Physicists tell us entropy is the natural state of the world, and that law seems especially true in today's multidivisional company.
"When you create organizational subunits of any form, they'll have a tendency to focus internally on their own things," says Toby E. Stuart, the Charles Edward Wilson Professor of Business Administration at Harvard Business School. "Interdepartmental coordination doesn't tend to happen organically. It needs some intervention to create collaborative networks."
What corporate-level strategy is about is creating a portfolio that is more valuable than the individual components of the portfolio.
In other words, it's a classic case of the left hand not knowing what the right hand is doing—each unit is trying to maximize its own performance, generally without much regard to what will increase value for the company as a whole. Enter corporate headquarters. Theoretically, it's the job of the executives upstairs to coordinate the activities of the various arms of a company to row in the same direction.
"What corporate-level strategy is about is creating a portfolio that is more valuable than the individual components of the portfolio," says Stuart. "All that is premised on the coordinating role of the corporate headquarters and corporate staff positions."
The key word is theoretically. As Stuart explains, there is actually very little empirical data about how a company's corporate staff members perform that coordinating function. "We have plenty of case studies, but given how important it is, there is remarkably little academic research that looks at the headquarters of a multiunit firm," he says. In fact, so mysterious is the inner functioning of corporate HQ that Stuart calls it a virtual "black box."
Peering Into The Black Box
In a new working paper written with Adam M. Kleinbaum (HBS DBA'08) of Dartmouth's Tuck School of Business, titled Inside the Black Box of Corporate Staff: Social Networks and the Implementation of Corporate Strategy, the researchers set out to peel back the cover of that box a little bit to shed light on how corporate staff members perform that coordinating function.
To do that, the researchers decided to focus on the communication networks between corporate staff members who direct the various divisions of the company, looking to see how they compare to the networks of line workers within those divisions. If theory is correct, corporate staff networks should be larger and more centralized than those of line workers.
Stuart and Kleinbaum hit upon the idea of using workers' e-mail networks, creating a real-time, virtual spider web of contacts based on who communicates with whom. A large, multidivisional information technology company (referred to in the paper as "BigCo") agreed to let the researchers have access to internal e-mails among more than 30,000 employees over two quarterly periods in 2006 and 2008. E-mails were stripped of content before the researchers saw them, but retained the addresses and positions of the recipients in order to track how the network was organized. Compared with more traditional research methods such as surveys, the data was relatively easy to obtain. "There is a remarkable wealth of e-mail data that is sitting on almost every company's servers, and it is inexpensive to extract," says Stuart. "All it involves is someone to write a relatively simple set of scripts. It's a surprisingly small favor to supply data like this."
In the case of BigCo, the company had two corporate staff units: a corporate headquarters and a corporate sales force. When the researchers applied a complex statistical analysis to the e-mail data, they found exactly what theories suggest: corporate staff had broader networks that reached farther and were more centralized than those in line positions. Not only did staff have more contacts, but they also were more likely to be brokers between individuals in discrete operations within the company as a whole.
The bigger question was why that was the case. The researchers had three hypotheses. One, individuals who take corporate staff positions might develop more centralized networks just as a function of the jobs they have. Two, employees who already have larger networks might be tapped for corporate staff jobs because of their networks. Three, employees who don't develop larger networks might not last in those positions and be weeded out over time.
The Network Builders
When Stuart and Kleinbaum crunched the numbers, they found little evidence to support the last hypothesis, but the first two hypotheses were borne out, with the second slightly stronger than the first.
In other words, those taking corporate staff positions tend to already have larger, more centralized networks, which then went on to broaden in their new roles. "BigCo appears to be doing a good job both in who it puts in these jobs and how it designs them," says Stuart. "What the company doesn't seem to be doing is removing people who don't develop those networks very quickly."
The more surprising thing about the data, however, is just how quickly networks changed once someone moved from a corporate staff position to a line position or vice versa.
The researchers assumed that workers would retain most of their networks over time, only changing them gradually. Instead, they found rapid turnover in much of the employees' networks when there was a change in job functions. This was especially the case when individuals/workers moved from a line to a corporate staff position. If workers returned to the line, they tended to retain some of their centralized network, reflecting a bit of the "corporate imprimatur."
Stuart cautions that these findings only apply to BigCo and that the picture might be different in other companies. But at least in BigCo, the data bears out the theory that corporate staff members have larger, more centralized networks, even if those networks are more malleable than previously thought.
The insights provide valuable information on what is really happening inside that mysterious black box. "For many, many years there have been theories about what a corporate headquarters does and why it does it," says Stuart. "[But] when we have theories about how something works but we have no data, we don't really know."
