ฐœ๋‹ฌ ์ƒํƒœ์ด๋‹ค.** --- ## Paragraph 3: The Game Structure (Plain Language Setup) ๋ณธ ๋…ผ๋ฌธ์€ ์ด ํ•ต์‹ฌ ๊ธด์žฅ์„ ๋“œ๋Ÿฌ๋‚ด๊ธฐ ์œ„ํ•ด ์˜๋„์ ์œผ๋กœ ๋‹จ์ˆœํ™”๋œ ํ˜•์‹ ๋ชจ๋ธ์„ ๊ฐœ๋ฐœํ•œ๋‹ค. ์šฐ๋ฆฌ๊ฐ€ ๋ถ„์„ํ•˜๋Š” ๊ฒŒ์ž„์€ ๋‹ค์Œ๊ณผ ๊ฐ™๋‹ค. ํ•œ ์ฐฝ์—…๊ฐ€๊ฐ€ ์ œํ’ˆ ์•„์ด๋””์–ด๋ฅผ ๊ฐ€์ง€๊ณ  ์žˆ์œผ๋ฉฐ, ์ž ์žฌ ๊ณ ๊ฐ(๋˜๋Š” ๋™๋“ฑํ•˜๊ฒŒ, ์ž๋ณธ ์ œ๊ณต์ž)์—๊ฒŒ ํ’ˆ์งˆ ์†์„ฑ์— ๋Œ€ํ•œ ์•ฝ์†์„ ํ•ด์•ผ ํ•œ๋‹คโ€”Tesla๊ฐ€ ํŠน์ • ๋ฐฐํ„ฐ๋ฆฌ ์ฃผํ–‰๊ฑฐ๋ฆฌ๋ฅผ ์•ฝ์†ํ•˜๋Š” ๊ฒƒ์„ ์ƒ๊ฐํ•ด๋ณด๋ผ. ์ด ์•ฝ์†์€ ๋‘ ๊ฐ€์ง€ ๊ตฌ์„ฑ์š”์†Œ๋ฅผ ๊ฐ€์ง„๋‹ค: **ํ’ˆ์งˆ์˜ ํ‰๊ท  ์ˆ˜์ค€(mean, $\mu$)**โ€”๋†’์€ ์•ฝ์†์€ ๋” ๋งŽ์€ ๊ณ ๊ฐ์„ ๋Œ์–ด๋“ค์ธ๋‹คโ€”๊ณผ **๊ทธ ํ‰๊ท  ์ฃผ๋ณ€์˜ ๋ถ„์‚ฐ(variance, $\sigma^2$ )**โ€”๋” ์ •๋ฐ€ํ•œ ์•ฝ์†(๋‚ฎ์€ ๋ถ„์‚ฐ)์€ ๋” ๊ตฌ์ฒด์ ์ด์ง€๋งŒ ๋” ์ œ์•ฝ์ ์ด๋‹ค. ์ฐฝ์—…๊ฐ€๋Š” ๋” ๋†’์€ ํ’ˆ์งˆ์˜ ์•ฝ์†์ด ์‹คํ˜„ํ•˜๊ธฐ ๋” ์–ด๋ ต๋‹ค๋Š” ๊ฒƒ์„ ์•Œ๊ณ  ์žˆ์œผ๋ฉฐ, ์ด๊ฒƒ์ด ํ‰๊ท ์—์„œ ๋ช…๋ฐฑํ•œ ์ƒ์ถฉ๊ด€๊ณ„๋ฅผ ๋งŒ๋“ ๋‹ค: **๋„ˆ๋ฌด ๋งŽ์ด ์•ฝ์†ํ•˜๋ฉด ์‹คํ˜„์— ์‹คํŒจํ•˜๊ณ (deliverability ์ €ํ•˜), ๋„ˆ๋ฌด ์ ๊ฒŒ ์•ฝ์†ํ•˜๋ฉด ์ž์›์„ ๋Œ์–ด๋“ค์ด๋Š” ๋ฐ ์‹คํŒจํ•œ๋‹ค(sellability ์ €ํ•˜).** ์„ฑ๊ณต์€ ์•ฝ์†์„ ํŒ”๊ธฐ(์ž์› ๋™์›, mobilizing support)์™€ ๊ทธ๊ฒƒ์„ ์‹คํ˜„ํ•˜๊ธฐ(์ ์‘์„ฑ ๋ณด์กด, preserving adaptability) ๋‘˜ ๋‹ค๋ฅผ ํ•„์š”๋กœ ํ•œ๋‹ค. --- ## Paragraph 4: Mean is Obvious, Variance is Not ("I Say" - Core Contribution) **ํ‰๊ท ์ด ์–ด๋–ป๊ฒŒ ์ž‘๋™ํ•˜๋Š”์ง€๋Š” ์ž๋ช…ํ•˜๋‹คโ€”๋ˆ„๊ตฌ๋‚˜ ์–ด๋–ค ์ตœ์ ์˜ ์ค‘๊ฐ„ ์ง€์ ์ด ์žˆ์–ด์•ผ ํ•œ๋‹ค๋Š” ๊ฒƒ์„ ์•Œ ์ˆ˜ ์žˆ๋‹ค.** ๋” ๋†’์€ ํ‰๊ท ์€ ๋” ๋งŽ์€ ๊ณ ๊ฐ์„ ๋Œ์ง€๋งŒ ๋” ์–ด๋ ต๊ฒŒ ๋งŒ๋“ค๊ณ , ๋” ๋‚ฎ์€ ํ‰๊ท ์€ ๋” ์‰ฝ์ง€๋งŒ ๋œ ๋งค๋ ฅ์ ์ด๋‹ค. **๊ทธ๋Ÿฌ๋‚˜ ์ด ์„ค์ •์—์„œ ์ž๋ช…ํ•˜์ง€ ์•Š์€ ๊ฒƒ์€ ์•ฝ์†์˜ ์ตœ์  ๋ถ„์‚ฐ(variance) ๋˜๋Š” ๋ชจํ˜ธ์„ฑ(ambiguity)์ด ๋ฌด์—‡์ด์–ด์•ผ ํ•˜๋Š”๊ฐ€์ด๋‹ค.** ์ด๊ฒƒ์ด ์šฐ๋ฆฌ๊ฐ€ ๋‹ค๋ฃจ๋Š” ์ค‘์‹ฌ ์งˆ๋ฌธ์ด๋‹ค. ๋ชจ๋ธ์—์„œ ์•ฝ์†์€ ํ‰๊ท ($\mu$)๊ณผ ๋ถ„์‚ฐ($\sigma^2$)์„ ๋ชจ๋‘ ๊ฐ€์ง„๋‹ค. **์šฐ๋ฆฌ์˜ ํ•ต์‹ฌ ๊ธฐ์—ฌ๋Š” ์ด ๋ถ„์‚ฐ์„ ๋Šฅ๋™์ ์œผ๋กœ ์„ ํƒ๋œ ์ „๋žต ๋ณ€์ˆ˜๋กœ ๋‹ค๋ฃฌ๋‹ค๋Š” ๊ฒƒ์ด๋‹ค.** ์ „ํ†ต์ ์ธ ๋ฒ ์ด์ง€์•ˆ ํ”„๋ ˆ์ž„์›Œํฌ๋Š” ์‚ฌ์ „ ๋ถ„ํฌ(prior)๋ฅผ ๋ฐ์ดํ„ฐ ๊ด€์ฐฐ ์ „์˜ ๋ฏฟ์Œ์œผ๋กœ ์ทจ๊ธ‰ํ•˜์ง€๋งŒ, ์šฐ๋ฆฌ๋Š” ์ฐฝ์—…๊ฐ€๊ฐ€ ์ด ์‚ฌ์ „ ์ž์ฒด์˜ ๋ถ„์‚ฐ์„ ์ „๋žต์ ์œผ๋กœ ์„ค๊ณ„ํ•œ๋‹ค๊ณ  ๋ณธ๋‹ค. ์šฐ๋ฆฌ๋Š” ์ •๋ฐ€์„ฑ์ด ์กฐ์ง์  ๋น„์šฉ์„ ์ˆ˜๋ฐ˜ํ•œ๋‹ค๋Š” ๊ฒƒ์„ ๋ณด์ธ๋‹ค: **๋ชจํ˜ธํ•œ ์•ฝ์†์„ ์œ ์ง€ํ•˜๋Š” ๊ฒƒ์€ ์‹ ๋ขฐ์„ฑ ๊ตฌ์ถ•, ๋ช…ํ™•์„ฑ ์œ ์ง€, ์ œ์•ฝ ๊ด€๋ฆฌ ์‹œ์Šคํ…œ์— ์ง€์†์ ์ธ ํˆฌ์ž๋ฅผ ํ•„์š”๋กœ ํ•œ๋‹ค.** ์ด๋Ÿฌํ•œ ๋น„์šฉ($i$๋กœ ํ‘œ๊ธฐ, ์ •๋ณด ํ†ตํ•ฉ ๋น„์šฉ)์€ ์ •๋ฐ€ํ•œ ์•ฝ์†์„ ๊ฒ€์ฆํ•˜๊ณ , ์†Œํ†ตํ•˜๊ณ , ๊ถ๊ทน์ ์œผ๋กœ ์ดํ–‰ํ•˜๋Š” ๋ฐ ํ•„์š”ํ•œ ์ž์›์„ ํฌ์ฐฉํ•œ๋‹ค. **์ฐฝ์—…๊ฐ€์˜ ๋ฌธ์ œ๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์ด ๋œ๋‹ค: ์„ฑ๊ณต์ ์ธ ์‹คํ˜„์œผ๋กœ๋ถ€ํ„ฐ์˜ ๊ธฐ๋Œ€ ๊ฐ€์น˜($V$)๋ฅผ ์ตœ๋Œ€ํ™”ํ•˜๋˜, ์•ฝ์† ์ •๋ฐ€์„ฑ์„ ์œ ์ง€ํ•˜๋Š” ์กฐ์ง์  ๋น„์šฉ์„ ์ฐจ๊ฐํ•œ๋‹ค.** ์šฐ๋ฆฌ์˜ ์ฃผ์š” ๊ฒฐ๊ณผ์ธ **"์ตœ์  ๋ฌด์ง€ ๊ทœ์น™(optimal-ignorance rule)"**์€ ์ด์‚ฐ์  ์ž„๊ณ„๊ฐ’ ๊ตฌ์กฐ๋ฅผ ๋“œ๋Ÿฌ๋‚ธ๋‹ค: ๋ฒค์ฒ˜ ๊ฐ€์น˜๊ฐ€ ํ†ตํ•ฉ ๋น„์šฉ์— ๋น„ํ•ด ์ถฉ๋ถ„ํžˆ ๋†’์ง€ ์•Š์„ ๋•Œ($V \leq 4i$, ๋‹จ์ˆœํ™”๋œ ๊ฒฝ์šฐ), ์ฐฝ์—…๊ฐ€๋Š” ์™„์ „ํ•œ ๋ชจํ˜ธ์„ฑ($\tau^* = 0$, ์ฆ‰ ๋ฌดํ•œ ๋ถ„์‚ฐ)์„ ์œ ์ง€ํ•ด์•ผ ํ•˜๋ฉฐ, ์ด๋Š” ์ตœ๋Œ€ ์œ ์—ฐ์„ฑ์„ ๋ณด์กดํ•˜๋Š” Haldane ์‚ฌ์ „(prior)์— ํ•ด๋‹นํ•œ๋‹ค. ๊ฐ€์น˜๊ฐ€ ์ด ์ž„๊ณ„๊ฐ’์„ ์ดˆ๊ณผํ•  ๋•Œ๋งŒ ์ •๋ฐ€์„ฑ์ด ์ƒ์Šนํ•ด์•ผ ํ•˜๋ฉฐ(๋ถ„์‚ฐ ๊ฐ์†Œ), ๊ทธ๋•Œ์กฐ์ฐจ๋„ ๊ฐ€์น˜ ๋Œ€ ๋น„์šฉ ๋น„์œจ์˜ ์ œ๊ณฑ๊ทผ์œผ๋กœ๋งŒ ์ฆ๊ฐ€ํ•˜์—ฌ ๊ฐ•ํ•œ ์ˆ˜ํ™•์ฒด๊ฐ์„ ์•”์‹œํ•œ๋‹ค. **์ด๊ฒƒ์€ ๋ฒค์ฒ˜๊ฐ€ ๊ฒ€์ฆ๋œ ์„ฑ์žฅ์„ ํ†ตํ•ด ์ •๋ฐ€ํ•ด์งˆ ๊ถŒ๋ฆฌ๋ฅผ ํš๋“ํ•ด์•ผ ํ•œ๋‹ค๋Š” ์ง๊ด€์„ ํ˜•์‹ํ™”ํ•˜๋ฉฐ, ํƒ์ƒ‰์  ๋ชจํ˜ธ์„ฑ(๋†’์€ ๋ถ„์‚ฐ)์—์„œ ์กฐ์ •๋œ ์•ฝ์†(๋‚ฎ์€ ๋ถ„์‚ฐ)์œผ๋กœ ์ „ํ™˜ํ•œ๋‹ค.** --- ## Paragraph 5: Positioning - Two Most Closely Related Papers **์šฐ๋ฆฌ์˜ ์—ฐ๊ตฌ๋Š” ๋‘ ๊ฐœ์˜ ๋…ผ๋ฌธ๊ณผ ๊ฐ€์žฅ ๋ฐ€์ ‘ํ•˜๊ฒŒ ๊ด€๋ จ๋˜์–ด ์žˆ๋‹ค(Our work is most closely related to two papers).** ์ฒซ์งธ, **Archibald, Thomas, Betts, and Johnston (2002)์€ Management Science์—์„œ ์Šคํƒ€ํŠธ์—… ๊ธฐ์—…์ด ์ƒ์กด ํ™•๋ฅ ์„ ์ตœ๋Œ€ํ™”ํ•˜๋Š” ์žฌ๊ณ  ์ •์ฑ…์„ ๋ถ„์„ํ–ˆ๋‹ค.** ๊ทธ๋“ค์˜ ํ•ต์‹ฌ ํ†ต์ฐฐ์€ ์‹ ์ƒ ๊ธฐ์—…์€ ํ‰๊ท  ์ˆ˜์ต์„ ์ตœ๋Œ€ํ™”ํ•˜๋Š” ์„ฑ์ˆ™ ๊ธฐ์—…๊ณผ ๋‹ฌ๋ฆฌ **์ƒ์กด ํ™•๋ฅ  ์ž์ฒด๋ฅผ ๋ชฉ์  ํ•จ์ˆ˜๋กœ ์‚ผ์•„์•ผ ํ•œ๋‹ค**๋Š” ๊ฒƒ์ด๋‹ค. **์šฐ๋ฆฌ๋Š” ์ด ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ๊ณต์œ ํ•˜์ง€๋งŒ, ๊ทธ๋“ค์˜ ์ดˆ์ ์ด ์žฌ๊ณ  ์ฃผ๋ฌธ๋Ÿ‰(inventory decision)์ธ ๋ฐ˜๋ฉด, ์šฐ๋ฆฌ์˜ ์ดˆ์ ์€ ์•ฝ์†์˜ ๋ถ„์‚ฐ(promise variance)์ด๋‹ค.** ๋” ์ค‘์š”ํ•˜๊ฒŒ, ๊ทธ๋“ค์˜ ๋ชจ๋ธ์—์„œ ๋ถˆํ™•์‹ค์„ฑ(์ˆ˜์š”์˜ ๋ถ„์‚ฐ)์€ ์™ธ์ƒ์ ์œผ๋กœ ์ฃผ์–ด์ง„ ๋ฐ˜๋ฉด, ์šฐ๋ฆฌ ๋ชจ๋ธ์—์„œ ๋ถˆํ™•์‹ค์„ฑ(์•ฝ์†์˜ ๋ถ„์‚ฐ)์€ ์ฐฝ์—…๊ฐ€๊ฐ€ ์„ค๊ณ„ํ•˜๋Š” ๋‚ด์ƒ์  ์ „๋žต ๋ณ€์ˆ˜์ด๋‹ค. ๋‘˜์งธ, **El-Zayaty, Ganco, and Khoshimov (2025)๋Š” Organization Science์—์„œ ์ฐฝ์—… ํŒ€์˜ ๋ชจํ˜ธํ•œ ์–ธ์–ด๊ฐ€ ์ž์› ํ™•๋ณด์— ์ „๋žต์ ์œผ๋กœ ์œ ์ตํ•  ์ˆ˜ ์žˆ์Œ์„ ๋ณด์—ฌ์ฃผ๋Š” ์ตœ์ดˆ์˜ ๋Œ€๊ทœ๋ชจ ์‹ค์ฆ ์ฆ๊ฑฐ๋ฅผ ์ œ๊ณตํ–ˆ๋‹ค**, ํŠนํžˆ ์ฐฝ์—…์ž๊ฐ€ ๊ฐ•ํ•œ ์ธ์  ์ž๋ณธ ์‹ ํ˜ธ๋ฅผ ๋ณด์œ ํ•  ๋•Œ ๊ทธ๋Ÿฌํ•˜๋‹ค. **๊ทธ๋“ค์˜ ์—ฐ๊ตฌ๋Š” ์ „๋žต์  ๋ชจํ˜ธ์„ฑ์ด ์„ฑ๊ณตํ•  ์ˆ˜ *์žˆ๋‹ค*๋Š” ๊ฒƒ์„ ๋ณด์—ฌ์ฃผ์ง€๋งŒ(They show *that* strategic vagueness can work), *์–ธ์ œ* ๊ทธ๋ฆฌ๊ณ  ์•ฝ์†์˜ ๋ถ„์‚ฐ์„ *์–ผ๋งˆ๋‚˜* ํฌ๊ฒŒ ๊ฐ€์ ธ๊ฐ€์•ผ ํ•˜๋Š”์ง€๋ฅผ ํ˜•์‹์ ์œผ๋กœ ๋ชจ๋ธ๋งํ•˜์ง€๋Š” ์•Š๋Š”๋‹ค(but do not model *when* and *how much* variance is optimal).** ๊ทธ๋“ค์˜ ์ž‘์—…์€ ์–ธ์–ด์  ๋ชจํ˜ธ์„ฑ(linguistic vagueness)์˜ ์กด์žฌ์™€ ํšจ๊ณผ๋ฅผ ๊ธฐ์ˆ ํ•˜๋Š” ๋ฐ˜๋ฉด, ์šฐ๋ฆฌ์˜ ์ž‘์—…์€ ์ˆ˜์น˜์  ๋ถ„์‚ฐ(numerical variance)์˜ ์ตœ์  ์ˆ˜์ค€์„ ์ฒ˜๋ฐฉํ•œ๋‹ค. --- ## Paragraph 6: Three Key Differences ("Our Contribution Differs") **์šฐ๋ฆฌ์˜ ๊ธฐ์—ฌ๋Š” ์„ธ ๊ฐ€์ง€ ํ•ต์‹ฌ์ ์ธ ๋ฐฉ์‹์—์„œ ๋‹ค๋ฅด๋‹ค(Our contribution differs in three key ways).** ์ฒซ์งธ, ์šฐ๋ฆฌ๋Š” **๋ถ„์‚ฐ์„ ์ „๋žต์  ์„ ํƒ์œผ๋กœ ๋‚ด์ƒํ™”ํ•˜๋Š” ํ˜•์‹ ์ตœ์ ํ™” ๋ชจ๋ธ**์„ ๊ฐœ๋ฐœํ•˜์—ฌ, ๋†’์€ ๋ถ„์‚ฐ(๋ชจํ˜ธ์„ฑ)์ด ์ง€๋ฐฐํ•˜๋Š” ์กฐ๊ฑด์„ ๋„์ถœํ•œ๋‹ค. Archibald et al.์€ ์ƒ์กด ํ™•๋ฅ ์„ ๋ชฉ์ ์œผ๋กœ ์‚ผ์•˜์ง€๋งŒ ๋ถ„์‚ฐ์„ ์„ค๊ณ„ํ•˜์ง€ ์•Š์•˜๊ณ , El-Zayaty et al.์€ ๋ชจํ˜ธ์„ฑ์˜ ํšจ๊ณผ๋ฅผ ์ธก์ •ํ–ˆ์ง€๋งŒ ์ตœ์ ํ™”ํ•˜์ง€ ์•Š์•˜๋‹ค. ๋‘˜์งธ, ์šฐ๋ฆฌ๋Š” ์•ฝ์†์˜ **ํ‰๊ท -๋ถ„์‚ฐ ๊ตฌ์กฐ**๋ฅผ ๋ช…์‹œ์ ์œผ๋กœ ๋ถ„๋ฆฌํ•˜์—ฌ, ์ตœ์  ๋ถ„์‚ฐ์ด ์ตœ์  ํ‰๊ท  ์ˆ˜์ค€์„ ์ง€๋ฐฐํ•˜๋Š” ์š”์ธ(๊ณ ๊ฐ ์ˆ˜์š”, ๊ธฐ์ˆ ์  ๋‚œ์ด๋„)๊ณผ ๊ตฌ๋ณ„๋˜๋Š” ์กฐ์ง์  ๋น„์šฉ ๋งค๊ฐœ๋ณ€์ˆ˜($i$)์— ์˜์กดํ•œ๋‹ค๋Š” ๊ฒƒ์„ ๋ณด์ธ๋‹ค. ์ด๋Š” "์–ผ๋งˆ๋‚˜ ๋†’๊ฒŒ ์•ฝ์†ํ• ๊นŒ"(ํ‰๊ท  ๋ฌธ์ œ)์™€ "์–ผ๋งˆ๋‚˜ ๊ตฌ์ฒด์ ์œผ๋กœ ์•ฝ์†ํ• ๊นŒ"(๋ถ„์‚ฐ ๋ฌธ์ œ)๋ฅผ ๋ถ„๋ฆฌํ•œ๋‹ค. ์…‹์งธ, ์šฐ๋ฆฌ๋Š” ๋ฒค์ฒ˜๊ฐ€ ์ตœ๋Œ€ ๋ถ„์‚ฐ(์™„์ „ ๋ชจํ˜ธ์„ฑ, $\sigma^* = \infty$)์—์„œ ์กฐ์ •๋œ ์ •๋ฐ€์„ฑ(๋‚ฎ์€ ๋ถ„์‚ฐ, $\sigma^* < \infty$)์œผ๋กœ **์–ธ์ œ ์ „ํ™˜ํ•ด์•ผ ํ•˜๋Š”์ง€**๋ฅผ ์šด์˜ํ™”ํ•˜๋Š” **์ž„๊ณ„๊ฐ’ ๊ทœ์น™**($V/i$ ๋น„์œจ)์„ ์ œ๊ณตํ•˜๋ฉฐ, ์ˆœ์ˆ˜ ์‹ค์ฆ ์—ฐ๊ตฌ์—์„œ๋Š” ๋ถ€์žฌํ•œ ์ฒ˜๋ฐฉ์  ์ง€์นจ์„ ์ œ๊ณตํ•œ๋‹ค. ์ด๋Š” "์ง€๊ธˆ ๋‹น์žฅ ๋ถ„์‚ฐ์„ ์ค„์—ฌ์•ผ ํ•˜๋Š”๊ฐ€, ์•„๋‹ˆ๋ฉด ๋” ๊ธฐ๋‹ค๋ ค์•ผ ํ•˜๋Š”๊ฐ€"๋ผ๋Š” ๊ฒฝ์˜์ž์˜ ์‹ค์ œ ์งˆ๋ฌธ์— ๋‹ตํ•œ๋‹ค. --- ## Paragraph 7: Broader Contribution and the Trade-off **๋” ๋„“๊ฒŒ๋Š”, ์šฐ๋ฆฌ์˜ ๋ชจ๋ธ์€ ์„ธ ๊ฐ€์ง€ ๋ฌธํ—Œ์— ๊ธฐ์—ฌํ•œ๋‹ค(More broadly, our model contributes to three literatures).** ์ „๋žต์  ์•ฝ์†(strategic commitment) ๋ฌธํ—Œ์€ ์˜ค๋žซ๋™์•ˆ ์•ฝ์†๊ณผ ์œ ์—ฐ์„ฑ ์‚ฌ์ด์˜ ์ƒ์ถฉ๊ด€๊ณ„๋ฅผ ์ธ์‹ํ•ด์™”์ง€๋งŒ, ์ผ๋ฐ˜์ ์œผ๋กœ ์ด๋ฅผ ์ด์ง„ ์„ ํƒ์ด๋‚˜ ์ปค๋ฎค๋‹ˆ์ผ€์ด์…˜ ๋ถ„์‚ฐ์˜ ๋ช…์‹œ์  ๋ชจ๋ธ๋ง ์—†๋Š” ์—ฐ์† ์ „๋žต ๋ณ€์ˆ˜๋กœ ๋‹ค๋ฃฌ๋‹ค(Ghemawat, 1991; Reuer & Tong, 2007). ๋ฒ ์ด์ง€์•ˆ ์ฐฝ์—…(Bayesian entrepreneurship) ๋ฌธํ—Œ์€ ํ•™์Šต๊ณผ ๋ฏฟ์Œ ๊ฐฑ์‹ ์„ ๊ฐ•์กฐํ•˜์ง€๋งŒ ์ผ๋ฐ˜์ ์œผ๋กœ ์‚ฌ์ „ ๋ถ„ํฌ(prior distribution)๋ฅผ ์ฃผ์–ด์ง„ ๊ฒƒ์œผ๋กœ ์ทจ๊ธ‰ํ•˜์ง€ ์„ค๊ณ„๋œ ๊ฒƒ์œผ๋กœ ๋‹ค๋ฃจ์ง€ ์•Š๋Š”๋‹ค(Alvarez & Parker, 2009; Camuffo et al., 2020). ์ž์› ๋™์› ๋ฌธํ—Œ์€ ์ฐฝ์—…๊ฐ€๊ฐ€ ๋‹ค์–‘ํ•œ ์‹ ํ˜ธ๋ฅผ ํ†ตํ•ด ์–ด๋–ป๊ฒŒ ์ž์›์„ ๋Œ์–ด๋“ค์ด๋Š”์ง€ ๋‹ค๋ฃจ์ง€๋งŒ ์•ฝ์† ๋งŒ๋“ค๊ธฐ์—์„œ ํ‰๊ท -๋ถ„์‚ฐ ์ƒ์ถฉ๊ด€๊ณ„๋ฅผ ํ˜•์‹ํ™”ํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋“œ๋ฌผ๋‹ค(Shane & Cable, 2002). **์‚ฌ์ „ ๋ถ„์‚ฐ์„ ๋ช…์‹œ์  ์กฐ์ง์  ๋น„์šฉ์ด ์žˆ๋Š” ์ „๋žต์  ์„ค๊ณ„ ๋ณ€์ˆ˜๋กœ ๋‹ค๋ฃธ์œผ๋กœ์จ, ์šฐ๋ฆฌ๋Š” ์ด๋Ÿฌํ•œ ๋ฌธํ—Œ๋“ค์„ ์—ฐ๊ฒฐํ•˜๊ณ  ๋ถˆํ™•์‹ค์„ฑ ํ•˜์—์„œ ์ฐฝ์—…๊ฐ€์  ์˜์‚ฌ๊ฒฐ์ •์„ ๋ถ„์„ํ•˜๊ธฐ ์œ„ํ•œ ๋„๊ตฌ ์„ธํŠธ๋ฅผ ํ™•์žฅํ•œ๋‹ค.** ํŠนํžˆ, ์šฐ๋ฆฌ์˜ ๋ชจ๋ธ์€ **์ง€์› ๋™์›(mobilizing support)๊ณผ ์ ์‘์„ฑ ๋ณด์กด(preserving adaptability) ์‚ฌ์ด์˜ ์ƒ์ถฉ๊ด€๊ณ„(the trade-off between mobilizing support and preserving adaptability)**๋ฅผ ์ •๋Ÿ‰ํ™”ํ•œ๋‹ค: ๋‚ฎ์€ ๋ถ„์‚ฐ์€ ํˆฌ์ž์ž์™€ ๊ณ ๊ฐ์œผ๋กœ๋ถ€ํ„ฐ ๋” ๋งŽ์€ ์ž์›์„ ๋Œ์–ด๋“ค์ด์ง€๋งŒ ํ”ผ๋ด‡ํ•  ๋Šฅ๋ ฅ์„ ๊ฐ์†Œ์‹œํ‚ค๋ฉฐ(adaptability โ†“); ๋†’์€ ๋ถ„์‚ฐ์€ ์œ ์—ฐ์„ฑ์„ ์œ ์ง€ํ•˜์ง€๋งŒ ์‹ ๋ขฐ์„ฑ ๋ถ€์กฑ (credibility) ๋น„์šฉ์„ ์น˜๋ฅธ๋‹ค(mobilization โ†“). **์–ธ์ œ ํ•œ์ชฝ ๋น„์šฉ์ด ๋‹ค๋ฅธ ์ชฝ์„ ๋Šฅ๊ฐ€ํ•˜๋Š”์ง€๋ฅผ ํŠน์ง•์ง“๋Š” ๊ฒƒ์ด ์šฐ๋ฆฌ ๋ถ„์„์˜ ํ•ต์‹ฌ์ด๋‹ค.** ์ด ์ƒ์ถฉ๊ด€๊ณ„๋Š” ์ฐฝ์—…๊ฐ€๊ฐ€ "์ž์›์„ ๋” ๋Œ์–ด๋“ค์ด๊ธฐ ์œ„ํ•ด ๋ถ„์‚ฐ์„ ์ค„์—ฌ์•ผ ํ•˜๋Š”๊ฐ€, ์•„๋‹ˆ๋ฉด ์ ์‘ ๋Šฅ๋ ฅ์„ ์œ ์ง€ํ•˜๊ธฐ ์œ„ํ•ด ๋ถ„์‚ฐ์„ ๋†’๊ฒŒ ์œ ์ง€ํ•ด์•ผ ํ•˜๋Š”๊ฐ€"๋ผ๋Š” ๊ทผ๋ณธ์  ๊ธด์žฅ์„ ํฌ์ฐฉํ•œ๋‹ค. --- ## Paragraph 8: Roadmap ๋…ผ๋ฌธ์€ ๋‹ค์Œ๊ณผ ๊ฐ™์ด ์ง„ํ–‰๋œ๋‹ค. ์„น์…˜ 2๋Š” ๋ชจ๋ธ์„ ์ ์ง„์ ์œผ๋กœ ๊ฐœ๋ฐœํ•˜์—ฌ, ๊ฐ ๊ตฌ์„ฑ์š”์†Œ(ํ‰๊ท  ์„ ํƒ, ๋ถ„์‚ฐ ์„ ํƒ, ๋น„์šฉ ๊ตฌ์กฐ)๊ฐ€ ์ตœ์  ๋ฌด์ง€ ๊ฒฐ๊ณผ์— ์–ด๋–ป๊ฒŒ ๊ธฐ์—ฌํ•˜๋Š”์ง€ ๋ณด์ธ๋‹ค. ์„น์…˜ 3์€ ๋น„๊ต ์ •ํƒœ๋ถ„์„์„ ์ œ๊ณตํ•˜๊ณ  Tesla ์‚ฌ๋ก€๋ฅผ ๋ชจ๋ธ์˜ ๋ Œ์ฆˆ๋ฅผ ํ†ตํ•ด ๋‹ค๋ฃฌ๋‹ค. ์„น์…˜ 4๋Š” ์ฐฝ์—… ์ „๋žต๊ณผ ์กฐ์ง ์„ค๊ณ„์— ๋Œ€ํ•œ ํ•จ์˜๋ฅผ ๋…ผ์˜ํ•œ๋‹ค. ์„น์…˜ 5๋Š” ๋ฏธ๋ž˜ ์—ฐ๊ตฌ ๋ฐฉํ–ฅ์œผ๋กœ ๊ฒฐ๋ก ์ง“๋Š”๋‹ค. --- ## Meta-Notes for Further Refinement ### They Say/I Say ์‚ฌ์šฉ ํ˜„ํ™ฉ: โœ… "It seems natural to think..." (Para 2) โœ… "However..." (Para 2) โœ… "Our work is most closely related to..." (Para 5) โœ… "They show that... but do not..." (Para 5) โœ… "Our contribution differs in three key ways" (Para 6) โœ… "More broadly..." (Para 7) ### ์ถ”๊ฐ€ ๊ฐ•ํ™” ํ•„์š” ์˜์—ญ: 1. **Para 2**: "Some might object that..." ์ถ”๊ฐ€ ๊ฐ€๋Šฅ (๋ฐ˜๋ก  ์˜ˆ์ƒ) 2. **Para 6**: ๊ฐ difference์—์„œ "While X does..., we..." ๊ตฌ์กฐ ๋” ๋ช…ํ™•ํžˆ 3. **Para 7**: "This matters because..." ์ถ”๊ฐ€๋กœ importance ๊ฐ•์กฐ ### Word Count: ~1,380 words (MS ์„œ๋ก  ํ‘œ์ค€ ๋ฒ”์œ„ ๋‚ด)