Stop Comparing Industries and Start Comparing What Actually Matters

Strategy scholars have become much more careful in demonstrating the validity of their findings, using stronger research designs and empirical methods. But the field has paid less attention to a different question: whether, and under what conditions, a finding from one research setting should also apply in another.

Strategy scholars have become much more careful in demonstrating the validity of their findings, using stronger research designs and empirical methods. But the field has paid less attention to a different question: whether, and under what conditions, a finding from one research setting should also apply in another.

So how can strategy scholars avoid that confusion and build knowledge that carries across contexts, while still recognizing what makes each context distinct? In our latest research article published in the Strategic Management Journal, “Mapping the Landscape of Research Findings: Generalization across Contexts in Strategic Management Research,” we address this basic challenge and offer both a diagnosis of the problem and a practical framework for moving the field forward.

We begin with a simple observation: strategy research usually relies on nominal categories, such as industry codes, calendar years, patent classes, and geographic regions, to describe research contexts. These labels are familiar and easy to use, but they are often weak stand-ins for the deeper conditions that shape behavior: two industries with nearby classification codes may differ sharply in levels of uncertainty or interdependence. Two adjacent years may reflect very different macroeconomic conditions. Two neighboring countries may have very different regulatory systems. When findings do not replicate across such contexts, researchers often end up with ad hoc boundary conditions instead of cumulative insight.

We argue that better generalization starts with a different way of thinking about context. Instead of asking whether a finding generalizes from one industry or time period to another, scholars should ask whether it generalizes across contextual attributes that reflect deeper conceptual categories. This abstraction — from nominal categories to theoretically meaningful attributes — matters greatly. Industry concentration, technological modularity, macroeconomic shocks, and societal norms are not just background conditions to control for. They shape how causal mechanisms operate, and they can determine whether a relationship should be expected to hold in a new context. Focusing on these attributes can make comparisons across contexts easier and help explain findings that seem contradictory when viewed only through nominal labels. In particular, we suggest that three higher-level conceptual categories — uncertainty, interdependence, and variance — can help identify useful contextual attributes across many settings.

We illustrate this argument with examples from recent studies in the Strategic Management Journal, including research on alliance formation, global diversification, team composition and innovation, and executive mobility. These examples show that results that initially look inconsistent across contexts become easier to understand when settings are reorganized around attributes such as technological uncertainty, industry interdependence, or societal variance. What first appears to be empirical noise may instead reflect systematic heterogeneity.

We also distinguish generalization from replication. The recent focus on replication and quasi-replication has improved empirical rigor, but replication by itself does not tell us where to look next. The number of possible contexts is enormous, and unguided quasi-replication can become inefficient or uninformative. To address this problem, we describe a “research landscape” in which each empirical finding is a point defined by both the observed relationship between variables and the context in which it is observed. Progress comes not from filling in random or convenient points, but from strategically sampling the landscape to identify where findings remain robust and where they break down.

This perspective changes how we think about future studies. Instead of treating new settings as simple robustness checks, scholars should choose settings that create clear contrasts on important conceptual dimensions. Contexts in which theory predicts a finding will fail can be just as informative as those in which the finding holds. In this way, generalization becomes an active part of theory development, rather than a passive extension of earlier work.

For a field that aims to build cumulative knowledge, but is often fragmented by context-specific findings, we hope this offers a timely and constructive contribution. We do not argue for broad claims that ignore nuance, or for narrow contextualism that resists abstraction. Instead, we offer a framework for disciplined generalization, one that respects the richness of context while supporting mid-range theorizing. Mapping the research landscape can help the strategy field accumulate knowledge more effectively.