How can you fit in and stand out at the same time? It’s a challenge we all face in our personal and professional lives. We want to be accepted by the group but also recognized as unique.
Businesses face a similar dilemma. They need to stand out to gain a competitive edge but also fit industry norms to be accepted by customers and investors.
So how can a business balance these competing pressures? One theory suggests a curve showing profitability as a function of typicality, where there’s an optimal level of differentiation—a balance between copying others and being too different.
This idea usually focuses on product differentiation, but a new paper from the University of California, Berkeley, argues it misses a key part of brand identity: how a company presents itself to outsiders.
Standing out
There are many ways to stand out, from meme-worthy ad campaigns like Old Spice’s to the Hawaiian shirts worn by Trader Joe’s employees. But the researchers focused on one specific signal: the language used by executives in earnings calls.
Using a deep-learning model, they analyzed transcripts of over 60,000 quarterly earnings calls from 2008 to 2016. They identified firms whose self-presentation was distinctly different from their competitors. Unsurprisingly, Tesla was much less conventional than Ford, and the mobile banking platform Green Dot was far removed from JP Morgan Chase. Conversely, Dell and Bank of America were among the most conventional.
When they examined earnings forecasts, they found that stock analysts who attended these calls projected higher earnings for the more atypical firms, likely boosting their stock prices at the time. The researchers called this the “performative atypicality premium.”
However, there was a catch: When they compared the consensus forecasts to actual financial results, they found that these atypical firms underperformed analysts’ expectations, leading to negative earnings surprises.
Intangible magic
To analyze the transcripts, the researchers used a word-embedding model. This model is trained to predict hidden words based on the context of adjacent words. For example, if the sentence is “I took my ___ to the veterinarian,” the model learns that “dog” is more likely than “father” or “sofa.”
Words are mapped as coordinates in a 300-dimensional space, where the distance between them reflects their semantic proximity. The researchers then calculated the geometric center of all the words a person has spoken—their “center of gravity”—and compared it to other speakers.
The word maps also showed that there are different ways to be different, and some are more effective than others. Specifically, the more a CEO’s language resembled that of celebrated innovators, the greater the performative atypicality premium their company received. This is akin to a CEO choosing to wear a black turtleneck instead of a suit, à la Steve Jobs.
In a rare scholarly nod to Monty Python, the authors cite a scene from “Life of Brian.” The protagonist, mistaken for Jesus, tells his thousands of followers that they are all individuals. They respond in unison, “We are all individuals!” except for one voice that shouts, “I’m not!”
Of course, there’s a fine line between being viewed as a genius or a goofball, and it’s not always easy to stay the right side of that line, but if leaders can manage it, then the data suggests that investors will reward them for their eccentricity.
