

Marketing research rarely says that something works everywhere. More often it says that a message, a discount, or a loyalty program works better for some customers or in some situations. Managers rely on these conditional findings for segmentation and budget decisions. Yet an effect can be statistically significant and still be too small to justify an investment, or hold only for part of the customer base.
Le Minh Thanh To addresses this problem in a new article published in the Italian Journal of Marketing. The author examines how moderators, the variables that change the strength of a relationship, such as customer age or sales channel, are used and reported in marketing research.
The article shows that conditional effects are typically small, and that many studies do not report their size or where along the moderator they matter. Charts with truncated axes can make trivial differences look substantial. The author proposes three questions to ask before acting on a finding: Can the moderating factor be observed in practice, for instance in CRM data? Can the company act on it? Is the effect large enough to matter for business results? Findings that pass can become explicit if-then rules, with thresholds beyond which a strategy pays off. The central message is that the question moves from whether a result is statistically significant to whether it is reliable, meaningful, and actionable. The framework is a proposal for assessing studies and has not been tested empirically.
We spoke with the author to learn how misreading these effects leads to faulty decisions, why observable customer traits can be more practical than psychological measures, how to turn statistical results into if-then rules, and what single change managers should demand when applying academic findings.
How does misinterpreting moderating variables lead executives to make faulty strategic decisions?
Failing to distinguish between “pure” and “quasi-moderators” can lead executives to underestimate contextual nuances, resulting in misallocated budgets between improving the contextual factor (Z) and strengthening the primary driver (X). Without evaluating substantive effect sizes (e.g., ΔR² or Cohen’s f²), firms risk investing in strategies with negligible incremental benefits. More subtly, because marginal effects may hold only within specific ranges of the moderator (Z), applying such strategies universally across the customer base inevitably drives waste.
Why should marketing leaders prioritize observable customer traits over complex psychological constructs?
Complex psychological constructs, such as need for cognition, self-construal, or risk tolerance, typically require multi-item surveys for measurement. Self-reported psychological constructs are susceptible to response biases and common-method effects, including social desirability, acquiescence, recall and self-perception biases, fatigue, and context effects, which can introduce measurement error and distort estimated moderation effects. Furthermore, storing and leveraging sensitive psychological attributes may create additional data privacy and compliance risks under regulations such as the European Union’s General Data Protection Regulation (GDPR), while also causing consumer discomfort or a sense of privacy intrusion. In contrast, observable characteristics, such as demographics (age, gender), transaction context (online/offline channel, product category), or CRM and behavioral data (customer tenure, purchase frequency), can often be readily retrieved from existing systems and used in near real time. This makes them more practical for operational segmentation and enables firms to apply targeting, messaging, or promotional rules immediately to relevant customer segments.
How can managers turn complex statistical moderation results into actionable “if-then” decision rules?
Rather than relying solely on the interaction coefficient, managers should examine marginal effects across the full range of the moderator to identify meaningful thresholds or regions of significance. For instance, as reported in a study cited in our paper, when customer value homophily exceeds Z = -1.241 (covering 86.9% of the sample), intensifying rational advertising messages is significantly and positively associated with information diffusion; below this threshold, the campaign is unlikely to generate a statistically significant incremental effect. In practice, multiple moderators may operate simultaneously, with some amplifying and others weakening the focal effect. When several contextual conditions must be satisfied simultaneously, managers should translate these findings into explicit “IF-THEN” rules using “AND” logic. For example: “Increase investment in the communication campaign IF message quality is high AND message complexity is low.” This converts a complex statistical interaction into a clear, testable, and operational decision rule.
What single change should managers demand when applying academic findings to real-world strategy?
Managers are encouraged to place greater emphasis on practical relevance and decision utility. In this regard, we recommend that managers assess three criteria before applying research findings on moderating effects: (1) the estimated effect and its associated uncertainty should be assessed in relation to the Smallest Effect Size of Interest and the Region of Practical Equivalence to determine whether the effect is sufficiently meaningful in practice to inform managerial decisions; (2) the characteristics used to operationalize the moderating effect should be observable or identifiable in the relevant decision context; and (3) where appropriate, the findings should be translated into clear IF-THEN rules to guide managerial action.
In summary, we need to move beyond asking, ‘Is the moderator statistically significant?’ to asking, ‘Is it reliable, practically meaningful, and bounded by actionable thresholds that translate directly into business decisions?’
Copertina: Foto di Shafin Al Asad Protic da Pixabay
