There’s a reason we assume hard work always pays off, that great products always find their market, and that bold decisions by bold leaders always pan out. We’re not drawing those conclusions from the full data set. We’re drawing them from the fraction that survived.

That’s the core of Survivorship Bias: when we evaluate outcomes, we instinctively look at the results that are visible to us. The failures, the shuttered companies, the abandoned projects, and the quietly shelved ideas don’t show up in our feed. So we treat their absence as if they never existed.

The most famous demonstration of this comes from World War II. Statistician Abraham Wald was part of the Statistical Research Group at Columbia University, tasked with helping the military figure out how to better protect bombers. The team examined returning aircraft and noted heavy damage to the wings, the tail section, and the fuselage. The instinct was to reinforce those areas.

Wald saw the flaw immediately. The planes that went down, that weren’t surviving, had much better information to provide on the areas most important to reinforce. Wald’s insight was to consider all the things that started on the same path but didn’t make it. The damage on the returning planes showed exactly where you could get hit and survive. The undamaged areas on returning planes were the critical vulnerabilities, because planes hit there never came back.

That logic should feel obvious in hindsight. But it almost wasn’t caught.

The psychological mechanism underneath this is straightforward: we can only reason about what we can observe. Failures are invisible. Successes are loud. So our brains build mental models of how the world works based almost entirely on the cases that made it through.

Nassim Taleb called the data obscured by survivorship bias “silent evidence.” We favor the visible, the tangible, and the concrete, and we discount what isn’t in front of us. This isn’t laziness. It’s just how attention works.

The business world is particularly prone to this. Companies that fail early on are ignored, while the rare successes are studied and celebrated for decades. Studies of market performance often exclude companies that collapsed, which can distort statistics and make success seem more probable than it actually is.

The same dynamic plays out with the college dropout mythology. A handful of extremely famous dropouts become the story. The millions of people who dropped out and didn’t build billion-dollar companies are never mentioned, not because their experience is irrelevant, but because it’s invisible.

This is the bias in action: the exception becomes the template.

For product teams, survivorship bias shows up quietly and often. And when it does, it tends to distort the decisions that matter most.

The most common version: using successful past projects as the default model for how to run new ones. A team ships a product under aggressive timelines, it somehow works out, and suddenly, aggressive timelines become the standard operating procedure. Nobody accounts for the dozen other projects that burned out under the same conditions and were quietly cancelled. The same thing happens with feature decisions. A team points to a competitor’s popular feature and says, “We need that.” But they’re only seeing the one feature that broke through. They’re not seeing the fifty features that same competitor shipped and buried because nobody used them.

Research practices on product teams can fall into this, too. When discovery focuses mostly on active, engaged users, you’re hearing from the people who stuck around. The users who churned, bounced after onboarding, or never returned after a confusing first session aren’t in your interview queue. Their experience is part of the product story. It’s just the part you’re not seeing.

Team culture can be shaped by it, too. High performers who thrive under certain conditions get promoted, get profiled, and get asked to speak at all-hands. Their approach to work gets codified as “how great work happens here.” But that framing completely ignores everyone who tried the same approach and burned out, disengaged, or left. The model gets built on the outliers, not the distribution.

There’s also a subtler version that shows up in how teams talk about process. Someone says, “We did this without documentation, and it turned out fine.” Maybe they did. But that one success doesn’t mean the approach is sound; it means it worked once, for those people, under those conditions. The times it didn’t work aren’t part of the story because nobody’s telling those stories.

The fix isn’t cynicism and not believing anything at all. Instead of asking “what worked?” ask “what are we not seeing?”