Update 7/24 at 6:30pm PT
A Harvard Business Review article about this study published on July 20th contained multiple factual errors. The way the study and its methodology were initially presented created confusion and serious concern about how user data was shared and analyzed. We want to sincerely apologize for this. One of our company values is to be worthy of trust, and to us, this means safeguarding our users’ data, first and foremost.
We’d like to emphasize that neither the Northwestern Institute on Complex Systems (NICO) nor Dropbox researchers ever had access to any user content. Dropbox anonymized any data that was shared with NICO.
Here’s our statement regarding this study:
In 2016, Dropbox formally enlisted NICO to help study shared folder collaboration at universities in order to better understand how effective research teams work. The study used anonymized data from May 2015 to May 2017, consisting of aggregated sharing activity data as well as publicly available information provided by NICO that resulted in a dataset of 16,000 researchers. At no time did NICO or Dropbox researchers have access to any user content.
What’s the secret to a high-performing team? A star player? Veteran experience? In a joint study by Dropbox and the Northwestern Institute on Complex Systems (NICO), we set out to answer questions like these by analyzing Dropbox collaboration at the top 100 universities in the world (based on the 2017 Center for World Universities Rankings) and cross-referencing academic citations according to the Web of Science database. To protect our users’ privacy, all data was anonymized and information like university ranks and number of citations were grouped into ranges.
For our study, we defined a “team” as a group of researchers collaborating within the same Dropbox shared folder. What sort of actions—like adds, edits, and number of collaborators—were typical of the “highest performing” teams? We measured team performance using two metrics: university rank and number of citations among the users. Teams with higher university ranks and higher numbers of citations among team members were deemed to be higher performing overall.
1. The highest performing teams have a mix of both old and new collaborators
Unsurprisingly, when the team members had never worked together in the same Dropbox shared folder before, the team wasn’t high performing. But the same was true of overly familiar teams: When every team member had already worked together in a past shared folder, performance was not necessarily higher. The data suggests that a mix of the two is the best choice—for example, a duo of familiar collaborators matched with two fresh faces.
2. Smaller teams perform better
High-performing teams tended to be between three and five people. Lower performing teams were larger on average and correlated with fewer citations and lower university rankings.
3. Equality of effort is key
In high-performing teams every member contributed a similar amount—adding, editing, and sharing files at similar rates. Teams where certain members did significantly more than others tended to be lower performing. This suggests that, generally speaking, a star player can’t overcome the deficiencies of a less motivated team. This also might help explain why larger teams tended to do worse: Teams with more freeloaders tended to be correlated with lower results.
4. Some experience is critical
The highest performing teams had at least one team member with a strong track record—in this case, a history of highly cited work. Teams with less established researchers underperformed in comparison. That said, there was no statistical advantage in teams with all veterans, suggesting that teams with mostly inexperienced members can still do well, so long as they have access to one or two experienced collaborators.
5. Great teams take more time
Among teams that checked all the boxes above, there was a final factor that separated the good from the great. Specifically, the best teams took a bit more time within shared folders—sharing, editing, and adding for a longer period of time than most other teams. This suggests if your team already has good chemistry and a strong sense of motivation, it’s probably worth sweating the details one last time.
Note that these takeaways apply specifically to teams at these universities. It’s possible that we’d see different results in other industries or settings. Secondly, keep in mind that these trends reflect correlation, rather than causation. It’s likely that some of these takeaways were influenced by other factors not present in the study. Still, at least some of these trends may hold for many team-based research projects—something you might keep in mind for your next company project or presentation.