Case Study Research: A State-of-the-Art Perspective
Summary and Keywords
Theory building from case studies is a research strategy that combines grounded theory building with case studies. Its purpose is to develop novel, accurate, parsimonious, and robust theory that emerges from and is grounded in data. Case research is well-suited to address “big picture” theoretical gaps and dilemmas, particularly when existing theory is inadequate. Further, this research strategy is particularly useful for answering questions of “how” through its deep and longitudinal immersion in a focal phenomenon. The process of conducting case study research includes a thorough literature review to identify an appropriate and compelling research question, a rigorous study design that involves artful theoretical sampling, rich and complete data collection from multiple sources, and a creative yet systematic grounded theory building process to analyze the cases and build emergent theory about significant phenomena. Rigorous theory building case research is fundamentally centered on strong emergent theory with precise theoretical logic and robust grounding in empirical data. Not surprisingly then, theory building case research is disproportionately represented among the most highly cited and award-winning research.
Case Study Research and Grounded Theory Building
Case study research has a long tradition within organization studies, strategy, and the social sciences broadly. Classic exemplars include Whyte’s (1943) study of Italian street gangs in Boston and Chandler’s (1962) study of the emergence of the M-form organizational design among firms like DuPont and General Motors. Moreover, the case study continues to be a central methodological approach in contemporary research.
A case study is a rich or “thick” empirical description of a particular instantiation of a phenomenon, typically with embedded (i.e., nested) levels of analysis and often relying on multiple data sources (Yin, 2017). Indeed, a signature feature of case study research is deep immersion in the phenomenon by the researcher. A second signature feature is replication logic—that is, the researcher analyzes each case as a stand-alone or unique experiment, and so does not pool cases together and use pooled logic (i.e., combine cases) as in traditional theory-testing research.1 In other words, the researcher first seeks to understand the focal research question within each case on its own, and then iteratively attempt to replicate these insights across each case. Finally, case studies can be used for a variety of research purposes along the continuum between theory building and theory testing. That said, our focus is on case studies used for theory building (Eisenhardt, 1989a), an appropriate approach when relatively little is understood about the focal phenomenon. Our comments also relate to theory-elaboration (McManus & Edmondson, 2007), a useful approach when the phenomenon is relatively more understood.
Theory building from case studies relies on grounded theory building. Grounded theory building simply refers to the process of generating theory from patterns that emerge within data (Glaser & Strauss, 1967). In other words, the researcher analyzes one or more cases to find patterns in the data that offer theoretical insights in the form of constructs, theoretical relationships among those constructs, and sometimes propositions (Eisenhardt, 1989a). While there are a variety of specific approaches to grounded theory building (e.g., Gehman, Eisenhardt, Gioia, Langley, & Corley, 2018), they all engage in the creative process of finding theoretical patterns that exist within data (Walsh, Holton, Bailyn, Fernandez, Levina, & Glaser, 2015a). The aim of theory building from case studies is to develop novel, accurate, parsimonious, and robust theory.
Theory building from cases is distinct from other approaches to theory building such as verbal theorizing, computational simulation, and mathematical modeling. A key distinction is that theory building case research rests on logic that is informed by the systematic analysis of empirical data, whereas the other approaches do not. Theory building case research is a broader approach than other inductive methods that also rely on deep immersion in data, and a grounded theory building process. These methods have a more specific focus and somewhat different assumptions. For example, interpretivist studies view knowledge as socially constructed, and prioritize the authentic representation of people’s voices and lived experiences (Gephart, 2004; Gioia, Corley, & Hamilton, 2012). These studies often explore constructs such as identity and sense-giving, and focus more on socially constructed perceptions about a particular phenomenon and less on creating generalizable theory. Ethnographies historically explore culture through immersive observation. Its interest in day-to-day practices provides insights into the actions that actors may not even know themselves (Bechky, 2011). Observing non-verbal cues, rituals, and artifacts is often central (Valentine, 2018; Van Maanan, 2011). Yet despite differences, multiple case, interpretivist, and ethnographic studies share commonalities in their use of grounded theorizing, theoretical sampling, and immersion in phenomena. This article primarily focuses on building theory from case studies (i.e., addressing when, why, and how to do so with rigor and quality).
When and Why to Use Theory Building Case Research
Theory building case research is particularly useful when existing theories are inadequate to explain the focal phenomenon (i.e., when current theory is limited, unlikely to be correct, conflicting, or simply does not exist). For example, Davis and Eisenhardt (2011) explored how established firms effectively engage in R&D collaborations. Prior work had examined the role of structural antecedents to R&D collaborations on performance, including the effects of partners of similar size and prior collaboration experience together. Yet, this research typically explained only limited variance in performance and did not effectively address the lack of success in these collaborations that many executives experienced. By emphasizing antecedent structural conditions, the extant research missed the “how” (i.e., the processes by which executives effectively (or not) create collaborative innovations). Through a multi-case theory building study of eight technology collaborations between ten organizations over a multi-year period, the authors identified the critical role of “rotating leadership” that facilitated the innovation process across organizational boundaries. By using case studies and controlling for the antecedent conditions identified in prior work, the authors unpacked the collaborative innovation process.
Theory building case research has become a popular and influential method in part because it bridges the divide between rich qualitative evidence and mainstream theory-testing research. While it involves pattern recognition to discover novel theoretical insights, it also emphasizes constructs, measures, and theoretical relationships among constructs—i.e., notions that are familiar to other types of researchers. Researchers can then empirically test this emergent theory, extend its insights using simulation to run experiments, or unpack its core insights through mathematical modeling. That said, the emergent theory can also stand on its own. In addition, theory building case research often has practical relevance that extends its influence. Finally, this method has become popular and influential because it often yields unusually novel insights. Indeed as others have observed (Bartunek, Rynes, & Ireland, 2006), articles using theory building case research tend to be more interesting, disproportionately winners of major awards (e.g., Dutton & Dukerich, 1991; Ferlie, Fitzgerald, Wood, Hawkins, Ferlie, & Hawkins, 2005; Gersick, 1988; Gilbert, 2005; Graebner, 2009), and highly cited (e.g., Baker & Nelson, 2005; Battilana & Dorado, 2010; Eisenhardt, 1989b).
Theory building case research is particularly effective for addressing research questions about “how” (i.e., process questions that can be difficult to study in other ways). In contrast, it is less effective for research questions about effect sizes and interaction terms. Theory building case research is also particularly effective for broad inquiries into “what’s novel.” Finally, theory building case research is flexible. That is, the researcher can adapt as the study progresses and address multiple, and even changing questions. As noted earlier, there is a continuum from theory building to theory elaboration (and theory testing) using case studies. For some research questions, there is relatively little prior theory and evidence. For example, Brown and Eisenhardt (1997) recognized the extensive prior research that addressed how firms develop a single product, but the contrasting absence of research on how firms develop multiple, successful products. In studying the latter, their approach was theory building. For other research questions, there is helpful extant research. For example, Gilbert (2005) built on prior research on “inertia” and used his cases to unbundle the inertia construct into its components (i.e., resource rigidity and routine rigidity). His approach was theory elaboration. Regardless of its position on this continuum, theory building case research is characteristically asking “big picture” research questions that are difficult to address with other methods.
How to Conduct Rigorous Theory Building Case Research
Rigorous theory building case research relies on a careful (and often iterative) process of developing an appropriate (i.e., fits the method) and compelling research question, designing the study including the use of theoretical sampling to select cases, collecting rich and triangulated data, and creatively analyzing the cases to find patterns, using replication logic. Like any rigorous research method, the process of theory building from case research involves both art and science. The following subsections (“Developing a Research Question,” “Conducting a Rigorous Literature Review,” “Designing the Study,” “Collecting Data,” “Analyzing Data and Building Grounded Theory,” and “Writing and Evaluating High-Quality, Rigorous Research”) provide some insights into that process and some guidelines for engaging in rigorous theory building from case studies, illustrated with examples from the literature.
Developing a Research Question
The first step in theory building from case studies is developing a well-defined and compelling research question that fits the method. As described earlier, theory building from cases is a particularly useful method for understanding a phenomenon for which existing theory is inadequate (i.e., current theory is limited, unlikely to be correct, conflicting or does not exist). While this question may change over the course of a research study and additional questions may be addressed, it is nonetheless important to begin with an initial question that fits the method. In order to find such a question, a literature review of the work related to the focal phenomenon is essential. Yet, a common error is forgetting to do a thorough literature review or doing a poor one.
Conducting a Rigorous Literature Review
To start, the researcher should read the literature very precisely in order to find gaps or dilemmas in the prior work that relate to the focal phenomenon. Theory testing best addresses incremental gaps. In contrast, large gaps fit theory building research, and, consequently the researcher should look for those gaps. For example, Graebner’s (2004) work on acquisitions began with a thorough review of the M&A literature which revealed that the buyer’s perspective was well-researched, but the seller’s perspective was almost completely neglected. This review led her to focus on the seller’s perspective, and the seller’s influence on how acquisitions occur, are negotiated, and are effectively implemented. This focus on sellers led to novel theoretical insights, including how sellers can create both expected and unexpected value for buyers in the acquisition process. A key point is that a thorough literature review not only reveals appropriate research questions but also helps to frame contributions effectively at the end of the study.
A thorough review of each stream of literature requires careful reading and evaluation of articles that are relevant to the focal phenomenon from core journals in the field from the past several years. A helpful rule of thumb is focusing on articles from the last five years. A thorough review also incorporates older, widely cited articles that are the relevant foundational work. Empirical articles are usually better because their insights are grounded in data whereas many theoretical articles are too close to opinions. If more than about 100 articles seem relevant, the researcher should probably narrow the scope of the research. It is important to read each article carefully to understand the research setting, key measures, results (if any), and any nuances of the contributions. It is usually helpful to systematically record research questions, hypotheses and their results, samples, and primary conclusions. This systematic record-keeping facilitates the synthesizing of the extant literature.
The next step is to develop an understanding of the literature by creating distinct research streams. The streams can be organized along one or several dimensions, such as dependent variables, independent variables, theoretical perspectives, points of view (e.g., the buyer or seller in acquisitions), and time period. The aim is to use the organizing dimension(s) that best give a clear picture of the literature. Yet, organizing the literature into streams is often the most challenging part of conducting a literature review. It may require several attempts that involve shifting abstraction levels to identify commonalities across seemingly disparate findings. This organization process typically reveals not only the findings, but also the relevant gaps and dilemmas in the literature. These gaps and dilemmas can then lead to one or more compelling research questions.
A crucial but often neglected step in the literature review process is actually writing the review. Writing is essential to clarify the researcher’s own understanding of the implications of prior research (i.e., both what is well-established and what is not). An effective literature review includes an introduction that argues why the broad research focus area is theoretically and practically meaningful. The core of the review is a description of each research stream, including description and illustration of selected articles, and a thorough summary of major results, strengths, weaknesses, and opportunities for future research for each stream. An effective review notes where research streams do and do not overlap. For theory building case research, the researcher also needs to move beyond incremental opportunities and instead identify major gaps and novel directions.
Designing the Study
Rigorous case study designs use theoretical sampling and often embedded designs to create variation, control, and improve generalizability. An effective research design fits both the setting and the research question. Theoretical sampling is an essential element in the design of theory building case research. In contrast to random sampling from populations used in theory testing studies, theoretical sampling emphasizes selecting cases based on their likely contribution to improved theory about the focal phenomenon. Ideally, the researcher chooses cases where the focal phenomenon is likely to occur. For example, Ozcan and Eisenhardt (2009) chose to study how entrepreneurs built an ecosystem of alliance relationships in the nascent mobile gaming industry, a setting where entrepreneur game publishers needed partnerships with key established firms (e.g., telecommunications carriers, brand owners, and software platforms) in order to survive and prosper.
Sometimes, it is helpful to recognize that the focal phenomenon may occur differently in different settings or have alternative paths and to sample accordingly. For example, Hallen and Eisenhardt (2012) studied how entrepreneurs form funding ties quickly and efficiently. A priori, they reasoned that experienced entrepreneurs in locations rich in venture capital, such as Silicon Valley, might use one process, while inexperienced entrepreneurs or entrepreneurs in other locations might use different processes. Therefore, they sampled firms from each of these conditions in order to improve the likelihood of more complete theoretical coverage.
More broadly, it is important to use theoretical sampling to improve variation, control, and generalizability. For example, Garg and Eisenhardt (2017) studied how CEOs manage their boards effectively or not. To do so, they chose cases where the focal phenomenon would likely be apparent. So, they studied first-time CEOs (to increase the likelihood of managerial variation) in ventures (to improve transparency) with Series B venture funding (to ensure meaningful board activity). In addition, they chose CEOs with similar ages and education to control for these factors that were not of theoretical interest. Further, they chose CEOs who varied in their experience (e.g., two were co-founders and two were not, and two had venture experience and two had large-corporate experience). This variation increased the generalizability of their emergent theoretical framework. Another example is Graebner’s (2004) study of M&A in which she theoretically sampled cases in three high-tech industry sectors (hardware, software, and services) to increase generalizability. Finally, Davis and Eisenhardt (2011) selected R&D collaborations which satisfied (and thus controlled for) the antecedent conditions for success (e.g., partners had worked together before). This sampling approach enabled the authors to control for known antecedents and isolate their primary interest, collaboration process. Overall, theoretical sampling is essential for theory building case research (i.e., it enables a focus on the phenomenon of interest, can control for variables that are not of theoretical interest, can add useful theoretical variation, and can improve generalizability).
Embedded design is another key feature of designing theory building case research. An embedded design refers to the use of multiple units of analysis (e.g., executives, business units, organizations) within each case (Yin, 2017). It is commonly used in theory building case research. Embedded designs afford a richer understanding of the overall phenomenon because they enable more tracking of the phenomenon. For example, Eisenhardt’s (1989b) study of strategic decision making focused primarily on the firm as the unit of analysis, but also integrated insights using individual decisions and the characteristics of specific executives and their relationships with one another as additional units of analysis. This embedded design led to a better grounded and more accurate emergent theory that related the strategic decision-making process to both the speed of strategic decision making and firm performance. In another example, Galunic and Eisenhardt (1996, 2001) studied charter changes (i.e., changes in product-market areas of responsibility) in multiple business units within a single firm. The authors used the firm as a unit of analysis as well as the business unit. Similarly, Ozcan and Santos (2015) studied ecosystem failure using the global payments industry as the setting. They studied both the global industry and country-specific markets, such as in Japan and Singapore, to develop their emergent theory about failure. Additionally, Bhakoo and Choi (2013) studied how healthcare organizations in multiple tiers, or levels, of a supply chain responded to institutional pressures. They created cases for organizations across levels in the supply chain and conducted within-level and across-level analysis. This embedded design allowed the authors to build theory about how institutional pressures influence organizations differently based on their relative positions in a system hierarchy.
The number of cases is also an important feature of study design. Research can be single-case, comparative case (i.e., two), or multiple case (i.e., two or more cases). With a single case, the researcher can convey particularly rich detail. Single case studies are especially useful for extreme or one-of-a-kind situations (e.g., what Siggelkow (2007) terms a “talking pig”; see Ozcan, Han, & Graebner, 2017). For example, Hargadon and Douglas (2001) studied the singular case of Edison’s introduction of the light bulb, while Tripsas and Gavetti (2000) took advantage of the unique case of Polaroid and its failure (despite having the capabilities) to succeed in a new era of photography.
In contrast, a multi-case design enables the researcher to understand rich details about a phenomenon while also gaining the advantages of replication logic. Replication logic involves using each case as a stand-alone experiment such that the researcher iteratively develops theory in one case, then tests the emergent theory in the others, and so repeats the process until there is a good fit between theory and data. Thus, replication logic allows comparison and contrast of findings across the cases as stand-alone experiments (Yin, 2017). A primary advantage of using multiple cases is that the researcher can more easily differentiate between consistent patterns and idiosyncratic details. Thus, replication logic typically improves the accuracy of constructs, including appropriate level of abstraction and measures, and usually leads to better grounded, more accurate, more robust, and more parsimonious theory than is possible with single cases (Graebner & Eisenhardt, 2007).
Multiple case designs typically include between two and twelve cases. The reason to limit the number of cases to twelve is that it is an approximate upper bound for the number of cases that a researcher can cognitively manage. Beyond that number, the researcher should probably move to a different analytic technique, such as qualitative comparative analysis (QCA). That said, the researcher can use subsets of the cases in different papers that address different research questions (e.g., Bingham & Davis, 2012; Bingham & Eisenhardt, 2011), especially when the omitted cases are not particularly revealing about the focal research question of the paper.
Finally, the researcher can use a two-case or comparative-case design. Sometimes, the researcher chooses cases that reinforce each other and so replicate the emergent theory in two settings. At other times, the researcher chooses cases that have different (e.g., high- versus low-performing) outcomes. A comparative-case design allows a detailed account of individual cases in the paper while maintaining replication logic across the stand-alone cases to develop the emergent theory. For example, Battilana and Dorado (2010) studied how hybrid organizations organized to balance their dual missions (e.g., financial versus social welfare). The authors studied two Bolivian microfinance firms that used markedly different hiring and socialization approaches. One firm worked toward a common identity for its financial and social welfare subgroups, and experienced internal stability and superior performance. The other firm reinforced the distinct identities of its subgroups, and experienced intractable identity conflicts and low performance. Thus, the authors were able to build theory about the organizational processes of hybrid organizations and their implications for performance. In another example, McDonald and Gao (2019) studied two ventures in the same nascent fintech market. Each reoriented its business (i.e., pivoted) after receiving funding. Yet, the leaders of the two firms diverged in how they communicated their pivots to the public and had similarly divergent outcomes. Making a “pivot without penalty” depended on how well firm leaders anticipated, justified, and staged changes for various audiences. Overall, comparative-case designs enable a compromise between rich presentation of the data and selection of unique cases versus the theory building benefits of more extensive use of replication logic.
Another important design feature is the choice between a process versus variance study. As noted above, process research questions (i.e., those that examine “how” questions and track the focal phenomenon over time) are both common and very well-suited to theory building case research. However, there is also a second meaning of process—i.e., a process study explores and emphasizes the similarities across cases with regard to the focal phenomenon, often across a variety of settings that serve to increase theoretical generalizability. Consistent with this use of the term “process,” the aim is to understand and highlight consistencies in emergent theoretical relationships and often common outcomes across cases. In contrast, variance studies seek to explore and highlight variations in either process or outcome, or both across cases.
For example, Bingham and Eisenhardt (2011) aimed to unpack the “black box” of what people learn from repeated experiences. Specifically, the authors studied what entrepreneurs learned from successive entries into new countries and purposely selected firms from three culturally-distinct countries (e.g., Finland, Singapore, and the United States) to increase the likelihood of generalizable theory beyond a single cultural context. The entire sample was twelve high tech firms (four from each country). The authors focused on the learning processes within the firms in which the executive teams learned simple rules for internationalization and were subsequently high-performing. In contrast, they omitted the cases in which executive teams learned little that was relevant across countries. They were then able to focus on unpacking the effective learning process using a process design that centered on similarity. In a later paper, Bingham and Davis (2012) then used a variance design that included executives who learned simple rules about internationalization and the poor learners who did not. Here, the authors highlighted how different learning processes lead to different outcomes.
Within variance designs, several specific designs are frequently used. One such design is polar types (i.e., cases with extreme positive or negative outcomes). This design helps to sharpen the contrast of the theoretical relationships among the relevant constructs with the focal outcome. For example, Ott and Eisenhardt (2018) used three matched pairs of firms (six in total) in different two-sided markets to unpack how executives effectively form strategies. Each pair of firms had a low performer and a high performer. The authors were able to develop an emergent theory that explained the differences in the strategy process between the high and low performers, and the similarities in the process, especially among the high performers. In another example, Eisenhardt (1989b) used eight computing firms to understand how executives make fast strategic decisions. The sample included those who made strategic decisions very slowly and those who made strategic decisions very quickly but did not include four intermediate-speed firms from the larger 12-firm sample. By focusing on polar types, she was able to draw a particularly striking contrast and provide added clarity regarding emergent theoretical relationships.
A second variance design is racing in which multiple cases begin at the same time and under similar conditions and then “race” over time to a specific outcome. For example, Hannah and Eisenhardt (2018) studied five firms that pioneered the nascent residential solar industry. The firms each began at about the same time with similar resources, such as funding. The patterns of competition and cooperation among the firms revealed distinct strategies to dominate the market, including a strategy to control bottlenecks of growth. The emergent theoretical framework indicates the strategies that were successful, and those strategies that do not work. Other examples of racing designs include Ozcan and Eisenhardt’s (2009) study of building the high-performing alliance portfolios and McDonald and Eisenhardt’s (2019) study of firms attempting to design a successful business model. In each example, firms in a nascent market raced to a particular outcome, with winners and losers. By controlling for starting positions, a racing design can sharply identify the constructs (and their logic) that lead to variance in outcomes.
Overall, rigorous case study designs rely on theoretical sampling to select cases that are likely to add to the strength of the theory by enabling replication logic, providing control of unwanted variation, and creating variance that adds to generalizability or sharpens distinctions in pathways and outcomes. Yet, while there are a variety of designs, the particular design is less important than choosing a design that supports the underlying aim of identifying theoretical relationships from the data using theoretical sampling and replication logic.
Finally, new advances highlight the surprising complementarities and similarities between theory building case research and machine learning. Case study research and machine learning approach data with few a priori assumptions and are both fundamentally concerned with pattern recognition in data. Both use techniques to avoid excess complexity and guard against overfitting. Further, both aim to provide robust, accurate, and generalizable theory (case studies) or prediction (machine learning). While case studies are particularly effective for identifying theoretical relationships among constructs in a small number of observations, case study research does not reveal effect sizes and nonlinearities. In turn, machine learning is able to provide these insights from quantitative data, but the method is limited in its ability to theorize the constructs and relationships. Recent empirical research combines theory building cases and machine learning. Tidhar and Eisenhardt (2019) used multiple cases to identify relevant constructs and theoretical mechanisms related to effective revenue models. The authors then used these constructs to seed machine learning. Machine learning then extended the theory-building insights from the case studies with non-linearities, equifinal paths, and size effects. Combining case research with machine learning offers new directions for grounded theory building research design, in particular using large data sets and in question-driven research.
Collecting relevant and accurate data is essential for theory building case research. A key point is that the researcher can be opportunistic in the sense that he or she may collect different data for one or more cases if it is helpful, even if it is not available for all cases. Additionally, the researcher may add cases and new data sources as they become apparent, available, or relevant.
A hallmark of case study research is deep immersion in the data. Thus, the data are usually rich and longitudinal, and often include a variety of sources, including common ones like interviews, observations, and archival data. In addition, newer forms of data have emerged including web archives (e.g., archive.org), online interviews (e.g., YouTube), and social media (e.g., Twitter). While the particular data that are relevant depend on the study (e.g., studies of culture often emphasize observations, histories emphasize archival sources), the semi-structured interview is often a central source. This type of interview is a particularly effective way to gather rich data from multiple informant perspectives who provide insights not obtainable through other sources.
It is often very helpful for the researcher to write an initial interview guide, and pilot test it with multiple individuals before going into the field for the focal study. Surprisingly, the most effective guides usually contain relatively few questions, perhaps less than a dozen. The aim is to use the questions to help the interviewer to guide respondents in providing chronologies of events around the focal phenomenon in their own words and with as little prompting or channeling by the interviewer as possible. In other words, the most effective interviews encourage the interviewees to describe the flow of events related to the focal phenomenon as they know it. Ideally, the researcher chooses interviewees from a variety of relevant perspectives, such as from different hierarchical levels, functions, and temporal periods. Internal and external informants are often useful to provide distinct perspectives.
It is usually effective for the researcher to emphasize open-ended questions focused on facts and events (an exception is interpretivist studies where the focus is on lived experience). One of the most effective approaches is to ask the interviewee to describe the events of the case as if telling a story. A helpful follow-up question is often simply, “What happened next?” When interviewees describe particular decisions, it can be helpful to ask what alternative options were considered and why those alternatives were not chosen (e.g., counterfactuals). Obtaining chronological facts and events throughout the interview helps to ensure completeness and aids the researcher later when writing the case. The interviewer may also ask quantitative questions, such as, “On a scale of 1–10, how would you rate the performance of this organization (or other unit of analysis)?” When asked to multiple interviewees, quantitative questions can complement qualitative data and help the researcher to understand the case better. It is usually effective for the interviewer to reserve “big-picture” questions for the end of the interview. An example of such a question is, “What was one of the best decisions made in this organization (or other unit of analysis)? What was one of the worst decisions?”
The researcher ideally conducts the interviews in temporal waves that allow capturing real-time events and changes over time. These waves of interviews also help to mitigate retrospective bias, and enable the researcher to gather new types of data as theoretical insights emerge. This approach supports critical iteration between data collection and emerging theoretical insights. Some studies rely on multiple interviews from a few informants (e.g., Graebner, 2004) while others rely on interviews from many informants (e.g., Davis & Eisenhardt, 2011). Regardless of the number of informants (which is usually related to how many people have knowledge about the focal phenomenon), the goal is to speak with the most relevant individuals to obtain rich and varied perspectives about the focal phenomenon from well-informed actors.
Observations can also be an important data source, especially when studying processes like work and day-to-day interactions (Bechky, 2011; Edmondson & McManus, 2007; Kellogg, 2011). For example, Garg and Eisenhardt’s (2017) study examined how CEOs effectively engage their boards in the strategy-making process. Since it seemed likely that board meetings played a central role in that process, an important source of data was observations of the board meetings. Combined with interviews, these observations allowed the authors to unpack processes that would have been less clear without firsthand experience in the meetings. Since the aim is deep immersion, observations can be important for seeing how processes unfold in real-time.
Archival data is another important source of data, one that often complements interview and observational data. It is particularly valuable when the study period is long. Examples include news articles, website archives, blogs, social media, and online video and audio interviews. For example, McDonald and Eisenhardt (2019) collected a wealth of data from YouTube and industry blogs about their sample firms. A recent explosion of archival data sources on the Internet provides an excellent source of easily accessible data, although it may have its own biases or even errors.
Surveys are another data source that can be useful, especially when the researcher has an a priori “hunch” about the relevance of some construct such as power or conflict. The researcher can administer a survey within interviews or separately. An advantage of surveys is that they can be given to a large number of respondents. Surveys can include both quantitative and qualitative questions. Overall, using multiple data sources allows the triangulation of data and perspectives during case analysis to get as close as possible to the facts.
Analyzing Data and Building Grounded Theory
Theory building from case studies relies on systematic, and yet creative, analysis to uncover patterns in the data and ultimately develop an accurate, robust, and novel theoretical framework from them. In other words, the researcher engages in grounded theory building. The analytic process includes within-case and cross-case analyses (for multiple case designs) to identify patterns within the data. Later stages of the analysis often involve bringing in related literature to sharpen constructs and theoretical mechanisms.
The analysis begins with writing each case. Individual cases are typically “thick” descriptions derived from deep immersion in the data related to the focal phenomenon. The researcher creates individual case histories at the level of the primary unit of analysis, with nested cases either folded into the main case or kept separate so that they can be analyzed as distinct cases. For example, Martin and Eisenhardt (2010) wrote case histories at the level of the firm, and then wrote nested cases describing each of the two collaborations within each firm in their study of how firms develop successful (or not) cross-business unit collaboration. Thus, cases are accounts (often chronological) of the broad focal phenomenon and the specific research question(s) in a particular situation.
The researcher often triangulates the data sources, frequently blending several types of data, including interviews, archival data, observations, and possibly surveys. Although there is no rule, archival data tend to be particularly helpful for addressing the facts of the case related to what, when, who, and where. Interview and observation data are particularly helpful for addressing how and why. For example, news articles and blogs can help a researcher understand which product innovations occurred in an organization and when. Interviews and observations can provide otherwise hidden insights into the processes and motivations related to those innovations. By weaving together primary (i.e., first-hand) and archival data, the researcher can arrive at a more accurate account of the case over time that includes insights about how, when, and why events unfolded in a particular way. Co-researchers working independently can corroborate the case narratives and help to identify emergent patterns in the data. Co-researchers often take separate roles in the case-writing process. For example, one researcher might write an initial draft of a case while another might read the original data to develop an independent perspective. The research team can then converge on a more accurate case, reconciling differences by returning to the data or gathering new data.
At the heart of the grounded theory process, the researcher seeks to identify patterns in the data (Walsh et al., 2015b). There are a variety of ways to undertake this creative process. Indeed, there is no cookbook for the creativity and insight of this method. Nonetheless, the theory-building process always involves using the data to identify initial patterns and then more abstract patterns (e.g., first-order concepts and second-order themes (Gioia et al., 2012), temporal bracketing (Langley, 1999), and construct and measures tables for each construct (Eisenhardt, 1989a; Eisenhardt & Graebner, 2007). Thus, deep immersion—which enables pattern identification and then recognition of theoretical relationships at appropriate abstraction levels and that are grounded in the data—is essential. In contrast, the particular approach and terminology are less germane.
In multi-case research, the researcher also conducts cross-case analysis after writing the within-case analyses. The comparison of multiple cases enables the researcher to “test” the emergent theory in each successive case using replication logic (Yin, 2017). Cases are compared and contrasted (e.g., A to B, A to C, B to C) using approaches that force the researcher to examine the data from multiple perspectives and in different combinations. For example, identifying similarities and differences across cases helps the researcher to identify relevant constructs. Tables are particularly helpful tools for summarizing measures of constructs (Miles, Huberman, & Saldaña, 2014). These construct tables can help the researcher to develop insights that lead to theoretical relationships among constructs.
The process of cross-case analysis involves iteration between data and emergent theory. Later in the grounded theory building process, the researcher often considers prior research that is either close or distant (or both) to the research question and phenomenon. This comparison with the literature sharpens constructs, theoretical mechanisms, and theoretical contributions. Thus, there is an iterative back-and-forth among data, theoretical arguments, and results from prior literature. This process helps to improve internal validity of the underlying logic of the emergent theoretical relationships. It also helps the researcher to develop an accurate, robust, and often parsimonious emergent theory.
Writing and Evaluating High-Quality, Rigorous Research
A consistent tension in writing case-based research is how much to focus on the data versus the theory. Because the researcher works within the spatial constraints given by publishers, there is typically a tradeoff between providing the data as empirical support for the emergent theoretical framework and describing that framework. This tension is particularly acute for multi-case research with its additional cases beyond a single case. For the multi-case researcher, the best way to address this tension and so balance “better stories” versus “better theories” is to frame the paper in terms of the theory and then support that the theory with empirical evidence demonstrated by at least some of the cases (Eisenhardt & Graebner, 2007). Extensive use of tables, appendices, and visual aids helps to accomplish this (e.g., Hallen & Eisenhardt, 2012; Hannah & Eisenhardt, 2018; McDonald & Gao, 2019; Pache & Santos, 2013). In contrast to multi-case studies, single case studies are less constrained. Here researchers often frame the paper as the case narrative followed by the theory. Although the researcher can use a variety of written formats, it is essential to tie supporting empirical data to individual constructs and provide the underlying theoretical logic for relationships.
Finally, it is important to understand how to write theory building case research that is high-quality and rigorous as well as to know how to evaluate this research. There are periodic calls to use specific writing formats, provide overly detailed accounts of the theory-building journey, or use a common analytic recipe. Some call for rigid replicability in an attempt to force theory-building research into the realm of theory-testing research, thus missing two key points: replication is at the heart of multi-case theory building and the aim of such research is insight into a phenomenon, not theory testing. Thus, many of these calls miss these points.
Instead, the most rigorous studies focus on theory (Eisenhardt, Graebner, & Sonenshein, 2016). They present a strong emergent theory that is internally coherent, accurate, robust, and parsimonious. This theory rests on a grounded analysis of rich and comprehensive data, an effective research design including appropriate theoretical sampling, and an intriguing research question. Each construct is grounded in well-measured and appropriate data. Effective research designs select cases using theoretical sampling where the focal phenomenon is likely. Rigorous designs control for theoretical variation that is not of interest, and help to establish generalizability. A thorough literature review to identify unanswered and novel questions and to sharpen theoretical contributions at the end of the study is essential. The best studies focus on “what’s novel” and often lead to emergent theory that tackles significant theoretical and practical phenomenon with accuracy, parsimony, and robustness. Table 1 provides an overview of the process of creating rigorous case research.
Table 1. Process of Conducting Rigorous Case Study Research
Identify an initial research question
Develop a well-defined, compelling question
Useful when existing theories are inadequate to explain focal phenomenon
Often focused on “why” or “how” questions
May address questions on a continuum from theory building to theory elaboration
Write a rigorous literature review
Organize the literature into streams and identify gaps or dilemmas
Write the review to clarify the implications of the literature gap and refine the research question
Helpful to move beyond incremental opportunities and instead identify major gaps and novel directions
Sometimes requires several attempts to get the right level of abstraction and blocking of streams
Design the study
Use theoretical sampling and replication logic to introduce variation, control, and generalizability
May focus on process or variance
Variance designs include but are not limited to polar types and racing
Collect the data
Become deeply immersed in the focal phenomenon
Gather data from multiple sources to triangulate
Typically rich and longitudinal
May include personal interviews, observations, surveys, financial records, news articles, web archives, social media, and many other sources
Analyze the data
Write individual cases
Conduct within-case analysis
Conduct cross-case analysis
Provides new constructs, the relationships among constructs, and the underlying theoretical logic of the relationships
Write the paper
Organize the paper in terms of the theory
Support the theory with evidence from the cases
Demonstrate how new theory builds on prior conversations in the literature
Often helpful to begin with tables to clarify constructs and their measures
Important to describe the theoretical logic for the relationships among constructs
An art is balancing theory and data because of page constraints
Theory building from case studies is an important research strategy because of its ability to provide rich and accurate theoretical insights into under-explored and under-theorized phenomena. Case research involves deep immersion into a focal phenomenon, and is well-suited to answer research questions about “how.” The best studies lead to new theoretical insights about “big picture” research questions that address significant theoretical gaps and dilemmas. Not surprisingly then, theory building case research is disproportionately recognized as some of the most cited and award-winning work.
Baker, T., & Nelson, R. E. (2005). Creating something from nothing: Resource construction through entrepreneurial bricolage. Administrative Science Quarterly, 50, 329–366.Find this resource:
Bartunek, J. M., Rynes, S. L., & Ireland, R. D. (2006). Academy of Management Journal Editors’ Forum: What makes management research interesting, and why does it matter? Academy of Management Journal, 49(1), 9–15.Find this resource:
Battilana, J., & Dorado, S. (2010). Building sustainable hybrid organizations : The case of commercial microfinance organizations. Academy of Management Journal, 53(6), 1419–1440.Find this resource:
Bechky, B. A. (2011). Making organizational theory work: Institutions, occupations, and negotiated orders. Organization Science, 22(5), 1157–1167.Find this resource:
Bhakoo, V., & Choi, T. (2013). The iron cage exposed: Institutional pressures and heterogeneity across the healthcare supply chain. Journal of Operations Management, 31(6), 432–449.Find this resource:
Bingham, C. B., & Davis, J. P. (2012). Learning sequences: Their existence, effect, and evolution. Academy of Management Journal, 55(3), 611–641.Find this resource:
Bingham, C. B., & Eisenhardt, K. M. (2011). Rational heuristics: The “simple rules” that strategists learn from process experience. Strategic Management Journal, 32(1), 1437–1464.Find this resource:
Brown, S. L., & Eisenhardt, K. M. (1997). The art of continuous change: Linking complexity theory and time-paced evolution in relentlessly shifting organizations. Administrative Science Quarterly, 42(1), 1–34.Find this resource:
Chandler, A. D. (1962). Strategy and structure: Chapters in the history of American industrial enterprises. Cambridge, MA: MIT Press.Find this resource:
Davis, J. P., & Eisenhardt, K. M. (2011). Rotating leadership and collaborative innovation. Administrative Science Quarterly, 56(2), 159–201.Find this resource:
Dutton, J. E., & Dukerich, J. M. (1991). Keeping an eye on the mirror: Image and identity in organizational adaptation. Academy of Management Journal, 34(3), 517–554.Find this resource:
Edmondson, A. E., & McManus, S. E. (2007). Methodological fit in management field research. Academy of Management Review, 32(4), 1155–1179.Find this resource:
Eisenhardt, K. M. (1989a). Building theories from case study research. Academy of Management Review, 14(4), 532–550.Find this resource:
Eisenhardt, K. M. (1989b). Making fast strategic decisions in high-velocity environments. Academy of Management Journal, 32(3), 543–576.Find this resource:
Eisenhardt, K. M., & Graebner, M. E. (2007). Theory building from cases: Opportunities and challenges. Academy of Management Journal, 50(1), 25–32.Find this resource:
Eisenhardt, K. M., Graebner, M. E., & Sonenshein, S. (2016). Grand challenges and inductive methods: Rigor without rigor mortis. Academy of Management Journal, 59(4), 1113–1123.Find this resource:
Ferlie, E., Fitzgerald, L., Wood, M., Hawkins, C., Ferlie, E., & Hawkins, C. (2005). The nonspread of innovations: The mediating role of professionals. Academy of Management Journal, 48(1), 117–134.Find this resource:
Galunic, D. C., & Eisenhardt, K. M. (1996). The evolution of intracorporate domains: Divisional charter losses in high-technology, multidivisional corporations. Organization Science, 7(3), 255–282.Find this resource:
Galunic, D. C., & Eisenhardt, K. M. (2001). Architectural innovation and modular corporate forms. Academy of Management Journal, 44(6), 1229–1249.Find this resource:
Garg, S., & Eisenhardt, K. M. (2017). Unpacking the CEO–Board relationship: How strategy making happens in entrepreneurial firms. Academy of Management Journal, 60(5), 1828–1858.Find this resource:
Gehman, J., Glaser, V. L., Eisenhardt, K. M., Gioia, D., Langley, A., & Corley, K. G. (2018). Finding theory–method fit: A comparison of three qualitative approaches to theory building. Journal of Management Inquiry, 27(3), 284–300.Find this resource:
Gephart, R. P. (2004). Qualitative research and the Academy of Management Journal. Academy of Management Journal, 47(4), 454–462.Find this resource:
Gersick, C. J. G. (1988). Academy of management time and transition in work teams: Toward a new model of group development. Academy of Management Journal, 31(1), 9–41.Find this resource:
Gilbert, C. J. G. (2005). Unbundling the structure of inertia: Resource versus routine rigidity. Academy of Management Journal, 48(5), 741–63.Find this resource:
Gioia, D. A., Corley, K. G., & Hamilton, A. L. (2012). Seeking qualitative rigor in inductive research: Notes on the Gioia methodology. Organizational Research Methods, 16(1), 15–31.Find this resource:
Glaser, B. G., & Strauss, A. L. (1967). The discovery of grounded theory: Strategies for qualitative theory. Chicago, IL: Aldine Publishing Co.Find this resource:
Graebner, M. E. (2004). Momentum and serendipity: How acquired leaders create value in the integration of technology firms. Strategic Management Journal, 25(89), 751–777.Find this resource:
Graebner, M. E. (2009). Caveat venditor: Trust asymmetries in acquisitions of entrepreneurial firms. Academy of Management Journal, 52(3), 435–472.Find this resource:
Hallen, B. L., & Eisenhardt, K. M. (2012). Catalyzing ties and efficient tie formation: How entrepreneurial firms obtain investment ties. Academy of Management Journal, 55(1), 35–70.Find this resource:
Hannah, D. P., & Eisenhardt, K. M. (2018). How firms navigate cooperation and competition in nascent ecosystems. Strategic Management Journal, 39(12), 3163–3192.Find this resource:
Hargadon, A. B., & Douglas, Y. (2001). When innovations meet institutions: Edison and the design of the electric light. Administrative Science Quarterly, 46(3), 476–501.Find this resource:
Kellogg, K. C. (2011). Operating room: Relational spaces and microinstitutional change in surgery. American Journal of Sociology, 115(3), 657–711.Find this resource:
Langley, A. (1999). Strategies for theorizing from process data. Academy of Management Review, 24(4), 691–710.Find this resource:
Martin, J. A., & Eisenhardt, K. M. (2010). Rewiring: Cross-business unit collaborations in multibusiness organizations. Academy of Management Journal, 53(2), 265–301.Find this resource:
McDonald, R., & Eisenhardt, K. M. (2019). Parallel play: Startups, nascent markets, and the search for a viable business model. Administrative Science Quarterly.Find this resource:
McDonald, R., & Gao, C. (2019). Pivoting isn’t enough? Managing strategic reorientation in new ventures. Organization Science.Find this resource:
Miles, M. B., Huberman, A. M., & Saldaña, J. (2014). Qualitative data analysis: A methods sourcebook (3rd ed.). Thousand Oaks, CA: SAGE.Find this resource:
Ott, T., & Eisenhardt, K. M. (2018). Decision weaving: Effective strategy formation in entrepreneurial settings research-oriented. Academy of Management Proceedings, 2018, 12137.Find this resource:
Ozcan, P., & Eisenhardt, K. M. (2009). Origin of alliance portfolios: Entrepreneurs, network strategies, and firm performance. Academy of Management Journal, 52(2), 246–279.Find this resource:
Ozcan, P., Han, S., & Graebner, M. E. (2017). Single cases: The what, why, and how. In Raza A. Mir & Sanjay Jain (Eds.), The Routledge companion to qualitative research in organization studies (pp. 92–112). New York, NY: Routledge,Find this resource:
Ozcan, P., & Santos, F. M. (2015). The market that never was: Turf wars and failed alliances in mobile payments. Strategic Management Journal, 36(10), 1486–1512.Find this resource:
Pache, A.‑C., & Santos, F. (2013). Inside the hybrid organizations: Selective coupling as a response to competing institutional logics. Academy of Management Journal, 56(4), 972–1001.Find this resource:
Siggelkow, N. (2007). Persuasion with case studies. Academy of Management Journal, 50(1), 20–24.Find this resource:
Tidhar, R., & Eisenhardt, K. M. (2019). Get rich or die trying… unpacking revenue model choice using machine learning and multiple cases. Academy of Management Proceedings, 2019(1), 12218.Find this resource:
Tripsas, M. & Gavetti, G. (2000). Capabilities, cognition, and inertia: Evidence from digital imaging. Strategic Management Journal, 21(10–11), 1147–1161Find this resource:
Valentine, M. A. (2018). Renegotiating spheres of obligation: The role of hierarchy in organizational learning. Administrative Science Quarterly, 63(3), 570–606.Find this resource:
Van Maanan, J. (2011). Tales of the field: On writing ethnography. Chicago, IL: University of Chicago Press.Find this resource:
Walsh, I., Holton, J. A., Bailyn, L., Fernandez, W., Levina, N., & Glaser, B. (2015a). Rejoinder: Moving the management field forward. Organizational Research Methods, 18(4), 620–628.Find this resource:
Walsh, I., Holton, J. A, Bailyn, L., Fernandez, W., Levina, N., & Glaser, B. (2015b). What Grounded Theory is . . . A critically reflective conversation among scholars. Organizational Research Methods, 18(4), 1–19.Find this resource:
Whyte, W. F. (1943). Street corner society: The social structure of an Italian slum. Chicago, IL: University of Chicago Press.Find this resource:
Yin, R. K. (2017). Case study research and applications: Design and methods. Thousand Oaks, CA: SAGE.Find this resource:
(1.) This article addresses the process of creating single case and multiple case studies, though certain aspects of the process such as replication logic and cross-case analysis apply to multiple case studies.