When we talked with Yammer CTO Adam Pisoni earlier this year, he dropped some iterative knowledge on us: The Microsoft-acquired social business startup forms and dissolves its engineering teams.
Members get drafted–and redrafted–continually. Pisoni told us that they iterate on software, so why not the organization?
It’s a similar case at GitHub, the happiness-optimized social coding site that bootstrapped for years before taking on $100 million in funding from Andreessen Horowitz. As CEO Tom Preston Werner told us, people across the company can be recruited to join project teams as they develop–and later disperse.
Instead of being bound to one another, the teams stay together as long as the project necessitates. Pisoni told us this was great because of the breadth of relationships that form between people, allowing for the implicit knowledge sharing between folks that predicts the most innovative–and most successful–teams.
But there’s another reason temp teams do so well when they take the stage: To understand, we’ll look at Broadway.
As David Burkus notes at Harvard Business Review: Most artists on Broadway work on more than one production a year.
“As such,” David Burkus notes at hbr.org, “artists develop a broad and interconnected network of relationships and can find themselves working with lots of old colleagues or teams of whole new people.”
That loosely networked grouping caught the attention of researchers Brian Uzzi and Jarrett Spiro, who used it for what’s become a formative paper, in which they analyzed every Broadway show from 1945 to 1989, totaling a database of 474 musicals and 2,092 artists. Writers, actors, producers, and choreographers spent all that time connecting, collaborating, and disbanding–the kind of fermentation you’d otherwise only see in kombucha.
Uzzi and Spiro called this latticed working world a “small-world network.” Then they set to calculate how a network of packed connections made for the most successful productions. Which gave some unexpected results, as Burkus notes:
. . . The pair calculated the level of repeat collaborations in any given production year, a value they called “small world quotient” or simply Q. When Q was high, the teams were densely interconnected, which meant that more artists knew each other and were working together on multiple projects. When Q was low, there wasn’t as much familiarity and multiple collaborations were rare. Uzzi and Spiro then compared each year’s Q score to the level of financial success and critical acclaim achieved by the shows that year. Given what we know about teams, it would be logical to assume that a higher Q would produce shows that were more creative and successful.
But the received wisdom–that the most veteran teams working with people they were the most familair with would do the best–didn’t jibe with the research. Burkus continues:
Instead, Uzzi and Spiro found that the correlation wasn’t a straight line, rising in success as it rose in collaboration; the trend line looked more like an inverted U. The most successful years had a Q score around 2.6 on a scale from 1 to 5, meaning the production teams had a good mix of old colleagues and new members.
This reconfirms what Yammer and GitHub found anecdotally: that we should encourage teams to recombine. This also reaffirms something we’ve found elsewhere: that you need deep-level diversity–like along levels of experience–for a team to do its best work.