Ethan S. Bernstein – Harvard University
Curtis Chan – Harvard University
Stefan Beljean – Harvard University
Article link: http://asq.sagepub.com/content/57/2/181
Question 1. A striking aspect of your project is the “nearly unlimited access” that you had to your field site of a Chinese mobile phone factory for PrecisionMobile (p. 185). You describe the process by which you first gained access to PrecisionMobile in terms of network connections with board members and executives. But can you give us more detail about how you managed your relation with your field site contacts to gain such wide access—especially at the level where you could gain permission from management to conduct a field experiment? For example: how did you frame your research project to upper management, was there any initial resistance from managers, and how did you deal with any such resistance or other roadblocks? Also, what were the reactions of the upper-level management to the results of your study, and how did you speak to these reactions, if applicable?
As scholars, we can easily fall into the trap of becoming so focused on the phenomena of study that we lose track of the needs of those who are generously giving us access to their work environment and, often, themselves. That is only made worse by the fact that our pace, approach, and even language can be very different from the pace, approach, and language of the executives who open the door to us.
First impressions were important. On November 15, 2007, I met for the first time with a board member at Precision. I already had a draft research proposal in hand. We discussed it, revised it together, and by December 4, we had a short, three-page document that he felt comfortable sharing with senior executives at the company and could stand on its own as it circulated. The first words in the title were “research proposal,” but the summary, project overview, hypotheses, intended results, and benefits from participation were written with management in mind—asking questions, framing ideas, and stating hypotheses from their perspective. I was clear about our objectives (advancement of social science through scholarly publication) but also clear in recognizing their interests (learning and productivity). The project was designed for rigorous scholarship but framed in the language, norms, and interests of PrecisionMobile executives, both in the short proposal and in the Powerpoint which ultimately accompanied it. This was, by no means, a core initiative at PrecisionMobile (indeed, it was highly peripheral), but at least senior executives knew enough about it not to be concerned. They also had clarity on what we intended to do: a Phase 0 (initial visit), Phase 1 (qualitative research), and Phase 2 (field experiment) approach, which not only set expectations on timing but also envisioned check-ins with management that were frequent enough to satisfy their organization’s norms of fast-paced work. My years in management consulting at the Boston Consulting Group (BCG) were excellent training for this aspect of academic field research, and I drew on that training heavily.
On December 19, the board member spoke with one of the most senior executives at the company, who was highly supportive and said that the company is “full of research projects” for this kind of work. In perhaps the greatest moment of luck in this project, he put us in touch with a rising-star VP within PrecisionMobile who was everything a researcher could hope for in a partner-sponsor. The hardest part of confidentiality in this project is not being able to publicly thank him by name. Years after the work (and several promotions later), he remains a thought partner.
All of that preparation at the senior executive level was enough to lay the groundwork for the participant-observation phase of the research (especially given how, by design, only three people at the site were supposed to know anything about the research at that stage). As you point out in your question, however, the field experiment required a new level of signoff. To make matters worse, our rising-star VP had risen into a new position that was no longer in the PrecisionMobile business, thus putting all of our access at risk.
The next step was a stroke of genius by Professor Willy Shih at HBS. Together, we wrote an HBS case on PrecisionMobile to be taught in the Building and Sustaining Successful Enterprises second-year MBA course at HBS. In relatively few pages, it captured one core aspect of what we observed via the embedded participant-observers. It also required CEO signoff. With signoff came engagement—and a desire to address the issues raised by the case. The challenge was that, given the phenomenon (hiding behavior that was so well hidden that they didn’t know about it before our research), it was not clear what should be done about the situation. We proposed a field experiment. After coming to HBS to see the case taught for the first time, company executives agreed. It was one of the most intricate (and time-intensive) processes I have witnessed for gaining permission to run a field experiment, but it worked.
That is not the full story. In re-reading the last five paragraphs, I must offer a caveat: that clean-cut retrospective masks what felt, on a day-to-day basis, much more like climbing a brittle rock face where you’re never sure what is going to happen next . If it weren’t for unconditional support from mentors and peers, deepening relationships with thoughtful individuals at PrecisionMobile, and (quite frankly) a good dose of luck, this would have been impossible. No day in the field is “normal,” and I believe being fast-on-your-feet (while remaining true to the fundamentals of rigorous research standards) is one of the most important virtues of any field researcher trying to do this kind of work.
I still remember when, with about two hours’ notice, I was asked to provide a five-minute briefing to the CEO (of this multi-billion dollar enterprise) on why I had put up curtains in one of his most important facilities. I thought that would be the end of the experiment. But he was extremely thoughtful and interested in the science behind it. Indeed, my view is that most managers are quite interested in this kind of scholarly work if they know we aren’t just going to muck about but rather actually help their organizations get work done better.
Question 2. Your paper demonstrates and documents a fascinating causal effect of privacy on higher performance. And yet, a question that burns with both managerial and theoretical implications is under what boundary conditions this relationship between privacy and higher performance might hold, and conversely under which conditions it may not. Would you like to share any thoughts about what these boundary conditions might be?
It is very interesting to speculate on the boundary conditions for these findings. Based on the conversations I have had over the past 18 months, much of the speculation around limits to generalizability involves either industry (manufacturing) or geography (China). However, one of the tremendous benefits of having this article published in ASQ has been the large number of individuals who have reached out to me with examples of how this theory applies in their own contexts—health care, financial services, professional services, government, technology startups—across a wide array of geographies. The evidence is, of course, purely anecdotal at this point. But the emerging view is that neither industry nor geography is likely to impose barriers to generalizability.
Far more likely to be fruitful, I believe, is an evolving set of boundary conditions around the presence of enablers required for privacy from observability to be productive (and, conversely, the absence of eroders which undermine that relationship). Recall, for example, that performance remained closely tracked at PrecisionMobile all the time, even when work activity became less observable with the curtains. Were that not the case, I could imagine increased privacy yielding a very different result. Of course, track performance too precisely and one ends up back in the transparency paradox—in a recently published article on “Analytics 3.0,” Thomas Davenport writes: “Just as analytics that are intensely revealing of customer behavior have a certain ‘creepiness’ factor, overly detailed reports of employee activity can cause discomfort. In the world of Analytics 3.0, there are times we need to look away.” So an interesting set of boundary condition questions exists around how adjusting various levers of transparency, not just observability, may enhance our understanding of the paradox.
A second important boundary condition is emerging around time. Rather than asking under what conditions the relationship between privacy and higher performance holds, a better question to ask may be how long that relationship will hold in different context.
I am continuing to investigate these questions of generalizability, and I would genuinely welcome both input and collaboration from others who are equally interested in such research.
Question 3. From your qualitative data, you considered the nature of the mechanisms for how privacy may increase performance. You focused on mechanisms largely concerned with learning (e.g., productive deviance, localized experimentation, distraction avoidance, continuous improvement). Given the extensive and detailed observations conducted by your embeds, we would imagine that there was a considerable amount of rich material that you left on the figurative “cutting room floor” of the paper-writing process. Was there evidence for other potential mechanisms in your qualitative data that you found interesting, but ultimately were not the focus of this paper for particular reasons? More generally, what else did you find intriguing in the data that might be a focus of future work?
Great question! The answer is an unqualified yes—there is certainly evidence in my qualitative for elaboration on the mechanisms. Indeed, in addition to the embed data, I have a factory-floor employee survey that was completed before and after the curtain intervention (both for control and treatment lines) which is the basis, along with the transcripts, for a new paper which is currently a work in progress.
Question 4. A key empirical strength of your paper is your field experiment at PrecisionMobile. In it, you were conscientious to control for potential Hawthorne effects, and you argue that workers on the night shift were convinced that the experimental manipulation of the privacy curtain was for the day shift, suggesting that Hawthorne effects were not the driving factor behind the results (p. 195-196). What might you have done, though, if it had not been the case that the night shift workers were oblivious to the manipulation? Did you consider other general strategies for controlling for possible Hawthorne effects either by a priori design or by post hoc arguments?
Every context offers idiosyncratic assets and liabilities for research. I chose PrecisionMobile in part because of its scale for a field experiment like this one, so any intervention would have been designed a priori to control for Hawthorne effects. Because the night shift and the day shift share the same space, it was the easiest control to accomplish, but any control would have had the same feature: similar space used by different groups with different perceptions on whether they were being studied. Given the nature of the field context, one could imagine other experimental designs that would have accomplished the same goal (although none as clean as the night shift / day shift design).
In my own view, post hoc arguments are best avoided, especially when one goes to such significant efforts to design and implement a field experiment. Perhaps another good question to ask is what I would have done if I had found a Hawthorne effect. On the one hand, that would have made the effect of privacy very hard to tease out. On the other hand, in words borrowed from my phenomenal Editor in the ASQ review process, the similarities between my study and the old Hawthorne studies could be viewed as “a feature, not a bug”—and I might have had a very different paper as a result.
Question 5. Given that a large amount of work in the contemporary knowledge economy involves IT and computers, what might be the equivalent of the “privacy curtain” in the digital realm?
I do think there are interesting digital analogues. Let me start with an example. I just finished teaching an extraordinary first-year section of MBAs at HBS. About halfway through the semester, I ran into one of my students on-campus as I was carrying my 15-month-old son to daycare. Shortly thereafter, I ran into another student who mentioned my son. Apparently, the first student had sent a “GroupMe” to the section about the rendezvous—something she would “never do on Twitter” because it would be too public, but something she would do within the boundary of the 92 other individuals in her section.
The moral of the story: in the digital realm, we actually have even more ways to design (and enforce) boundaries that keep selected information “within these four walls” (at least metaphorically). When we do anything digitally, we consider how widely it will spread (and if we don’t consider it, as demonstrated by a number of notable situations, we often learn our lesson a little too late). How many times do you open an email and, even before reading the body of the note, see who is on the “To:” and “cc:” lines (not to mention wondering about the “bcc:” line) before deciding what to write in a “reply all” message—or deciding to pick up the phone and call someone to chat “offline” about something? A promising, new Silicon Valley startup is based on that premise and may provide a research opportunity to learn more about this phenomenon in the digital space over the next few years. In the meantime, one thing seems clear: just like in the physical realm, the presence and location of boundaries in digital spaces may indeed partially dictate how we perform.