In a recent essay, Eugene Yan draws on a familiar maxim from Elon Musk — "If things are not failing, you're not innovating enough" — to make a case about how data science teams should operate. The core argument is that a healthy innovation culture cannot exist if failure is punished or hidden. Instead, teams need explicit permission to try approaches that might not work, because that is the only way to discover what does.
Yan suggests that the role of leadership is to create an environment where experimentation is safe. This means setting expectations that not every project will yield a positive result, and that the value of a failed experiment often lies in the knowledge gained. For data science, where uncertainty is inherent, this is especially important: models and hypotheses frequently fail, and that is a sign of genuine exploration rather than incompetence.
The essay does not offer a step-by-step playbook, but it points to a cultural shift. Teams that celebrate learning from failures, rather than only celebrating wins, are more likely to take the risks that lead to breakthroughs. The message is straightforward: if your data science team never fails, you are probably playing it too safe to innovate.