## Abstract
Entrepreneurs make consequential decisions under extreme uncertainty, yet traditional startup advice lacks personalization and actionability. This research introduces STRAP (Sequential Threshold Reduction for Actionable Perceptions), a decision framework built on a powerful insight: reducing uncertainty directly increases your probability of successâthese are mathematically equivalent goals. This fundamental principle reshapes how founders should approach venture-building, transforming it into a systematic cycle of perception (understanding stakeholder uncertainties), action (running targeted experiments), and confirmation (aligning stakeholder expectations). STRAP provides a practical rule for choosing which experiments to run next: always target the "bottleneck" uncertainty that offers the highest information gain per dollar spent. This approach reveals that coordinating stakeholder expectations breaks deadlocks, using resources efficiently matters more than raising large amounts, and successful ventures naturally evolve from unpredictable to predictable outcomes. Case study on comparing Segway's STRAP-guided (balanced uncertainty reduction across stakeholders) and its actual (over-investing in technology while neglecting market validation) operation, demonstrate that STRAP-guided ventures achieve better outcomes with fewer resources by systematically eliminating critical uncertainties in the optimal sequence.
cInstead of relying on a static map and a one-dimensional compass, entrepreneurs need the equivalent of a GPS that dynamically combines sensor and motion â in other words, a navigation system that adapts in real time to the founderâs perspective and surroundings. Bayesian Entrepreneurship is an attempt to orient the field around entrepreneurs' prior beliefs about their opportunity space and how those beliefs evolve over the course of their entrepreneurial journey. If we take a Bayesian perspective, then entrepreneurship is more about "establishing hypotheses rather than having the right "vision", and running the right experiments rather than writing one-hundred-page business plans" (Stern et al, 2025).
This thesis contribute to Bayesian entrepreneurship, by suggesting a blueprint for Bayesian decision making tool for entrepreneurs. After identifying there's no entrepreneurial decision making model that are personalized and realistically tractable I identify three core needs for developing such an adaptive tool: đ˝ď¸ **Perception** â the ability to clearly sense and interpret how different stakeholders view the venture; ⥠**Sequencing** â the foresight to take the right steps in the right order under resource constraints; and đ **Confirmation** â the capacity to align and synergize these stakeholdersâ actions and expectations. These three needs form the backbone of my framework. In the following **Need Analysis** and **Solution Design** sections, I delve into each need and show how a rigorous decision model can function as that adaptive entrepreneurial GPS, I call this đŞ˘STRAP (Sequential Threshold Reduction for Actionable Perceptions) framework.
To motivate the economic effect of my work, I apply this framework to the mobility sector. Mobility ventures require complex, multi-stakeholder alignment across regulators, infrastructure partners, and technology providersâmaking them ideal for testing Confirmation strategies under real-world constraints. Their moderate clockspeedâslower than software but faster than heavy industryâgrants entrepreneurs time to strategically sequence experiments without being overtaken by market shifts. The sectorâs remarkable breadth, from batteries to built environment (micromobility, electric vehicles, fleet/sharing, autonomy, elecaviation, marintime and public transit/infrastructure, allows us to validate generalizability across heterogeneous contexts. Crucially, mobility sits at the intersection of AI, robotics, and climate techâfast-evolving, signal-rich domains that challenge my ability to model stakeholder perception under uncertainty. These qualities make mobility ventures not just an appropriate case, but a high-leverage proving ground for adaptive, evidence-based decision support.
## 2. Entrepreneurial Decision-Making Under Uncertainty
This section defines the core challenges â differing stakeholder perceptions, circular dependencies, and resource constraints â and maps them to the requirements for a solution. It reviews literature on entrepreneurial decision-making under uncertainty, pointing out gaps that STRAP will address.
### 2.1 Entrepreneurial Decision-Making Under Uncertainty Literature Review
This section introduces the fundamental challenge: entrepreneurs lack decision making model that is personalized, realistically tractable, actionable â promoting imitating other's behavior and unsystematic experimenting that lowers entrepreneuring quality.
Entrepreneurial decision models span a spectrum of complexity defined by two key dimensions: **operational complexity** (decisions unfolding over time) and **multi-stakeholder complexity** (multiple interacting agents or criteria). At the simplest end of this spectrum are single-stakeholder static strategy models, which consider a one-off strategic choice by a lone decision-maker. These simple models are highly tractable but offer poor reality fit, since they ignore dynamic feedback and the involvement of other stakeholders (e.g., Sarasvathy, 2001; McMullen & Shepherd, 2006). Increasing the operational complexity yields single-stakeholder dynamic models that incorporate sequences of decisions and learning over time. Such models capture how an entrepreneur adapts through multiple stages (for instance, via staged investment or real options reasoning) while still focusing on one primary actor (HĂĽkansson, 1971; McGrath, 1999).
To better reflect real venture conditions, other models introduce **multi-stakeholder complexity** even in static settings. These multi-stakeholder strategy models consider how an entrepreneur coordinates or negotiates with various actors (investors, customers, competitors) within a single decision context, adding richer criteria to the choice (Gans, Hsu & Stern, 2002; Van den Steen, 2016). This boosts reality fit by acknowledging diverse interests, but tractability remains manageable only because the decision is not sequential. The most comprehensive models embrace both interacting stakeholders and dynamic operations, portraying entrepreneurship as an ongoing process co-created with partners, markets, and investors (Schindehutte & Morris, 2009; Garud & Karnøe, 2003; Roundy, 2018). These high-complexity models achieve a strong fit to entrepreneurial reality (phenomenological richness) but become computationally intractable â indeed, the full **EDMNO** formulation that accounts for all stakeholders over time is NP-complete (as shown in Section 1.1). In essence, the progression from static single-actor to dynamic multi-actor models reveals an engineering-versus-phenomenological trade-off: as I add realism, I lose tractability.
Bridging this tractabilityâreality gap requires new approaches that retain high reality fit while restoring usability. Rather than abandoning formal modeling, I propose an **engineering** solution that balances these extremes through targeted simplifications and optimization strategies. In this thesis, a three-part framework is introduced to tackle each complexity dimension: **phase-based learning** breaks the decision process into stages to manage operational complexity over time, **proactive hypothesis proposal** uses probabilistic experimentation to navigate multi-stakeholder uncertainty, and **calibrated federated learning** allows entrepreneurs to collectively improve models without sacrificing individual context. Together, these strategies form a mid-complexity decision framework designed to maintain realism (capturing multi-stage, multi-stakeholder dynamics) while remaining computationally tractable. This approach sets the stage for the use cases and solution scope detailed in Section 1.3, aiming to empower entrepreneurs with decision tools that match the actual challenges they face.
| Model Type | Multi-stakeholder Complexity | Dynamic operational Complexity | Tractability | Reality Fit | Key Need | Reference |
| ----------------------------- | ---------------------------- | ------------------------------ | ------------ | ----------- | ------------------------------ | ------------------------------------------------------------------ |
| Single-Stakeholder Static | No | No | High | Poor | More realistic representation | Sarasvathy (2001); McMullen & Shepherd (2006) |
| Single-Stakeholder Dynamic | No | Yes | Medium | Medium | Multiple stakeholder view | HĂĽkansson (1971); McGrath (1999) |
| Multi-Stakeholder Static | Yes | No | Medium | Medium | Sequential decision capability | Van den Steen (2016); Gans, Hsu & Stern (2002) |
| **Multi-Stakeholder Dynamic** | Yes | Yes | LowâMedium | High | **Computational tractability** | Schindehutte & Morris (2009); Garud & Karnøe (2003); Roundy (2018) |
### 2.2 The Entrepreneurial Decision-Making Under Uncertainty Need Analysis
In this section, I identify the root causes of entrepreneurial decision-making challenges across three levels and three dimensions, structured as a need matrix, and show how each need maps to specific elements in My mathematical formulations:
Each cell in this matrix represents not just a conceptual challenge but also maps directly to specific mathematical elements in My formulations, connecting entrepreneurial needs to formal optimization structures:
1. **Perception Challenges** (đ˝ď¸): These map to uncertainty terms $\textcolor{#3399FF}{U_j}$ in the probabilistic formulation and entropy terms $H(p_j|\textcolor{red}{a})$ in the primal-dual approach. Perception asymmetry is captured by stakeholder-specific distributions $p_j$ and expectation terms $\textcolor{#3399FF}{\mu_j}(\textcolor{red}{a})$. Entrepreneurs struggle to understand how stakeholders perceive their ventures. Different stakeholders (investors, customers, partners) project identical startup attributes onto different perceptual dimensions, creating a fundamental information asymmetry problem. Traditional approaches either assume perfect information or rely on intuitive reading of stakeholder signals. My framework uses perceptual projection models to decode how observable venture attributes map to stakeholder decision spaces, enabling entrepreneurs to optimize information gathering and signal presentation.
2. **Sequencing Challenges** (âĄ): These map to the resource constraints $C,\textcolor{red}{A} \leq \textcolor{#8B0000}{R}$ and action variable $\textcolor{red}{a}$, along with the state transition function $D(\textcolor{green}{S},,\textcolor{red}{a})$. The dual decision rule provides a computationally tractable approach to this otherwise NP-complete problem by determining which experiments to run based on their information value per cost. Entrepreneurs must sequence actions optimally under severe resource constraints. However, they cannot directly compute the uncertainty reduction per cost for each possible action sequence due to combinatorial explosion. Standard optimization approaches become computationally intractable as variables increase. My bottleneck-driven framework allows entrepreneurs to make near-optimal myopic decisions by decomposing complex metrics into estimable components, focusing resources on experiments that provide maximum information value. This mapping demonstrates how My mathematical formulations capture each dimension of the entrepreneurial decision challenge. The probabilistic formulation emphasizes uncertainty minimization across stakeholders, while the primal-dual approach transforms this into maximizing stakeholder satisfaction likelihood under resource constraints.
3. **Confirmation Challenges** (đ): These correspond to the interdependencies between stakeholder states in $B,\textcolor{green}{S}$ and the consistency constraints on expectations $\textcolor{#3399FF}{\mu_j}(\textcolor{red}{a})$ across stakeholders. In the dual formulation, Confirmation is addressed through likelihood maximization that aligns all stakeholders' expectations. Entrepreneurs face circular dependencies where stakeholders make simultaneous, interdependent decisions. Investors wait for customer validation, customers require operational proof, and partners need investment signals, creating deadlock situations. Conventional methods tackle stakeholders sequentially, but this approach cannot resolve inherent circularity. My framework enables entrepreneurs to act as central coordinators who leverage information spillovers across stakeholder networks, breaking decision deadlocks through parallel engagement strategies.
This dual formulation provides critical insights by transforming the uncertainty minimization problem into a likelihood maximization problemârevealing how entrepreneurs can maximize the probability of stakeholder satisfaction while optimizing resource allocation.
## 3. Solution design
This section introduces how the needs identified in the previous section is addressed with a **Perception â Action â Confirmation** logic. STRAP integrates two complementary mathematical approaches â probabilistic Bayesian inference and optimization (primal-dual formulations) â to systematically guide entrepreneurial decisions. In essence, STRAP behaves like a âventure GPSâ that perceives the stakeholder landscape, plans optimal actions (experiments), and confirms stakeholder alignment, iterating this cycle as the venture learns. I first discuss the theoretical foundations behind this approach, then detail each component (perception modeling, bottleneck sequencing, and stakeholder Confirmation) in turn. The following table presents My comprehensive solution approach across the three dimensions and three levels of entrepreneurial decision-making:
## 2.2 The Primal Formulation: Uncertainty Minimization
At its core, entrepreneurial decision-making involves reducing critical uncertainties with limited resources. I formalize this as an optimization problem where the entrepreneur seeks to minimize a weighted sum of stakeholder-specific uncertainties:
$\begin{aligned} \min_{\textcolor{red}{a} \in \textcolor{red}{A}} \quad & \textcolor{violet}{W_d}\cdot\textcolor{#3399FF}{U_d} + \textcolor{violet}{W_s}\cdot\textcolor{#3399FF}{U_s} + \textcolor{violet}{W_i}\cdot\textcolor{#3399FF}{U_i} \ \text{subject to} \quad & \sum_{j} c_j \textcolor{red}{a_j} \leq \textcolor{#3399FF}{R} \end{aligned}$
Where:
- $\textcolor{#3399FF}{U_d}$, $\textcolor{#3399FF}{U_s}$, $\textcolor{#3399FF}{U_i}$ represent uncertainties facing demand-side, supply-side, and investor stakeholders
- $\textcolor{violet}{W_d}$, $\textcolor{violet}{W_s}$, $\textcolor{violet}{W_i}$ are the weights reflecting each stakeholder's importance
- $\textcolor{red}{a_j}$ indicates whether action/experiment $j$ is chosen
- $c_j$ is the cost of action $j$
- $\textcolor{#3399FF}{R}$ is the available resource budget
This primal formulation captures the essence of entrepreneurial decision-making: systematically reducing the most important uncertainties given limited resources. Each uncertainty $\textcolor{#3399FF}{U_j}$ can be mathematically represented as the entropy of a stakeholder's belief distribution: $\textcolor{#3399FF}{U_j} = H(p_j(\textcolor{red}{a})|\textcolor{red}{a})$, where $p_j$ is stakeholder $j