It is crucial that the design of your research allows you to answer your research question. You want to prevent the situation that you have conducted your study and have to conclude that the data or measures did not allow you to answer the research question. That is why it is important to align your research design to your research question. A common typology (De Vaus, 2001) makes a distinction between four types of research designs: the case study, the cross-sectional study, the longitudinal study, and the experiment. The table below shows the key characteristics of the four designs and their use.

Case studies can be used to answer exploratory and descriptive research questions, and they can generate leads for potential explanations (see #5 here). Sampling procedures are very important in this case when you want to generalize your findings to a broader population. If you use the case study method, it is crucial to think about the rules you will apply for the selection of your cases (Seawright & Gerring, 2008). Selecting a group of ‘successful’ cases and contrasting them with ‘unsuccessful’ cases for instance helps you draw up a list of leads to potential causes, but it does not generate a representative sample of all cases.

A case study is hardly ever suitable to provide stringent evidence on causal questions. An example of a research question you cannot answer with a case study is: “To what extent do partnerships between corporations and nonprofit organizations solve social issues?” By studying a case of a successful partnership such as the development of the AstraZeneca vaccine against COVID-19 you may learn a lot about the collaboration between Oxford university and the pharmaceutical company AstraZeneca, but it remains unclear to what extent partnerships work. If you’re interested in this particular partnership, put it in the research question so that your case study aligns with it. Your revised research question could be: “How did the partnership between Oxford University and AstraZeneca develop and successfully produce a vaccine?” This is a purely descriptive question.

You should be dissatisfied with this question because many studies have been published already suggesting ‘success and failure’ factors for cross-sector partnerships (Selsky & Parker, 2005; Koshmann, Kuhn & Pfarrer, 2012; Babiak & Thibault, 2018; Clarke & Crane, 2018). So in addition to the descriptive question you could ask: “Which features of the partnership between Oxford University and AstraZeneca are inconsistent with theories on cross-sector partnerships?” With this focus on inconsistencies, you may learn about conditions in this particular case that invalidate insights from previous research, or indicate boundary conditions in which insights from previous research do not hold. This is a better approach than to ask: “Which features of the partnership between Oxford University and AstraZeneca are consistent with theories on cross-sector partnerships?” With such a positive focus you may miss the opportunity to learn from anomalies (see #3 here).

If you’re generally interested in partnerships or want to test theories about it, design your study to collect data on a larger number of partnerships. For instance, you could design a survey asking corporations about their experiences with partnerships involving nonprofit organizations. This is an example of a cross-sectional design. You could also study the annual reports of a sample of corporations to see whether they mention partnerships with nonprofit organizations. The data could tell you what kinds of corporations engage in which kinds of partnerships. Also they may indicate to what extent the corporations claim that the partnerships were successful. This design may be more informative than the single case because it tells you something about a larger population of cases.

However, you would still not learn much about the effects of partnerships from the surveys or annual reports, because you do not observe the counterfactual: what would have happened if the corporation had not engaged in this partnership? At the very least you need a sample of corporations that do not engage in partnerships. Consequences of partnerships should be more prominently reported by corporations that engaged in them than among those that did not. It would be better still to also collect data on failed partnerships. Because corporations are unlikely to describe their failures in annual reports you may need to design alternative data collection strategies to get information about them. Even in surveys, however, respondents may be reluctant to admit that they made mistakes at work or talk about the failures of colleagues.

Because it is so difficult to get good quality data that are informative for the original research question, it may be better to formulate an alternative research question on partnerships. For example: you may ask what kind of corporations are more likely to engage in partnerships than others, and how they evaluate them.