On Failure

“Know Thyself” is a maxim inscribed on the temple of Apollo in the ancient Greek city of Delphi. Its origin is not fully settled. While some may attribute it to the poet and philosopher Ion of Chios, others mention it as the core of Socratic philosophy. Stoic philosophers consider it a stepping stone in the search for wisdom, which constitutes “the mother of all virtue,” according to Marcus Tullius Cicero. The injunction means that one should know oneself through the practice of self-awareness and self-examination. And what’s the best way to do that than learning from mistakes and failures?

Nowadays, no one disputes that true learning unquestionably occurs through making mistakes. Some even encourage people to make as many as possible in order to learn fast. In business, much ink has been spilled over the subject, and it doesn't get any more popular than in the startup world and tech industry. In fact, the concept of the lean startup, popularized by the book of the same name, suggests that one should develop a product or service with incremental features and validate the value with customers so that one can learn from mistakes and correct them early on in the process. The culture of encouraging and learning from failures largely permeates business discourse, and rightly so. This is an important ingredient of the success recipe where agility helps address an ever-changing environment, countering the rigidity of bureaucratic organizations.

To learn from failure, one needs to ensure understanding of what went wrong and trace back the outcome to the specific input. By doing so, one can easily change the input in hopes of a better outcome. For instance, in developing software, a feature may be designed to enhance the user’s experience, but testing might demonstrate otherwise. In other words, the direct and linear relationship between the input and outcome is fundamental in the learning process. This is called a deterministic relationship, where the response is directly and unequivocally caused by the action. For example, if you don’t know how to drive a car and you drive it anyway, crashing while leaving the garage, you can be pretty sure that you crashed because you didn’t know how to drive, and learning how to drive would be the next step.

The question worth asking then is: is it always easy to isolate the cause in things that didn’t work very well, or for that matter, things that worked well, so one can learn from the experience? Is the world so deterministic as to make learning from failure something easy and repeatable? These are important questions whose answers are crucial for learning but makes it difficult to achieve. Let me explain:

There is a domain in the business world and in life in general where determinism applies, where the response can easily be traced back to the specific stimulus. This domain is characterized by a mechanistic relationship; it’s the world of machines obeying strict physical laws. It’s also an engineering world. Time also plays an important factor in determinism when the response occurs quickly. Putting your hand on a hot stove, for example leads to quick reaction and the learning is straightforward.

However, there is another domain, vastly larger than the first, where randomness takes over. It’s the locus of multiple moving parts with very complex relationships, and it doesn’t get any more complex than in a business setting. Developing and introducing a product to a market, making, and implementing a strategic choice entails a set of people and activities with internal complex dynamics along with relationships with external environments. The result of these choices, good or bad, can have multiple and complex causes, making the whole learning process difficult. You can make the best strategic choice given the available information, and the result can be disastrous. Or, you can end up with a good outcome without having to make any strategic choice at all.

Moreover, let’s assume, for the sake of argument, that one can confidently determine the specific cause of an outcome. You decided to invest resources in developing a new product or service, and despite early positive feedback from customers, the enterprise didn’t take off. The decision wasn’t good obviously, but only in hindsight. Does that mean that making the decision again would lead to the same result? The answer would be yes if and only if the context in which the decision was made doesn’t change. But we know that things don’t stand still; the world changes, time progresses. In other words, one can make the same decision and wind up with a different outcome simply because the context changes. This whole idea about learning from experience draws on the concept of induction, where one can apply knowledge acquired from a certain situation to a completely different one. Induction, viewed as a heuristic, can definitely help us to address the uncertain world, but it can also lead to fundamental biases, and experience is one of them.

Overall, learning from mistakes and failures can be easy in very specific conditions where determinism is the ruling feature, and knowledge is derived mechanically from experience1, but is very hard to do in complex and entangled situations. The learning, therefore, starts by identifying the context in which the situation occurred. This requires drawing upon deep thinking and adopting a complexity perspective, both of which unfortunately are lacking in today’s business world where people want quick answers, usually drawn from conventional and popular ideas. One should acknowledge that due to the sheer level of complexity involved, this is no guarantee for identifying the cause of the failure, but the thinking process is very valuable.

That being said, learning from failure should leave room for learning regardless of failure. This happens when one reflects on what happened irrespective of the outcome because the latter tends to cloud the learning. A negative outcome does not necessarily mean an initial bad decision. So, what we are left with is analysing decisions using a different yardstick, that of convexity. A good decision is one that has a limited downside and high or open upside. It’s a convex decision. The opposite is a concave decision. One should strive for the former and let randomness work its magic and avoid the latter at any cost.

Previous
Previous

On Inequality

Next
Next

Decision Making Under Uncertainty