Last week Lego posted a 25% jump in revenues and a 31% rise in net profits for 2015.

This is a far cry from 2003 when the company was in trouble, having lost 30% of its sales from a year earlier. In 2004, another 10% vanished. As Lego’s CEO Jørgen Vig Knudstorp put it: “We are on a burning platform, losing money with negative cash flow, and at a real risk of debt default, which could lead to a breakup of the company.”

The solution to Lego’s problems–the thing that may have rescued it from potential bankruptcy–lay in an old pair of sneakers.

Every big data study Lego commissioned concluded that future generations would lose interest in Lego. “Digital natives”–men and women born after 1980 and came of age in the Information Era–lacked the time and patience for Legos, and would quickly run out of ideas and storylines to build around. Each Lego study showed that the generational need for instant gratification was more potent than any building block could ever hope to overcome.

small data cover Cover image courtesy of St. Martin’s Press LLC

In the face of such a prognosis, it seemed impossible for Lego to turn things around – but it did. It sold off its theme parks; continued successful brand alliances with the Harry Potter, Star Wars, and Bob the Builder franchises; and reduced the number of products while entering new and underserved global markets.

But probably the biggest turnaround in Lego’s thinking came as the result of a visit Lego marketers paid in 2004 to the home of an 11-year-old German boy. Their mission? To figure out what really made Lego stand out.

In addition to being a Lego aficionado, the boy was also a passionate skateboarder. Asked which of his possessions he was most proud of, he pointed to a pair of beat-up Adidas sneakers with ridges and nooks along one side. Those sneakers were his trophy, he said. They were his gold medal, his masterpiece. More than that, they were evidence. Holding them up so everyone in the room could see and admire them, he explained that one side was worn down and abraded at precisely the right angle. The worn-down sneakers signaled to him, to his friends, and to the rest of the world that he was one of the best skateboarders in the city.

At that moment, it all came together for the Lego team. It realized that children attain social currency among their peers by playing and mastering their chosen skill, whatever that skill happens to be.

Until that point, Lego’s decision making was predicated entirely on reams of big data. Ultimately it was a small, chance insight – a pair of sneakers belonging to a skateboarder and Lego lover – that helped propel the company’s turnaround. From that point on,Lego refocused on its core product, and even upped the ante. The company not only re-engineered its bricks back to their normal size; it began adding even more, and smaller, bricks inside their boxes.

Ten years later, in the wake of the worldwide success of The Lego Movie and sales of related merchandise,Lego’s sales rose 11% to exceed $2 billion. For the first time,Lego surpassed Mattel to become the world’s largest toy maker.

Big data might find it hard to find meaning, or relevance, in insights like these. In every big data study, there is a missing question: How might these findings be combined with small data to affect or transform a brand or business?

Case in point: a few years ago an unnamed banking institution faced major challenges comprehending the behavior of its customers even after leveraging a big data analytics model designed to prevent churn, customers who move money around, refinance their mortgages, or generally show signs they are on the verge of exiting the bank. Thanks to the analytics model, the bank found evidence of churn, and promptly drafted letters asking customers to reconsider. Before sending them out, though, the bank executive who had hired me discovered something surprising. Yes, “big data” had seen evidence of churning. But it wasn’t because customers were dissatisfied with the bank or its customer service. No: most were getting a divorce, which explained why they were shifting their assets. A parallel small-data study could have figured this out in a day or less.

As accurate as big data can be while connecting millions of data points to generate correlations, it is often compromised whenever humans act like, well, humans. As big data continues helping us cut corners and automate our lives, humans in turn will evolve simultaneously to address and pivot around the changes technology creates. Big data and small data are partners in a dance, a shared quest for balance–and information.

In our small data lies the greatest evidence of who we are and what we desire, even if, as Lego executives found out more than a decade ago, it’s a pair of old Adidas sneakers with worn-down heels.

This article is adapted from the New York Times bestseller Small Data: The Tiny Clues That Uncover Huge Trends by Martin Lindstrom. Copyright 2016 by the author and reprinted by permission of St. Martin’s Press, LLC.