Cultural renaissance: How Haas faculty are leading a new wave of data-driven culture research

Unraveling the mystery of culture

Your company has a mission statement and carefully crafted values. Maybe they’re painted on the office walls. But what’s the real, day-to-day culture like? If you’re like a lot of managers, the answer has often been, “I know it when I see it.”

Not anymore. The same digital services that have upended how workers communicate—think G Suite, Slack, and Zoom—are also helping to unravel one of the great mysteries of the workplace: culture. And they’re ushering in a new wave of research based on how employees act that until now has been elusive.

“Any time employees interact with each other, they are leaving digital traces and creating sources of data that weren’t available before,” says Sameer Srivastava, a Haas associate professor of organizational behavior. “This not only helps us to better understand how employees behave but to do so in ways that are dynamic.”

Jennifer Chatman and Sameer Srivastava
Prof. Jennifer Chatman and Assoc. Prof. Sameer Srivastava

Srivastava and Prof. Jennifer Chatman are at the forefront of this new wave: This year, they co-founded the Berkeley Haas Culture Initiative to help make Berkeley a hub for advancing the field. They launched the effort with a groundbreaking conference last January, bringing together leading academics with executives from “culture-forward” companies like Netflix, Zappos, Deloitte, and Goldman Sachs to potentially collaborate on new, data-driven studies. The second Berkeley Haas Culture Conference is scheduled for January.

Chatman, the co-creator of the Organizational Culture Profile (OCP)—a survey tool that’s been a dominant measure of organizational culture for more than 30 years—wants to ensure that new approaches as not only methodologically rigorous but also useful to managers.

The quantitative side of culture research

For too long, Chatman says, culture research amounted to little more than “advanced gossiping,” with scholars visiting companies and writing reports based on observations. That all changed a few years ago when, buoyed in part by new data sources and research methodologies, a group of academics set out to jumpstart the quantitative side of culture research.

“Managers, too, eventually realized that they would rather learn about the links between organizational culture and performance in analyses of millions of emails, thousands of anonymous reviews, or other data on large swaths of employees rather than a few stories about what did or did not work in one specific organization or another,” says Chatman.

In fact, a dataset of over 5 million emails from a technology company has been fodder for a series of studies by Srivastava and Amir Goldberg of Stanford, who co-directs the Computational Culture Lab—a joint Berkeley-Stanford project that harnesses data science and computational methods to study culture. In one paper, Srivastava and collaborators found that the extent to which workers’ language matched their peers’ communication styles predicted performance and retention. A more recent paper co-authored by Chatman, Srivastava, Goldberg, and PhD student Richard Lu demonstrates the power of combining culture measures based on surveys with natural language analysis through machine learning to predict culture fit and key performance outcomes.

Both Srivastava and Chatman have analyzed hundreds of thousands of company reviews on the job site Glassdoor: Chatman has recently published a study showing that employees perceive a more negative environment in companies with “fixed-mindset” rather than “growth-mindset” cultures. Srivastava found that the most highly valued and innovative firms were those whose employees embraced a variety of cultural values—suggesting that people can be creative and innovative, but still conform to company norms and beliefs.

Srivastava is also co-authoring a study that shows how managers trying to shape culture after mergers can anticipate—from employee emails and other forms of communication—where integration challenges might lie.

The researchers say this new era of culture research extends far beyond text analyses. With the right privacy and confidentiality protocols in place, the tone of voice and even facial expressions people use when communicating might become common sources for understanding subtle, real-time shifts in culture—enabling managers to better manage culture as a strategic asset.

“By combining new digital tools with established survey-based approaches, we can get a more nuanced view of culture than ever before,” Chatman says.