How Open-ended Exploration Leads to Success: a $15 Million Lesson in Entrepreneurship

Those who have read my previous article "How I Lost 15 Pounds in 6 Months and Beat My Weight Loss Plateau for Lasting Results" might know that I ended a five-year entrepreneurial journey a year ago. During those five years, I led my company through several rounds of financing, from angel to pre-seed and seed, achieving a valuation of $15 million. However, I ultimately chose to close the company and embark on a new journey of self-healing. In this year and a half of physical and mental healing and continuous reflection, I finally found the core reason for this entrepreneurial failure. And this core reason is not unique to my previous startup because human weaknesses are universal. I chose to share this experience, hoping it would provide some insights to other startups and help them avoid detours. This experience is also relevant to personal career exploration and upgrading.

Summarizing the reasons for a startup's failure is relatively easy. I once quickly wrote hundreds of pages outlining the various decision-making issues encountered in the previous entrepreneurial process. Still, I could never penetrate the essence of its failure until I read the book "Why Greatness Cannot Be Planned." Suddenly, I understood the real reason for my previous startup's failure – I used to be "plan-oriented" rather than "exploration-oriented." I was "limited by goals."

"Why Greatness Cannot Be Planned," by Kenneth O. Stanley and Joel Lehman, argues that greatness and innovation are often not achieved through detailed planning and linear thinking but through exploration, serendipitous discovery, and non-linear paths.

Kenneth Stanley and Joel Lehman have extensive work experience at OpenAI and other well-known institutions. While at OpenAI, Kenneth Stanley focused on artificial intelligence and machine learning, advancing evolutionary algorithms and open-ended exploration methods. Joel Lehman, a research scientist at Uber AI Labs, worked on generative design and adaptive systems. Together, they emphasize pushing beyond traditional goal-oriented strategies through open-ended exploration and innovation, driving significant advances in AI.

The book describes a robot maze experiment comparing "goal-oriented" and "non-goal-oriented" approaches. The goal-oriented group of robots was programmed to find the maze exit directly, resulting in only a 30% success rate and an average time of 45 minutes; as such, robots tend to get stuck in local optima and need more flexibility. In contrast, the non-goal-oriented group of robots used random exploration and trial-and-error methods, achieving an 85% success rate and finding the exit in just 15 minutes on average. This demonstrates that open-ended exploration significantly enhances adaptability and efficiency in complex environments.

Kenneth and Joel also provide examples showing that greatness and innovation often come from open-ended exploration and serendipitous discoveries rather than detailed planning. For instance:

• Alexander Fleming discovered penicillin after accidentally finding a mold with antibacterial properties, showcasing that significant discoveries often stem from unexpected occurrences.

• Slack initially started as a gaming project but pivoted to an enterprise communication platform after the game's failure, leading to its unexpected success.

• Wikipedia's success originated from the spontaneous contributions and collaboration of its user community. It evolved from an open online encyclopedia project to one of the most popular knowledge platforms worldwide.

• Wilhelm Röntgen discovered X-rays accidentally while researching cathode rays, revolutionizing medical diagnostics.

• Werner Heisenberg and Erwin Schrödinger advanced quantum mechanics through serendipitous discoveries and theoretical breakthroughs while studying microscopic particle behavior.

• McDonald's founders, the McDonald brothers, unintentionally created a new fast-food service model by continually optimizing service processes and menus.

• Kevin Systrom and Mike Krieger initially created a check-in app but, driven by user feedback and market demand, focused on photo-sharing features, leading to Instagram's global popularity.

• Drew Houston created Dropbox to solve his own file synchronization problem across devices. He continually improved the product to attract many users and investors.

• Reed Hastings and Marc Randolph initially founded Netflix as a DVD rental service but pivoted through continuous adjustments and business transformations, making Netflix the world's largest streaming service provider.

• During a hackathon, Twitter's founding team proposed the "status update" idea, which eventually evolved into today's Twitter.

• OpenAI also went through periods of uncertainty and exploration without a specific development direction but gradually found its path through continuous experimentation and exploration.

These examples illustrate that many successful business outcomes and scientific discoveries are not achieved through detailed planning and goal-oriented methods but through open-ended exploration, flexible adaptation, and serendipitous inspiration.

Kenneth Stanley and Joel Lehman use a mountain-climbing analogy to explain the different results of "goal-oriented" and "exploration." They describe climbers searching for the highest peak: the first climber, seeing a higher peak at the starting point, might choose a seemingly direct path to the summit. However, as the climb progresses, the path becomes steeper, ultimately reaching a seemingly high point, but the climber is unable to continue climbing and gets stuck in a local optimum.

The second climber chooses a winding path, which may not appear to ascend initially, but through continuous exploration and adjustment, eventually finds a path leading to the highest peak. This climber might experience temporary descents and detours, discovering seemingly unrelated trails and routes, but each attempt and exploration gradually guides them closer to the final goal. During this process, the climber not only gains more experience and skills but also develops the ability to adapt to different terrains and environments.

The second climber's experience demonstrates the importance of open-ended exploration. They are not limited by predefined paths and goals but gradually find the optimal solution through continuous experimentation and adjustment. This approach is essential for innovation and solving complex problems, as it allows us to discover new paths and possibilities instead of being confined to existing limitations. Through such exploration, we can find better solutions, gain more experience and abilities, and enhance our adaptability and creativity.

During our financing process, I learned that investors value whether you have clear goals and a detailed financial and company development plan for the next three or even five years. They like to see competitive analysis and understand if you have a strong team. Society encourages "certainty" because humans fear the risks of "uncertainty." Ironically, entrepreneurship itself is an adventure full of uncertainty. Therefore, if entrepreneurs habitually use various "analyses" and "plans" to create a false sense of certainty, they often focus on local optima and, lacking flexibility, end up dead ends.

I participated in various entrepreneurial planning competitions from my first year of undergraduate studies. In my second year, I won first place in the China region of the Oval China-Japan-Korea Entrepreneurship Planning Competition, becoming one of the only 30 members selected from the China region to participate in the competition in Japan. Tracing back to earlier times, from middle school (when I was 12), I proactively made daily study and life plans without any guidance or requirements. I have shown strong goal-planning and execution abilities since childhood. People have always said that I know what I want, referring to my goal orientation, which I once took pride in. I have always instinctively made various plans because they gave me a sense of security and control over my life.

But this is also why I suffered from depression when I decided to start my own business in 2016 – nothing went according to my plan. Human civilization is a history of exploration, just as no one can predict earthquakes and financial crises, and no one can genuinely plan their life. Humans, out of self-protection, make various plans; when those plans fail, they go mad. The ultimate reason is the lack of an open exploratory mindset. Open exploration sounds dangerous and counterintuitive; it sounds risky and might threaten our survival.

ByteDance is one of the world's highest-valued internet startups, worth $300 billion. In its early days, the company had a general direction – "focusing on information distribution" – but did not determine what specific application to create. Instead, it launched hundreds of apps simultaneously for market testing, focusing on the ones with the best performance and cutting the underperformers. This laid an essential foundation for TikTok's later success as a mainstream short video app globally – it wasn't based on the founder's specific plan but on large-scale testing under a core goal (focusing on information distribution).

A famous counterexample is Quibi, founded by Hollywood producer Jeffrey Katzenberg and former HP CEO Meg Whitman in 2018. Quibi aimed to provide high-quality, short-duration (5 to 10 minutes) mobile video content, targeting millennial and Gen Z users. The platform officially launched in April 2020, collaborating with famous producers and stars to create many original content. It was said that before the app launched, it had already raised $1.7 billion based on a "concept" or "goal," with $100 million in advertising commitments from significant advertisers like Google, Procter & Gamble, and Walmart, showing great confidence in the platform's potential.

However, Quibi operated for only six months. Due to the significant gap between paid subscribers and expectations, the founders returned the remaining $350 million to investors and shut down the platform. They didn't choose to continue using the $350 million for market testing and finding new solutions because the company strictly followed the founders' product "vision" and plan. When unexpected events occurred, the founders were thrown off balance, lacking flexibility in market response, leading to an instant collapse.

Although the company leaders blamed the dismal outcome on the "COVID-19 pandemic," I disagree with this excuse to save face. I believe the company's problem was that the leaders launched the product entirely according to their "vision," neglecting extensive real-market experiments and testing. Previously, Jeffrey Katzenberg made movies, and Meg Whitman made computers. Their industries (movies and hardware devices) share a common characteristic – they require significant upfront resource investment to develop a product, hoping it will succeed in the market.

However, Quibi belongs to the software business, which can leverage lower upfront investments to gain honest market feedback and then determine the direction and expand production. Yet, Jeffrey and Meg applied traditional rules from the movie and hardware industries to run the software business, leading to a high failure rate.

Of course, this is not to say that "goals" are meaningless; on the contrary, they are still essential. Goals are like our compass for navigating the seas or the North Star in our hearts. Without an exciting goal, we would not begin exploration or take action. However, we must understand that goals are merely our guides or hypotheses that need to be validated through experiments. They do not represent the answers themselves. As for what the answers are, we need to discover them. Each test, experiment, and exploration brings us closer to the truth.

Therefore, embrace an "experimental" life. Through open-ended exploration, bold experiments, and attempts, we will surely find our "new continent."