What Ethnography Actually Demands of an IS Researcher

There is a version of "qualitative research" in IS that means: I did twenty interviews, I coded them with NVivo, and I themed them into three categories. That is not nothing. But it is not ethnography. Ethnography is something considerably more demanding, and I think the word gets used loosely enough that IS researchers sometimes claim its credibility without doing its work.

The anthropological tradition that ethnography comes from requires the researcher to be present in a setting for long enough to understand it from the inside. Not from the outside looking in. Not from a protocol-driven interview with a tape recorder. From sustained, immersive presence that lets you see the gap between what people say they do and what they actually do. That gap is, very often, the most interesting thing in the room. And it only shows up if you stay long enough.

In IS research, this means going into organizations. It means watching how people actually use systems, the workarounds they have invented, the things they never tell the help desk, the shortcuts that only work because someone who left four years ago set up a macro nobody has touched since. Surveys ask people to report their behavior. Ethnography watches it. Those are different research objects, and they produce different kinds of knowledge.

Wanda Orlikowski's early work brought this kind of thinking into IS with real rigor. Her studies of technology in organizational contexts took seriously the idea that technology use is enacted, not fixed. People do not simply adopt a system and use it as designed. They appropriate it, adapt it, work around it, and through that use they produce something that is neither what the designers intended nor what the users planned. My reading of her work on Lotus Notes in law firms, which appears frequently in the IS research methods literature, is exactly this: she went into the organizations, she observed, and what she found was that the same groupware system produced radically different practices in different firms. Not because the software was different. Because the organizational context, the existing work practices, the management expectations, and the norms around collaboration were different. No survey would have surfaced that finding. It required presence.

That is the core argument for ethnography in IS. Surveys tell you what people think they do. Interviews tell you what people are willing to say they do. Ethnography tells you what people actually do. And in IS, where technology use is deeply embedded in organizational practice, the distance between those three things can be enormous.

But ethnography is genuinely hard to do in IS, and I want to be honest about why. The time commitment alone is a structural problem. A proper ethnographic study takes months, sometimes longer than a year. PhD students working on dissertations and junior faculty trying to publish before tenure review are not well served by a method that requires that much fieldwork before you can write a word of analysis. The practical reality is that top IS journals still favor large-sample survey studies with clean measurement models and path coefficients. Ethnographic work is harder to evaluate with standard validity criteria, takes longer to review, and produces findings that are difficult to generalize statistically. That is a real barrier, not a pretend one.

There is also the IRB problem. Ethnographic research requires access, and sustained organizational access is not easy to negotiate. You have to convince a real organization to let a researcher embed with their employees, observe their systems, and take notes on their work practices. Organizations are often skeptical. They worry about confidentiality, about what the researcher will publish, about what employees might say. And IRB review for observational fieldwork can be complicated, especially when the researcher moves through multiple departments and speaks to many people over a long period. The ethical issues around informed consent in naturalistic observation are not trivial.

And then there is researcher bias. The ethnographer is not a neutral instrument. She brings her own theoretical commitments, her own organizational experiences, and her own interpretive habits into the field. She notices what her framework trains her to notice. She is moved by what resonates with her. Klein and Myers (1999) addressed this directly in their principles for interpretive IS field research: the principle of interaction between researcher and subjects acknowledges that the fieldwork is a social relationship, not a data extraction exercise. What the participants say and do is shaped partly by the presence of the researcher, and the researcher's interpretation is shaped by the relationship she builds with them. This is not a flaw to eliminate. It is a feature of the method that has to be acknowledged and managed through reflexivity.

None of this means ethnography is not worth doing. The argument for it is precisely that the findings it produces cannot be replicated through other means. When you do ethnographic research well, you produce knowledge about the texture of practice, the specificity of context, and the complexity of technology use that simply cannot be compressed into a five-point Likert scale. You find things that no hypothesis-testing study would have been designed to discover, because they were not on anyone's prior agenda. The best ethnographic IS research I know of tells you something that changes how you see an entire class of phenomenon, not just the setting you studied.

The challenge for IS as a field is whether it has the institutional patience for that kind of knowledge. Journal review processes are not designed for it. Citation counts reward breadth over depth. The comparative advantage of a large-n survey study, in terms of publication speed and generalizability claims, is real and rational for individual researchers trying to build careers. Ethnography asks researchers to accept a much slower feedback loop, a longer time to publication, and findings that are harder to quantify.

I think about this when I read papers that cite "ethnographic observation" in their methods section and then describe two weeks of site visits. That is not ethnographic in the sense Orlikowski or the anthropological tradition would recognize. It might be valid field research. It might produce useful data. But calling it ethnography without doing the immersive sustained work is borrowing a word's credibility without earning what that word means. The distinction matters because how we name our methods tells our readers how to evaluate what we have done. And if IS researchers routinely call brief fieldwork ethnographic, we lose the ability to distinguish between work that earns genuinely deep, contextual knowledge and work that is just survey research with site visits.

The deeper question, and the one I keep coming back to, is what kinds of IS phenomena actually require ethnography to understand. Some things really can be studied with surveys. If the phenomenon is stable, the constructs are well-defined, the causal mechanism is theorized in advance, and variation in outcomes is what interests you, a well-designed survey is the right tool. But when the phenomenon is emergent, when you do not yet know what the relevant constructs are, when practice is what you want to understand and not just attitudes toward it, ethnography is not a methodological preference. It is the appropriate response to the research question.