How the Study Makes “Gut Feel” Testable
This chapter reconstructs how the focal study turns a vague dispute about venture-screening intuition into a defined comparison. The subject is the early-stage NTBF seeking support; the focus is the study’s sample, survival outcome, baseline, subjective benchmark, and predictive model. Understanding those choices lets a reader judge the later percentages as evidence about a particular design rather than as a general verdict on human judgment or venture quality.
The design matters because “objective” and “intuitive” are not self-explanatory categories. A business plan can be read by a person, coded into a relation score, or ignored in favor of the sample’s overall survival rate. Each route answers a different question. Before comparing their performance, we need to know who entered the comparison, what counted as survival, and how each prediction was generated.
A cohort that begins before incorporation
The study uses 272 business plans submitted to a high-tech business-plan award in Baden-Württemberg, Germany, between 2006 and 2014. The ventures are new technology-based firms, a population that includes classic start-ups and research spin-offs, although company spin-off projects were excluded from the final sample because their access to resources differs from that of other ventures. The competition’s standardized template meant that the plans shared a common structure, which made the text-based coding more comparable.
A consequential feature of the dataset is its timing. Some plans describe ventures before a legal entity has been created. That lets the researchers include projects that fail before formal registration, rather than observing only the ventures that survive long enough to appear in a public business register. In studies built from registered firms, this early selection can be invisible: the dataset begins after some fragile projects have already disappeared. Starting with applications avoids one form of survivor bias, although it does not make the sample representative of every venture in every region.
The sample was assembled through a network approach and comes from a particular innovation-intensive German region and competition. The authors compare its characteristics with a broader study of the technology-based venturing landscape in Baden-Württemberg and regard the subgroup as reasonably representative of that local setting. That is useful context, but it is not a license to generalize the result to ordinary small businesses, later-stage companies, or ecosystems with different institutions and markets. The population being screened is part of the result.
Five years after submission of a plan, each venture is assigned a binary outcome. It counts as surviving if it is still operating without signs of insolvency and is registered in the German trade register or has made a successful exit. It counts as non-surviving if it has been removed from the register, is insolvent, or has exited through a forced sale caused by imminent bankruptcy.
The successful-exit rule deserves attention. For the study’s purpose, survival means that the venture continued as an operating project or completed a form of exit rather than being forced out by failure. For an investor, however, a successful exit can carry a very different financial value from a small company that continues operating. The binary label is therefore a continuity measure, not a return measure. It is a practical common denominator for heterogeneous ventures, not a complete definition of success.
The authors give two reasons for choosing survival. First, the ventures may pursue different goals.