Entrepreneurial Experimentation Design for Venture Finance

This paper examines how entrepreneurs strategically design experiments to convince venture capitalists (VCs) to fund their projects when investors interpret data through heterogeneous statistical frameworks. Drawing on Liang (2021)'s model of games with incomplete information played by statistical learners, we translate her abstract theoretical framework into a practical venture capital setting. We characterise how entrepreneurs balance funding plausibility against equity dilution by strategically choosing experiments along two critical dimensions: sample size (precision) and dimensionality (breadth of metrics). Our analysis reveals that for genuinely high-quality projects, increased precision helps by forcing VC beliefs to converge toward true quality. For low-quality projects, funding depends on preserving disagreement through sparse, high-dimensional experiments. The entrepreneur's optimal design depends critically on their prior beliefs: confident entrepreneurs choose high-precision, low-dimensional experiments that minimise equity dilution, while uncertain entrepreneurs opt for sparse, high-dimensional experiments that maximise the probability some VC will hold sufficiently optimistic beliefs. The framework provides a rigorous foundation for understanding how entrepreneurs and investors can rationally disagree when observing identical evidence, with significant implications for strategic information design in entrepreneurial settings.