Advancing the Value of Ethnography

The De-skilling of Ethnographic Labor: Signs of an Emerging Predicament


Download PDF

Cite this article:

Ethnographic Praxis in Industry Conference Proceedings 2009, pp. 41–49. © American Anthropological Association, some rights reserved.

An oft-stated rule in design and engineering is, “Good, fast, cheap: pick two”. The success of ethnography in business has forced this rule into action with a vengeance. As a result, ethnographers now face a threat experienced by many categories of worker over the past two centuries: job de-skilling. Some mechanisms of de-skilling in business-world ethnography are reviewed, including:

  • simplifications that invert the conventional depth-vs.-breadth balance of ethnographic knowledge;
  • standardizations that permit research to be distributed among workers of varying cost;
  • the rise of ethnographic piecework suppliers who rely on pools of underemployed social scientists.

I argue that pressures leading in this direction must be contested, and that only by altering the cost-time-quality paradigm that controls our work can we restore its value to our employers and clients.

Manufactures prosper most where the mind is least consulted, and where the workshop may, without any great effort of imagination, be considered an engine, the parts of which are men…. and thinking itself, in this age of separations, may become a peculiar craft.

Adam Ferguson, An Essay on the History of Civil Society: 327,330.


Each of us at this conference is living proof that the demand for ethnographic knowledge is at an all-time high. From the many backgrounds that we represent, the knowledge we generate is being disseminated across ever-wider fields of activity. That is a good thing. But increased demand exerts a powerful force, tending to transform any production resource into a commodity. That force applies even when the resource in question is an intangible one like knowledge. This in turn has serious implications for the way we do our work, and my talk today will review some warning signs of a particular problem known to labor economists as de-skilling.

For a long time, the status of ethnographic labor in the business world benefited from the relative naïveté of many of our clients, which allowed us to behave like artisans with secret skills—like a pre-capitalist European guild or a traditional craft union at the peak of the industrial age. Our work was rooted in our professional traditions, passed on through various kinds of apprenticeship, and was essentially under our direct control. But as more businesses want more of what we offer, they need to more tightly integrate us with the control systems that underpin all modern companies. Our work needs to be more predictable, more standardized, have tighter linkage to ROI, faster throughput, and quicker turnaround. One client put it to me this way: “How can I get 80% of the value of a full-blown ethnographic study, for 60% of the cost, in about half the time?”

Those numbers tell a familiar story. In fact, most of business history can be summarized in that one sentence. Companies must continually ratchet up their resource efficiency or they will lose the game. When the resource in question is people—us, in this case—I have observed that there are three strategies for achieving this:

  1. Increase the number of workers, which may include simplifying the definition of a qualified worker to make it easier to be one—in other words, expand the labor pool.
  2. Require workers to work longer and/or faster for the same amount of money— speed up the labor pool.
  3. Increase efficiency by transforming an “artisanal” and self-directed work process into a fully rationalized one—de-skill the labor pool.

Option one is not something the business world has much control over yet, although it constantly tries by accepting the vaguest possible definition of what the work consists of and who can or should do it. Option two is self-explanatory: I suspect all of us have had to work longer hours and more intensively in recent years, without always seeing our salaries go up correspondingly. Option three is a bit more complicated, and it is what I want to focus on with examples from my own experience.

But first, a definition of de-skilling. Here’s how to de-skill any job: First, you break it down into pieces, and parcel those pieces out among multiple actors who get paid variable rates based on the minimum expertise needed to do their part. Then you remove human autonomy and variability by making people work according to standard routines. To the degree possible, you take the humans out altogether and replace them with machines (of the hard or soft variety). If you can extract the creative mental elements from the job and turn them into private property, you do that too.

Despite my alarmist description, de-skilling can be beneficial. By embedding human knowledge into machines and routines, de-skilling allows its opposite, namely re-skilling. People can concentrate on higher-order activities and leave the boring work to the robots. But at the same time it usually has negative impact on the people most directly affected. For one thing, average wages go down. For another, people tend to lose their workplace autonomy—how they do what they do, and how they relate to their co-workers who may be doing some other portion of the work. As usual, Marx has an appropriate bon mot: “In order to make the collective laborer, and through him capital, rich in social productive power, each laborer must be made poor in individual productive powers” (Marx 1967 [1887]:361). In factories, the enforcer of this regime may be a time and motion specialist who determines the optimal sequence of steps for a job. There still are such people, and there’s a steady business in time and motion analysis software. For business ethnographers, it is done by a more subtle set of forces intrinsic to the way the global economy works, and for the most part we do it to ourselves.


A few years ago, together with two colleagues at a leading market research company, I helped create a product called HomeView®, in which a database of ten thousand photographs of domestic interiors in twelve countries was linked to a global consumer survey. It was designed as an open-ended creative tool for generating connections and hypotheses, a hybrid device that blended some of the benefits of qualitative material culture research with the rigor and certitude of quantitative studies. To work as intended, we had to first analyze the contents of all ten thousand images, making judgments about things like people’s propensity to create collections of objects, or the degree of intentionality in how their spaces are decorated, or the presence of a “node” where activities intersect, and then tagging those judgments as meta-data. Complexly interrelated criteria for the analysis of material culture had to become multiple choice entries in a coding interface.

Coding each photo took at least ten minutes, if you worked fast and somewhat unthinkingly. Ten minutes times ten thousand is sixteen hundred and sixty-six hours, so to be cost effective, we planned to send the job to our company’s outsourcing partner in India, where it would be done by general-purpose research workers paid far less than my colleagues and me, and minimally trained to produce fairly consistent interpretations (we hoped) of what the photos were telling them about hundreds of brands and products in situ.

Set aside for a moment the interesting epistemological questions this raises. From a purely labor standpoint, what is happening? Let’s go back to that checklist of prerequisites for job de-skilling:

  • Break the job down into pieces: check.
  • Parcel the pieces out among variably paid actors, ensuring the cheapest possible worker is doing each task: check.
  • Transfer parts of human activity to a machine system: check.
  • Extract the creative mental elements and privatize them: check.

Numerous tools exist to codify and organize visual material for ethnographic analysis. What is different here is that, to my knowledge, HomeView® was the first such tool to be predicated on de-skilling the labor that produced its value. This was an unavoidable outcome of our need to give it a viable business model in addition to just making it functional. Part of the traditional skill of analyzing visual culture becomes a piece of clever software, a number of people wind up with the fairly low-paid, stressful and mind-numbing task of being image coders—with unpredictable levels of accuracy along the way—and a small number of us are thereby granted the opportunity to concentrate on higher-order analytic activities.

One of my main worries is the implication this has for creating a multi-tiered labor force with permanently asymmetrical access to money and opportunities. But even for those who don’t care about that issue, there is reason for concern in terms of the covert stultifying effects that such a labor regime can have on the enterprise. Arguably, a key contribution of ethnographers in industry is that they create the conditions for corporate learning. That is a goal that most companies at least pay lip service to, often explicitly. But in a study of why corporations fail, University of Southern California business professor Jonathan Klein took aim at job de-skilling, warning that it

in many respects mimics rather than applies the learning process. In it purest sense, learning provides a new, expanded way of looking at old things. De-skilling is the reverse: it simply narrows the application of current skills…. Hence, de-skilling undermines the very efficiency for which it is intended (Klein 2000:78, 80).

You can surmise why this is so in the case of HomeView®: once it was congealed into software and a set of database entries, this body of knowledge and its mode of interpretation become more resistant to evolution and revision, as well as to contestation and dispute. The novel connections we envisioned would be at least partly offset over the long term—undermined, in Klein’s words—by the distortion of workers’ relationship to the information they are manipulating in their various ways. The fixed role of the image coders, and their structural position as outsourced contractors, places unknowable constraints on the output of their work, and leaves them in no position to ever “get in front of” the data as its masters. Meanwhile, the hopefully masterful role of the ethnographic analysts back at the home office is mediated through the work already done by those coders, again with unknowable effects. The decision to use a de-skilling strategy thus places in question the organization’s ability to maximize the payoff of its new tool, even though at first sight that strategy may have looked like money in the bank.


Here is another example from a different angle. Historically, after artisanal production is de-skilled, certain tasks are replaced by piecework—people get paid per unit of output regardless of how much time they have to spend on it. In corporate ethnography, a small number of piecework brokers have come into existence as hiring halls for the reserve corps of under-employed social science grad students and excess Ph.D.s. What they do is different in kind from the fixed-fee arrangements that are common for many practicing ethnographers, as the unit of production in piecework is no longer the entire project but a tiny sub-unit of it. I have been on both sides of the piecework system—as buyer and seller—and have learned how it contributes to the de-skilling of labor even while it provides for cost-effective production of a certain type of information. Field researchers in this system receive a fixed fee for a specified number of home visits, observations, or whatever their task may be, suitably recorded in standardized templates. They rarely find out who the client is or what their goals are, and often don’t get to see the final results: they are alienated from the context of their work and restricted to knowing only a minor piece of the process. In their modest corner of each project, if they spend the time it really takes, they often wind up earning a fairly low hourly wage. In other words, by doing their job well, they lower its time value. Their choice is between being a “good” researcher and being a “well-paid” researcher. It becomes hard to have it both ways, because the piece-rates are generally set with an eye on creating the most attractive budget for the client.

At the same time, the reporting templates—the bridge from data to interpretation—are essentially textual machines that shape data into uniform ethnographic propositions. Like the coding software for those ten thousand photographs mentioned earlier, ethnographic piecework templates always exemplify what Marxian economists call “dead labor”, which is found wherever the source of value has been removed from people and transferred into artifacts which embody the accumulated but now-static creative ability of previous generations of workers. With dead labor in a key position in the value chain, value is generated faster and more predictably, but the amount of it that is contributed by the living worker decreases to the extent that the amount attributable to the machinery increases. So ethnographic piecework not only intensifies the fragmentation of the work process but also reduces the relative contribution of the researcher—and thus provides continual justification for lowering piece-wages. Once this becomes possible it becomes probable, given the cost-driven logic of labor allocation decisions. An example of what can happen, from a different industry, is the 2008 reduction in pageview payments by Gawker Media’s owners to its freelance writers (Avent 2008; Golson 2008). Pageview payments are an Internet version of a piecework system, and the ease with which such payments can be calibrated so as to force ever-greater effort for the same amount of money provides a foretaste of what ethnographic workers might encounter in the near future.


Piecework is just one dimension of the simultaneous rationalization and simplification of our work that I am sure we have all noticed, as we ratchet down our definition of what we are doing so it matches the level of complexity that our clients are ready to absorb. We are de-skilling ourselves. For those who have studied de-skilling in other contexts, this comes as no surprise. Production methods and production outputs co-evolve. Gene Rochlin, a physicist who writes about automation, summed it up in his analysis of factory production in the 20th century:

In the classic mass-production plant, increased production was achieved not only by dividing and specializing tasks, but by preprocessing away into the mechanisms of control much of the information contained in the final products…. items to be manufactured were therefore increasingly selected, and designed, according to the ease with which they could be subjected to the new techniques of mass production (Rochlin 1997:ch.4, paragraph 18).

If we consider ethnographic praxis in industry to be the manufacture of ethnographic knowledge, then the analogy holds. When the logic of business requires us to do our work ever-faster and ever-cheaper, that justifies calibrating the goals of the research so they are the goals most readily produced by the fastest, cheapest methods. Many of us have observed that we can’t always do the kind of work we think the situation calls for: what we are noticing is the tandem downshifting of our work processes and the type of information our clients are capable of desiring, given their constraints.

Techniques for mass-producing ethnographic knowledge include things like HomeView® and the piecework system. The goal of de-skilled information production is also aided by disintermediating technologies that create a simulacrum of identity with the consumer’s point of view. The benefit here is that data interpretation—a bulwark of ethnographic craft skill—becomes apparently unnecessary. An example is the spy-camera eyeglasses that a French market research firm uses in one of its proprietary methods (PLM Marketing n.d.). With a video camera hidden in the nose-piece, the glasses allow one to experience the wearer’s visual reality, remotely and in real time. Put the glasses on a selected consumer, and a new kind of armchair ethnography becomes possible—a return to the discipline’s 19th century roots, but now shaped by the needs of harried business managers. It’s time-consuming and costly to let ethnographers analyze ethnographic data that have already taken a long time to collect. No wonder clients prefer to “be constantly wading ankle-deep in data”, as my colleague Simon Pulman-Jones once expressed it to me. The immediacy of direct experience substitutes nicely for the unacceptable time-sink of analysis. Thus, instead of helping clients interpret complex data about complex situations, I am increasingly asked to produce an experience of getting to know consumers and end-users on a pre-analytic level that looks and feels new, but which must, for reasons of business efficiency, dovetail as much as possible with existing ways of conceptualizing those consumers.

One can think of this as the postmodern condition coming back to haunt us in a perverse way. Recall Lyotard’s keywords from his 1979 report on knowledge: “I define postmodern as incredulity towards metanarratives….The grand narrative has lost its credibility…. Where, after the metanarratives, can legitimacy reside?” (Lyotard 1984:xxv, 37). Analysis produces metanarratives, and the fact is, some clients are not very interested in my ethnological metanarratives anymore. It may be for lack of time or lack of perceived relevance, or both, but clients can be thoroughly postmodern in Lyotard’s sense. I would speculate that corporate pressure toward a de-skilled ethnographic workforce comes as much from this as from the financial considerations. Together, the two are synergistic and mutually justifying.


I have briefly described some evidence that the commodities called “ethnographic labor” and “ethnographic knowledge” are in such high demand that the business world would have us stretch them to, and beyond, the limits of efficiency. I focused on the strategy of labor de-skilling, and suggested that this results logically from actions that align our work with business imperatives. I also observed that while de-skilling is a normal way business evolves, it is not always good for those who do the work, because it forces down the market value of their labor, distorts the flow of knowledge, and alters work practices in a self-reinforcing symbiosis with the demand for mass-produceable output.

The anecdotes that illustrated my points are rare occurrences at present, but they are not random occurrences. If history is a guide, they are early signs of a process that is likely to intensify. At the final stage of that process, ethnographic praxis will be something other than what it is today. Ethnography transformed into a factory system will probably not be as rewarding for many of its practitioners, and possibly less valuable to the businesses we serve.

I’m sure we have all encountered and contemplated the meaning of the well-known design and engineering Project Triangle:

Good, Fast, Cheap: pick two.

One way of summarizing the point of my discussion is to ask what is happening to “Good”? Here are a few questions to sharpen that point:

  • Can ethnographers working in industry artificially restrict the ethnographic labor supply, as a way to boost its market value? If so, how, and will that give us a stronger platform for demanding improvements in the way we do our work?
  • What happens to the emerging community of ethnographic workers when it is stratified by structures like outsourcing schemes and piecework systems, which enforce multi-tiered wages and differential access to information?
  • What happens to ethnographic praxis when the only outputs that can be envisioned are ones that can be produced in a regime of fragmented and partially de-skilled work?

These are classic labor-management questions without a classic labor-management solution. In the community represented by the EPIC meetings, some people earn their livelihood by selling their ethnographic labor to others, some hire those ethnographic laborers, and some, like me, move back and forth between buying and selling ethnographic labor, or do both at the same time. That is just one of several factors that make it unproductive to say, “Okay let’s organize the Ethnographic Workers of the World into one big union and start making demands on our employers”. That would be easy compared to what we actually need to do. We might turn for guidance to someone like Lewis Mumford, the humanistic historian and critic who anticipated many current debates about work, technology and the environment. He argued that unless the quality of working life is explicitly addressed as an aspect of the production process, economic pressures will always force its degradation in countless ways (Mumford 1964:1970). That is what is happening to us. Our mission is to help make goods and services that improve other people’s lives. Ironically, the trend toward de-skilling promises to push our own lives in the opposite direction. So when I say let’s restore a little more “Good” to the equation, I mean it with respect to the conditions of our own work as well as to the outputs we produce, which usually still manage to be worthwhile.

This means our work needs to take more time and cost more money. You may ask how I can be so naïve as to believe that we can convince our companies and clients of the wisdom of this position. But we have precedents for optimism. Certain business practices that are widely accepted today seemed almost laughable only a few years ago. Think of the tectonic shift toward bottom-of-the-pyramid consumer strategies. Think about recent corporate commitments to carbon neutrality, sustainability and the development of less harmful energy sources. What they have in common is that they recognize there are ultimate physical limits to growth, and that our global economy needs to evolve its way out of the expand-at-all-costs phase of its existence. Fast and Cheap without Good simply reveals the labor dimension of the trajectory that sends us straight toward that brick wall of inescapable limits.

Most of us think of ourselves as humanists. If we take our humanistic mission seriously, we need to insist that the principles we apply in our research should be turned around and applied to the conditions of our own labor. And when I say “our” labor, I mean all of us whose work touches on the ethnographic enterprise in any way. Otherwise, the work is not worth doing, and we should consider laying down our tools.

Gerald Lombardi is a director of the healthcare practice at Hall & Partners, a brand and communications research agency. Prior to that he was an experience modeler at Sapient, then co-founded the ethnographic practice at GfK, a global market research firm, and became the practice’s Vice President for North America. He has been a private sector anthropologist since 1991, when still a graduate student. His Ph.D. is in cultural and linguistic anthropology from New York University.


Ferguson, Adam
1819 [1767] An Essay on the History of Civil Society. 8th edition. Philadelphia: A. Finley.

Klein, Jonathan I.
2000 Corporate Failure by Design: why organizations are built to fail. Westport CT: Greenwood Publishing.

Lyotard, Jean-François
1984 [1979] The Postmodern Condition: a report on knowledge. Geoff Bennington and Brian Massumi, trans. Minneapolis: University of Minnesota Press.

Marx, Karl
1967 [1887] Capital: a Critique of Political Economy. Vol. 1: a critical analysis of capitalist production. Samuel Moore and Edward Aveling, trans. New York: International Publishers.

Mumford, Lewis
1964 Authoritarian and Democratic Technics. Technology and Culture 5(1): 1-8.
1970 Reflections: The Megamachine III. The New Yorker, Oct. 24: 55–127.

Web resources

Avent, Ryan
2008 Blogonomics: Valleywag’s pay. Electronic document,, accessed 10 July 2009.

Golson, Jordan
2008 It’s April 1 and I don’t know what my salary is. Electronic document,, accessed 10 July 2009.

PLM Marketing
n.d., accessed 20 July 2009.

Rochlin, Gene
1997 Trapped in the Net: the unanticipated consequences of computerization. Princeton NJ: Princeton University Press. Electronic document,, accessed 14 July 2009.