Promoting sustainable behavior: addressing user clusters through targeted incentives

The study examines whether distinct clusters of users can be identified based on sustainability-related values, attitudes, intentions, and perceived barriers, and whether tailored interventions can effectively promote sustainable behavior within these clusters. Motivated by the high share of household emissions and the potential of demand-side solutions, the work situates behavior change within established frameworks (COM-B, Theory of Planned Behavior, Protection Motivation Theory) and acknowledges cognitive mechanisms (e.g., attitude–behavior gap, misperceptions of high-impact actions). The purpose is to derive personas and validate them with data-driven clustering to inform individualized, scalable interventions for product design and communication.

The paper synthesizes key behavioral frameworks and determinants relevant to sustainable behavior: COM-B (capability, opportunity, motivation) for behavior change; Theory of Planned Behavior (attitudes, subjective norms, perceived behavioral control) and Protection Motivation Theory (self-efficacy, rewards, costs). It reviews psychological and social factors influencing sustainable behavior, such as attitudes, values (Schwartz), emotions, self-efficacy, personality traits, convenience, involvement, motivation, intentions, demographics, social norms, knowledge, and greenwashing. It highlights cognitive aspects including the attitude–behavior gap and misestimation of high-impact behaviors. Prior research on segmentation and personas in sustainability and user-centered design is referenced to justify cluster-based targeting and persona development.

Mixed-methods design with two parts. Part 1 (Qualitative): Semi-structured interviews with 10 participants (6 women, 4 men; age 26–64, Mage=36.5, SD=11.44) recruited via an agency in Germany; compensation €25. Interviews (~65 min) covered everyday washing habits, sustainability attitudes and social environment, and sustainability in purchasing decisions. Data were recorded (Teams), transcribed, GDPR-compliant, and coded in MAXQDA. Reflexive thematic analysis and the KJ (affinity mapping) method were used by two raters to derive themes (prerequisites, enablers, barriers) and to synthesize five qualitative personas. Part 2 (Quantitative): Online survey (Qualtrics) with adults in Germany recruited via an access panel. Initial N=444; exclusions for consent/age/incompletion/inattentive processing and very fast completion (<10% of average, <553 s) yielded final N=342 (174 female, 165 male, 3 diverse; Mage=49.9, SD=17.16). Demographics recorded (household size, children, income). Measures: validated scales from literature on five subtopics—Sustainable Attitudes and Values (personal environmental responsibility, environmental concern, Schwartz values, CSC environmental/social/economic), Knowledge and Information (technology affinity, pro-environmental product knowledge, skepticism toward environmental ads, trust), Perceived Consumer Effectiveness, Social Factors (social status, social norms), and Price/economic benefits. Data processing: standardization and MinMax scaling. To limit dimensionality (recommended sample-to-variable ratio ~70:1), variables were split into clustering features vs. covariates; low-variance variables (Var ≤ 0.045) removed; final set of five variables for dimensionality reduction and clustering (social status, trust, skepticism, economic benefit, care for sustainability). Clustering: HDBSCAN implemented in Python (SciPy, scikit-learn) following a two-stage procedure; evaluated with Calinski–Harabasz index and silhouette score. An unsupervised decision tree (depth=5 via grid search) provided interpretability of cluster structure; performance reported with entropy (H ≥ 0.91) and Gini coefficient (≈83.5). Post hoc differences among clusters assessed with Tukey HSD. Ethics: approval from the University of Stuttgart ethics committee (Az. 22-023); informed consent; GDPR compliance; preregistration with noted deviations explained.

• Qualitative themes: prerequisites (understanding of sustainability, knowledge, self-efficacy, sustainable attitudes/values), enablers (social factors, measurability/tangibility), and barriers (stress/time, price, convenience; plus other context-specific barriers). A conceptual "behavioral weighing pan" illustrates how factors combine to tip decisions toward or away from sustainable actions. Five qualitative personas were drafted: Activist, Sustainability-interested, Hedonistic, Indifferent, and Dismissive.
• Clustering outcomes: HDBSCAN clustered 326/342 samples (silhouette Sc < .0583). A depth-5 decision tree (entropy H ≥ 0.91; Gini ≈ 83.5) helped explain cluster boundaries using features: care for sustainability, skepticism, trust in labels, social status, and economic benefit. Five clusters emerged and were mapped to validated personas: • Socially Sustainable (Cluster 1; n=39): High belief in human-caused climate change; high care/concern; high perceived consumer effectiveness, environmental knowledge, trust in eco-labels; higher social status/norms; strong environmental responsibility and conservation values; relatively tech-savvy and collaborative; low skepticism. • Responsible Savers (Cluster 2; n=86): Strong belief in climate change; high environmental responsibility and conservation; value simplicity, collaboration, debt-free consumption; prioritize sustainable purchases; knowledgeable; more sustainable but less tech-savvy than average. • Unconcerned Spenders (Cluster 3; n=84): Low belief in climate change, low care/responsibility; hedonism and spending prioritized over sustainability; low awareness of sustainable consumption; average social status/trust; low preference for socially sustainable products, collaboration, simplicity, debt-free consumption; focus on immediate gratification. • Comfort-Oriented (Cluster 4; n=62): Lowest engagement; low concern, knowledge, and tech affinity; below-average openness to change, belief in climate change, care, and perceived consumer effectiveness; slightly above-average social status/norms, environmental responsibility, self-enhancement, skepticism and trust in labels; money and comfort are primary hindrances. • Skeptical Consumers (Cluster 0; n=55): Lowest trust and care for sustainability combined with low skepticism and low engagement; below-average perceived consumer effectiveness, social norms, tech affinity, conservation values, belief in climate change; average environmental concern; slightly above-average knowledge, responsibility, openness to change, self-enhancement, self-transcendence; highest cost barrier; not motivated by environmental concern or social recognition.
• Statistical differences: Tukey HSD indicated significant pairwise differences across clusters for care for sustainability, economic benefit, social status, skepticism, and trust (multiple comparisons reported). Costs and convenience emerged as the most influential barriers across clusters.
• Persona mapping: Hedonists ≈ Unconcerned Spenders; Dismissive ≈ Skeptical Consumers; Indifferent ≈ Comfort-Oriented; Activists/Interested distributed across Socially Sustainable and Responsible Savers.

The findings support the central premise that individuals differ meaningfully in motivations, beliefs, social drivers, and perceived barriers related to sustainability, and that identifying clusters enables targeted intervention design. By validating qualitative personas with data-driven clusters, the study bridges descriptive insights and quantitative segmentation, offering actionable guidance for user-centered product development and tailored communication. The clusters suggest differentiated leverage points: social-image and norm-based appeals for Socially Sustainable; simplification and feedback for Responsible Savers; monetary framing and hedonistic incentives for Unconcerned Spenders; convenience-enhancing design and cost reduction for Comfort-Oriented; and trust-building, dialog-based engagement with cost-savings feedback for Skeptical Consumers. The work underscores the importance of combining knowledge interventions with choice architecture and socio-cultural strategies to address misperceptions and the attitude–behavior gap, particularly among socially oriented groups.

The study integrates qualitative and quantitative methods to derive and validate five sustainability-related user personas/clusters, offering a practical framework to individualize interventions for behavior change. By aligning design and communication with cluster-specific motivators and barriers, organizations can promote high-impact sustainable behaviors more effectively. The work contributes validated personas, a replicable clustering approach, and guidance for targeted incentives and messaging. Future research should test personalized interventions on real behavior (e.g., through randomized controlled experiments), expand to other cultural contexts and larger samples, periodically update segmentation variables, and refine personas to maintain representativeness.

Key limitations include: reliance on self-reported intentions rather than observed behavior; sample restricted to Germany and an online access panel with exclusions that may introduce selection bias; potential respondent fatigue due to survey length and matrix format; limited number of variables retained for clustering, omitting factors (e.g., time pressure, stress) identified qualitatively; modest silhouette score; and possible bias toward more salient factors. Although HDBSCAN handles varying densities and noise better than some alternatives, variables may not capture full cluster complexity. The authors note plans for brief screening instruments, behavioral measures, and broader, culturally diverse samples in future work.