Abstract
We study the tradeoff between fundamental risk and time. A time-constrained agent has to solve a problem. She dynamically allocates effort between implementing a risky initial idea and exploring alternatives. Discovering an alternative implies progress that has to be converted to a solution. As time runs out, the chances of converting it in time shrink. We show that the agent may return to the initial idea after having left it in the past to explore alternatives. Our model helps explain so-called false starts. To finish fast, the agent delays exploring alternatives, reducing the overall success probability.
Introduction
In high-stakes environments such as venture-backed startups or rapid R&D cycles, the clock is often the most significant constraint on innovation. Decision-makers face a recurring dilemma: should they continue to execute a risky initial plan, or should they pause to search for a more reliable alternative? This paper formalizes this tension as a tradeoff between fundamental risk and time pressure. By modeling the decision as a dynamic allocation of effort, the authors uncover why even rational agents often succumb to "false starts" and why high confidence in an initial idea can paradoxically lead to a lower probability of ultimate success.
The Model: Doing vs. Thinking
The researchers develop a multi-armed, continuous-time bandit model to analyze how a finite horizon—a hard deadline—alters optimal strategy. The model features two primary modes of action:
- The "Doing" Arm: This represents the implementation of a risky initial idea. Success is stochastic; the agent does not know if the idea is fundamentally viable until it either succeeds or the deadline passes.
- The "Thinking" Arm: This represents the search for a pivot or alternative. Unlike the doing arm, success here provides "progress" rather than an immediate solution. This progress must then be converted into a final product through further effort.
Crucially, the value of the "thinking" arm is time-dependent. As the deadline approaches, the window to convert progress into a solution shrinks. If an alternative is discovered too late, it becomes worthless because the agent lacks the remaining time to implement it. Using Pontryagin’s Maximum Principle, the authors characterize the optimal control path for an agent navigating these constraints.
The "Hail Mary" Effect and the Do-Think-Do Pattern
The most striking theoretical finding is the emergence of a Do-Think-Do behavioral pattern. In many scenarios, an agent begins by "doing"—attempting to solve the problem with their initial idea. If this fails to yield results, they switch to "thinking" to find a better alternative while they still have a sufficient time buffer to convert that discovery.
However, as the deadline looms and the time required for conversion disappears, the value of the thinking arm drops to zero. In a final act of desperation, the optimal strategy often dictates a return to the original risky idea. This "Hail Mary" pass occurs because, despite the low probability of success, the initial idea remains the only path that can still yield a solution in the final moments. This explains why projects often return to previously abandoned concepts when under extreme duress.
The Confidence Trap and False Starts
The model provides a rigorous explanation for two common phenomena in entrepreneurship:
- The Confidence Trap: Counterintuitively, the authors find that a higher initial belief in the first idea can decrease the total probability of success. High confidence encourages the agent to persist with the "doing" arm for too long. By the time they realize the first idea might be a failure, they have exhausted the time buffer necessary to make a pivot (the "thinking" arm) viable.
- False Starts: The model explains why agents often rush to market with unrefined ideas. To minimize the expected cost of effort (the "burn rate"), agents may frontload the risky work. While this increases the chance of finishing early, it significantly reduces the overall probability of success compared to a "backloaded" strategy where the agent thinks first and does later.
Strategic Implications and Limitations
For stakeholders like venture capitalists, the research suggests that deadlines are a double-edged sword. While tight deadlines can prevent a waste of resources on unproductive "thinking," they can also crowd out the necessary research phase, forcing entrepreneurs into premature implementation.
As a theoretical work, the paper acknowledges several simplifications. The model assumes a unidimensional search for alternatives and abstracts away from exponential discounting to focus purely on the deadline effect. Furthermore, it assumes the agent is a single entity, leaving the complex dynamics of Principal-Agent conflicts (e.g., between a founder and an investor) for future research.
Conclusion
By introducing a finite horizon to the classic bandit problem, this paper provides a nuanced framework for understanding how time pressure distorts decision-making. The discovery that success probability is non-monotonic relative to both time and initial confidence challenges standard economic intuitions and offers a new lens through which to view the high-failure rates of time-sensitive innovative ventures.
Summary updated