ÌÇÐÄÊÓƵ

Research uses AI to showcase the role of pivots in startup performance

For startups, a pivot is challenging and risk-driven, and the big question is not "whether to pivot?" but how to identify those pivots that will create a meaningful difference in performance. Consider the origin of X/Twitter: podcast platform Odeo. Entry to its space by iTunes and Apple prompted a pivot—to microblogging. The rest is history.

But for every successful pivot like that producing Twitter, countless founders and startups are forgotten or never heard of because they do not engage in purposeful pivots–those that were focused, coherent and impactful. Rather, they abide by the "pivot often in response to customer feedback" mantra to undertake reactive or remedial pivots for superficial changes in their business model, thus failing to unlock the value-creation potential of their underlying ideas.

Tackling the deficiency is new research co-authored by Rajshree Agarwal, the Rudolph Lamone Chair of Strategy and Entrepreneurship at the University of Maryland's Robert H. Smith School of Business. Using AI tools in the form of large language models (LLMs), the in Strategy Science provides new evidence for how entrepreneurs pivot, and why it matters.

Either a "bias for action" (acting without articulating reasons why) or "paralysis by analysis" (overthinking to a point of indecision) can stymie a startup's ability to engage in the right type of pivot, one that delivers on its survival and growth goals.

To counter this, "we show how and why entrepreneurs who take the time to articulate the 'why' and utilize experiments to test their assumptions are more likely to make pivots that are purposeful and coherent," says Agarwal, who directs Smith's Ed Snider Center for Enterprise and Markets and co-authored the findings in Strategy Science with Smith Ph.D. student Jacob Valentine and Elena Novelli, professor of strategy at Bayes Business School (U.K.).

The paper, Agarwal says, "unpacks the two dimensions of theorization and experimentation by showing how they are complementary for pivots by mixing LLM techniques with business case analysis."

The researchers used these mixed methods of examining rich data from more than 1,600 interviews with 261 entrepreneurs in London over a nine-month period in 2019 and 2020.

More broadly, the work has created a publicly available AI-generated dictionary of words and algorithms for future research. This has "important implications for practitioners (e.g., entrepreneurs, incubators/accelerators and business leaders) and policymakers (e.g., government agencies, such as the National Science Foundation and the National Institutes of Health that sponsor entrepreneurial and strategic training programs)," the authors write.