Introduction

Humankind is already experiencing the consequences of unsustainable lifestyles in numerous areas. Climate change and the crossing of other planetary boundaries (Rockström et al., 2009) are more present in our daily lives than ever. Heat waves, crop failures, wildfires, droughts with severely decreasing water levels of lakes, floods, and loss of biodiversity are just a few examples of many. One way to address this urgency is through individual behavior change. Mitigating the demand could lead to reductions of 62% (5.8 Gigatons of Carbon Dioxide equivalent [GtCO2e]) in the transport sector and 41% (7.3 GtCO2e) in the food sector (Creutzig et al., 2021).

Various groups and initiatives have been working on sustainability-related behavior change of individuals, such as consumer choices. The Intergovernmental Panel on Climate Change (IPCC) describes three factors that mitigate the demand on the consumer side: technology adaptation, infrastructure use, and socio-cultural adaptation. First, technology adaptations like electric vehicles and efficient, lightweight cars could mitigate the greenhouse gas-intensive demand (Creutzig et al., 2022). Second, infrastructure use could be enhanced by utilizing sharing concepts like pooled mobility (Creutzig et al., 2022), which involves using shared transportation, such as carpooling or public transport. Third, adjusting socio-cultural factors holds significant potential, such as avoiding long-haul flights and transitioning to train travel (Creutzig et al., 2022). Systemic change is required to enable individuals to behave more sustainably (Abson et al., 2016) as it alters a system’s structures, policies, and processes. Impactful opportunities exist in the interaction of governments, the private sector, civil society, and individuals (Shukla et al., 2022). Nevertheless, changing behavior regarding technology adaptation, infrastructure use, and socio-cultural factors is not trivial due to the versatility of strategies and arguments that motivate people to think and act sustainably.

Understanding the parameters of sustainable behavior and the associated theoretical frameworks is crucial for gaining insight into behavioral adaptation possibilities. Behavior is considered sustainable if it meets present needs without jeopardizing the ability of future generations to meet their needs (Brundtland, 1987). Several theories and frameworks attempt to explain people’s behavior, each incorporating different behavioral factors. These frameworks can help to identify barriers and develop targeted interventions for sustainable behavior change. Approaching various factors influencing an individual’s behavior can encourage behavioral changes more effectively. The COM-B model proposes Capability, Opportunity, and Motivation as critical components for Behavior change (Michie et al., 2014; Michie et al., 2011). Capability refers to the ability to do something, opportunity is determined by external factors, and motivation describes the willingness to engage in the behavior. An alternative framework is the theory of planned behavior (Ajzen, 1985, 1991), which specifies attitude, subjective norm, and perceived behavioral control as factors influencing intention and, consequently, the resulting behavior. Lastly, the protection-motivation theory (Rogers, 1995) focuses on influencing factors such as self-efficacy, rewards, and costs. These theoretical frameworks help to determine factors that highly influence sustainable behavior, which can vary between individuals. These factors can be grouped into psychological and social components. Psychological factors include attitudes (Haustein & Hunecke, 2013; Vermeir & Verbeke, 2008), values (McCarty & Shrum, 1994; Schultz et al., 2005; Schwartz, 2021; Vermeir & Verbeke, 2008), emotions (Brosch & Steg, 2021), self-efficacy (Hanss & Böhm, 2010; Hanss & Doran, 2020) and personality traits such as narcissism (Lin et al., 2021), product involvement (Tarkiainen, Sundqvist 2009), convenience (McKenzie-Mohr & Schultz, 2014), general interest in sustainability (Matthes & Wonneberger, 2014), motivation (McKenzie-Mohr, 2011), and behavioral intention (Wang et al., 2014). Furthermore, social circumstances such as demographics such as gender (Tautscher et al., 2020; Bloodhart & Swim, 2020) and social norms (Salazar et al., 2012); Schultz et al., 2007) can influence behavior. Finally, other factors, such as knowledge (Steg & Nordlund, 2019) and greenwashing (Hameed et al., 2021), may also impact behavior. These factors have the potential to either support or inhibit sustainable behavior.

Moreover, sustainable behavior might be influenced by cognitive mechanisms of perception and processing. For instance, the attitude-behavior gap, or attitude-action gap, describes a discrepancy between people’s stated environmental concerns and their actual behavior (Juvan & Dolnicar, 2014; Terlau & Hirsch, 2015; Young et al., 2009). For example, an individual can be aware of the high air travel emissions but still choose to travel by plane. Second, individuals are likely to make error-prone assumptions regarding behavior that causes significant harm to the environment. Most individuals overestimate, for example, the effect of CO2 emissions from the plastics industry, while underestimating the impact of meat consumption or thermal insulation (Bilstein & Rietmann, 2020). Interventions can address specific behaviors influenced by the highlighted factors and cognitive effects. According to the behavior change wheel, behavior can be controlled or changed by strategies such as incentives, persuasion, constraints, restrictions, or training (Michie et al., 2014). Incentives can provide rewards or benefits for people who behave sustainably, and persuasion can lead to the transmission of social norms that promote sustainable behavior.

People have unique needs due to individual factors, social circumstances, and cognitive mechanisms. These individual influencing factors suggest the effectiveness of personalized targeting. For instance, incentives may be more effective than restrictive strategies for individuals with hedonistic tendencies, as they are primarily motivated by seeking pleasure and personal satisfaction. By contrast, a socially orientated person might be more motivated by feedback from their social environment. Therefore, we need an enhanced understanding of individual factors for sustainable intentions and behavior for different user groups. The analysis of individual influencing factors could help change individuals’ behavior. Clustering of users can account for individual differences and is an appropriate way to identify and describe distinguished groups. Here, personas are a valuable tool for understanding users’ attitudes and behaviors at a descriptive level. A persona represents an individual within a group (Chang et al., 2008). This approach is well-established in user-centered research, which aims to create products or services that meet users’ needs more effectively. Employing a cluster analysis approach to create personas and subsequently design targeted interventions is relevant both to research (Brickey et al., 2012; Pruitt & Grudin, 2003; Tu et al., 2010) and practice, for example, in market research institutes (Czioska et al., 2021; Gatterer & Tewes, 2023; Kroth, 2019; Brincken, 2022; Tautscher et al., 2020).

Building on the outlined evidence, we investigate the effectiveness of an individualized targeting approach to promote sustainability among user groups. To this end, we explore which clusters of sustainable intentions can be identified among individual user groups. More precisely, we are interested in (a) the main differences between personas, and (b) how these differences can be validated with data-driven user clusters (Balderjahn et al., 2018; Malatesta & Breadsell, 2022; Tabianan et al., 2022). Consequently, our investigation consists of two successive parts: a qualitative interview aims to identify sustainable personas with individual influencing factors, and a quantitative survey aims to validate the identified personas by data-driven clustering. The second method employs statistical algorithms to identify patterns and cluster the data systematically. Five personas were identified across both parts, which differed in terms of underlying strategies towards sustainable behavior. Specifically, we identified the Unconcerned Spenders, Skeptical Consumers, Socially Sustainable, Comfort-Oriented, and Responsible Savers. Some of these personas respond to sustainability benefits, while others might be better approachable with benefits appealing to their hedonistic side. The personas also differ in their general affinity for technology, collaboration, and novelty factors. Overall, this work supports the envisioned effectiveness of personalized behavioral interventions.

Methods

Part 1: Interview to create sustainable personas

Existing literature extensively examines individual factors, yet formal analyses about different clusters of users are scarce. Awe conducted semi-structured interviews to learn about the needs of different user groups and ultimately create personas. Taking washing behavior as an example, the underlying aim was to identify differences in sustainability factors, enablers, and impediments to sustainable behavior.

Participants

We interviewed ten participants about behavioral characteristics and strategies related to washing habits. The participants (n = 10) included women (n = 6) and men (n = 4) between 26 and 64 years (Mage = 36.5 years, SDage = 11.44 years). Participants were recruited through an agency (Lämmler, 2020) and were intended to represent a small sample of the German population. In our pre-screening process, we inquired about the participants’ regular washing habits. This ensured that participants were accustomed to regular washing and could provide accurate self-reports based on their memory. Participants received a monetary compensation of 25 € for participating in the interview.

Materials

We examined three topics with sub-questionsFootnote 1 to explore individual differences related to potential sustainability influencing factors. The interview started by reporting on washing habits as an example of everyday behavior, thereby ensuring that other questions did not influence answers to this question. The second topic included participants’ attitudes and social environment concerning sustainability in everyday life. The last topic focused on whether and how they consider sustainability a criterion in their purchasing decisions. The questions focused on factors from the literature concerning individual and social determinants of sustainable behavior, alongside insights derived from preliminary test interviews. On average, interviews lasted for Mlength = 65.4 min (SD = 8.34 min).

Data collection and preparation

The interviews were conducted and recorded with Microsoft Teams, an online collaboration application. Each participant gave written consent. We ensured to closely follow the General Data Protection Regulation (GDPR) principles for storing personal data and checked compliance beforehand. The interview data obtained was transcribed with Adobe Creative Cloud and tagged in MAXQDA a program for qualitative data analysis.

Data exploration

Reflexive thematic analysis (RTA) was utilized for data analysis inductively to interpret qualitative datasets (Braun & Clarke, 2019). This method includes rereading the interview transcripts to analyze overreaching themes across participants. Related excerpts were collected, reread, and categorized by two independent raters. We conducted a systematic and iterative review of the transcripts and their associated categories using the KJ Method (Scupin, 1997), also known as affinity mapping, to identify and extract themes. First, excerpts were sorted and comparatively organized to discover emerging themes or patterns between excerpts. Afterward, we created personas by affinity mapping and assigning experts according to the main themes. Core differentiating characteristics of participants were synthesized, and in a final step, resulting clusters were supplemented by demographic information and a name for each persona.

Part 2: Online questionnaire to validate sustainable personas

Following the outlined phase of conceptual development, a quantitative exploration served the purpose of validating the obtained personas by formalized clustering. During the subsequent stages of the research, the conducted study adhered to the applicable guidelines and regulations as outlined in Standard 8 of the Ethical Principles and Code of Conduct for Psychologists (American Psychological Association, 2010)Footnote 2. This involved obtaining informed consent from all participants and securing ethical approval. Furthermore, compliance with the General Data Protection Regulation (GDPR) was ensured.

Participants

Tying in with related prior research (e.g., Balderjahn et al., 2018; Lee & Haley, 2022; Niedermeier et al., 2021), the final sample included 342 participants aged 18 years or older (174 females, 165 males, and three diverse; Mage = 49.9 years, SDage = 17.16). Participants were recruited by an agency (Bidou, 2020). The utilized age distribution was aligned with a report by the Federal Statistical Office to represent the age demographics of Germany’s adult population (Minage ≧ 18) (Destatis, 2022). Most participants lived in either 2-person households (n = 144, 42.1%) or alone (n = 94, 27.5%). Furthermore, most participants had no children living in the same household (n = 248, 72.5%). The household income per month was mainly in the range of 3600 € to 4999 € (n = 74, 21.6%), followed by ranges of 2600 € to 3599 € (n = 66, 19.3%), 2000 € to 2599 € (n = 47, 13.7%) and 1500 € to 1999 € (n = 39, 11.4%).

Materials

Building on the main themes emerging from the interviews and recent research on segmenting users’ sustainable behaviors (Balderjahn et al., 2018; Lee & Haley, 2022; Niedermeier et al., 2021), we selected a set of validated scales corresponding to five subtopics. These cover aspects such as individual attitudes, social factors, price, understanding of sustainability, and knowledge. More precisely, regarding the subtopic Sustainable Attitudes and Values, we collected data on personal environmental responsibility (Lai & Cheng, 2016; Lee, 2008), environmental concern (Lee, 2008; Straughan & Roberts, 1999), human values (Schwartz et al., 2001), and consciousness for sustainable consumption (CSC) scale, including an environmental, social, and economic dimension (Balderjahn et al., 2018; Ziesemer et al., 2016). Regarding the subtopic Knowledge and Information, we collected data on affinity for technology (Neyer et al., 2016), knowledge of pro-environmental products (Flynn & Goldsmith, 1999), skepticism towards pro-environmental advertising (Mohr et al., 1998; Obermiller & Spangenberg, 1998), and trust (Voon et al., 2011). We further included the Perceived Consumer Effectiveness scale, which builds on the planned behavior theory (Roberts, 1996). Regarding the subtopic of Social Factors, we collected data on social status and social norms (Lee, 2008; Suki & Suki, 2015). Additionally, we obtained data on economic benefits regarding the subtopic Price.

Data collection and preprocessing

Participants were asked to complete an online survey presented in a matrix-questionnaire format utilizing Qualtrics, an online survey platform. The survey was designed for 10–20 min and required a median duration of 19.05 min (SD = 5.97 min) for completion across participants. To ensure appropriate data quality, we had to exclude several of the initially recruited 444 participants, as they did not consent to data collection, did not meet the age requirements (n = 8), did not complete the survey (n = 16) or admitted inattentive processing (n = 6). In addition, participants with a survey completion time lower than 10% of the sample average, corresponding to an engagement of less than 553 s, were also excluded (n = 41).

Clustering procedure

Utilizing Python 3 (Van Rossum & Drake, 2009), Anaconda 3 (Anaconda, 2020), and the Scipy (Virtanen et al., 2020) and Sklearn (Pedregosa et al., 2011) libraries, we implemented a Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) clustering algorithm in Jupiter Notebook (Kluyver et al., 2016). Across all included variables, data was standardized and scaled with the MinMax-Scaler. Following general recommendations that the sample size should outperform the number of variables by a factor of 70 (Dolnicar et al., 2013), we first separated cluster variables and covariates to avoid including an overly high number of variables (Fraiman et al., 2008). Considering our sample size, we included a final set of five variables for dimensionality reduction. Thereby, we eliminated variables with a low variance (\({Var} < =0.045)\) in the process of feature selection.

Following established two-stage clustering procedures (e.g., Balderjahn et al., 2018; Hellwig et al., 2015; Lee & Haley, 2022), we applied HDBSCAN, as this algorithm incorporates the advantages of hierarchical and density-based clustering approaches. Results were evaluated utilizing both the Calinski-Harabasz index (Calinski & Harabasz, 1974) and the silhouette scores (Rousseeuw, 1987). To enhance the interpretability of clustering outcomes, we employed an unsupervised machine-learning approach in the form of a decision tree (Laber et al., 2023). Decision trees can generally be helpful because they provide an approximate clustering result, facilitate examining feature interactions in the data, and offer reasonable explanations for clustered contents (Molnar, 2022). To evaluate the relationship between clusters, we employed descriptive statistics. The differences between obtained clusters were calculated using a Tukey Honestly Significant Difference (HSD) Test (Tukey, 1949), providing the foundation for deriving the personas.

Results

Part 1: Identification of sustainability personas

Thematic analysis (Braun & Clarke, 2019) revealed three primary themes: prerequisites for sustainability, enabling factors, and barriers, as presented in Table 1. More precisely, the prerequisites for sustainability included understanding sustainability, knowledge, self-efficacy, and sustainable attitudes and values. These conditions were present in the more sustainable personas, while the less sustainable personas lacked one or more of these prerequisites. Enablers and barriers can work in both directions, as social factors and tangibility can promote and hinder sustainable behavior. Participants reported that social factors, such as a sustainable peer group, increased sustainability intention, while stress and convenience hindered sustainable behavior.

Table 1 Main themes and selected associated factors emerging from the qualitative interviews.
Full size table

Interactions of the themes outlined in Table 1 are presented in Fig. 1. According to the depicted model, a weighing pan can be built for each user and sustainability decision. The example displays a user who can inform themselves but has lower environmental knowledge. The primary enabler for the user is their social group; the most significant hindering factor is the price. Factors could be switched in position (enabling or hindering) for another user or situation or appear on both sides. For each new decision concerning behavior, the scale can look different and swing to the other side. In this case, the user has performed the sustainable behavior since the enablers outperform the hindering factors. Analyzing different situations can help identify the main influencing factors across individuals. In the depicted example, unsustainable behavior emerges from the predominating right side, while a predominating left side results in sustainable behavior. Based on the model and the factors included, five qualitative personas were created.

Fig. 1: Behavioral weighing pan of sustainability.
figure 1

The depicted example shows how the prerequisites, enablers, and hindering reasons play together, thereby emphasizing the significance of portraying mutual aspects of each entity rather than focusing solely on absolute values.

Full size image

The Activist persona is 25 years old, male, and works in the solar industry. Typical statements from him include, “I am willing to spend more money on sustainable products.”, “I am very well able to inform myself about sustainability.”, and “I feel addressed by sustainable products and question them concerning greenwashing.” Factors influencing this persona are values, social contacts, political opinions, and belief in climate change.

The Sustainability-interested persona is 46 years old, female, has one child, and works as a florist. Typical statements from her include, “I can inform myself about sustainability, but I do not have enough time to do so.”, and “I feel addressed by sustainable products.” Values, social contacts, and time pressure mainly influence this persona.

The Hedonistic persona is a 30-year-old female who works as a cultural manager. Typical statements from her include, “I am attracted to products that I enjoy.”, “I can inform myself about sustainability, but I do not have enough time to do so.” and “I am only willing to spend more money on sustainability if it pays off for me, e.g., through money or fun.” Factors influencing this persona are convenience, monetary incentives, and time pressure.

The Indifferent persona is 37 years old, male, and works as an event technician. Typical statements from him include, “I am not willing to spend more money on sustainable products” and “I find it hard and not interesting to inform myself about sustainable products.” Factors influencing this persona are convenience, social contacts, monetary incentives, and specific situations.

The Dismissive persona is a 65-year-old retired male with two grown children. Typical statements from him include, “I am attracted to products that I enjoy.”, “I am not willing to spend more money on sustainability.”, and “I am not influenced by the opinions of others about sustainability”. Convenience, social contacts, and monetary incentives mainly influence this persona.

The personas, enablers, and barriers were used to select validated scales from the literature to enable the collection of quantitative data for executing a more formalized clustering related to behavioral prerequisites underneath.

Part 2: Validation of sustainable personas by clustering

Building on the results of the first part, we selected the features social status, trust, skepticism, economic benefit, and care for sustainability for clustering. Clustering the data with HDBSCAN (Campello et al., 2013) provided a silhouette coefficient of \({Sc} < =.0583\), and the algorithm clustered 326 from the total of 342 samples. In general, a silhouette score can range from a perfect representation \(\left({Sc}=+1\right)\) to a wrong representation \(\left({Sc}=-1\right)\) (Shahapure & Nicholas, 2020). Based on the clustering results, a decision tree was used to explain cluster descriptions. Based on a grid search, the tree’s preferable depth was set to 5. The five layers pruned decision tree resulted in an entropy of \(H > =0.91\) and a Gini coefficient (Breiman et al., 2017) of \({G}^{{ML}}=83.5\). While an entropy value close to \(H=1\) is considered ideal (Celeux & Soromenho, 1996), a value \(H > =0.80\) is considered acceptable. Figure 2 depicts one selected path from the decision tree as example, and the complete decision tree can be found in the supplement.

Fig. 2: Exemplary path from process for applying the decision tree algorithm.
figure 2

A decision tree based on the clustering results shows which factors lead to the clustering of certain groups. Clusters can include multiple paths of a decision tree. The averages of the scales are scaled to values between 0 and 1.

Full size image

Table 2 outlines scaled results of the five clusters derived from the decision tree in terms of attitudes, values, and behavioral implications. Results for the post hoc Tukey’s HSD analysis are presented in Table 3, exploring significant differences between clusters.

Table 2 Descriptive characteristics of sustainability clusters according to selected validated scales.
Full size table
Table 3 Pairwise comparisons of differences between clusters related to overarching constructs.
Full size table

The Socially Sustainable Cluster (1) comprises 39 participants with strong beliefs in human-made climate change, concern for sustainability, engagement, perceived consumer effectiveness, environmental knowledge, trust in eco-labels, social status, social norms, environmental responsibility, and conservation values. Participants assigned to the Social Sustainable Cluster display a strong awareness of sustainable consumption across all areas of the CSC. They are motivated to protect the environment and recognized by their social peer group for exerting responsible behavior. Additionally, this cluster is more tech-savvy and collaborative than the average consumer and shows low skepticism toward sustainable labels.

The Responsible Savers Cluster (2) comprises 86 participants with a strong believe in human-made climate change, environmental responsibility, and conservation values. This cluster also values simplicity, collaboration, and debt-free consumption. Participants assigned to the Responsible Saver Cluster prioritize sustainable purchases and are motivated by environmental concerns, with a high level of knowledge about sustainable products. They are more sustainable and less tech-savvy than the average consumer.

The Unconcerned Spenders Cluster (3) comprises 84 participants exhibiting low beliefs in human-made climate change, care for sustainability, and environmental responsibility. This cluster is further characterized by valuing hedonism and spending money over sustainability, with little awareness of sustainable consumption practices. Participants assigned to the Unconcerned Spenders Cluster show average values for social status and trust in eco-labels but low values for all other subscales, including preference for socially sustainable products, collaboration, simplicity, and debt-free consumption. This cluster demonstrates little environmental concern and a focus on immediate gratification rather than long-term sustainability.

The Comfort-Oriented Cluster (4) cluster comprises 62 participants exhibiting the lowest engagement levels across all clusters and further scoring low on environmental concern, knowledge, and technical affinity. Furthermore, the cluster is characterized by below-average values for openness to change, belief in human-made climate change, care for sustainability, and perceived consumer effectiveness. Additionally, participants assigned to the Comfort-Oriented Cluster display slightly above-average values for social status, social norms, environmental responsibility, self-enhancement, skepticism towards sustainable labels, and trust in eco-labels. The Comfort-Oriented Cluster reports a low awareness of sustainable consumption and identifies money and comfort as the primary hindering reasons toward exerting sustainable behavior.

The Skeptical Consumer Cluster (0) comprises 55 participants with the lowest values across all clusters for skepticism towards sustainable labels, trust in eco-labels, and care for sustainability. They also indicate below-average values for engagement, perceived consumer effectiveness, social norms, affinity for technology, conservation values, and beliefs in human-made climate change. Participants assigned to the Skeptical Consumers Cluster display average values for environmental concern and rank slightly above average for environmental knowledge, environmental responsibility, openness to change, self-enhancement, and self-transcendence. The Skeptical Consumers Cluster is not motivated to act sustainably by environmental concerns or social recognition and exhibits the highest cost values as a barrier to adopting sustainable behavior.

The qualitative-derived personas from the first part to the quantitative clusters from part two are mapped in the following. First, the Hedonists compare well with the Unconcerned Spenders, while the Skeptical Consumers roughly correspond to the Dismissive Cluster. The Indifferent persona corresponds to the Comfort-Oriented persona, and the Activists and Interests are distributed among the Socially Sustainable and Responsible Savers clusters. Generally, some contributing factors can be found in both analysis steps, but not all descriptively emerging aspects were included in the formal clustering process. For example, barriers like time pressure or stress were not part of this inspection. When considering all clusters, the factors of costs and convenience emerged as the most significant obstacles to behavior change.

Discussion

Considering the substantial impacts of private household emissions on climate change, understanding the drivers of individual sustainable behavior gains increasing importance. Consequently, our research aimed to advance understanding of users’ thinking and acting related to sustainability practices and concerns. We conducted both qualitative interviews and a quantitative survey based on validated scales. Diverse influencing factors among individuals were identified as a critical challenge. To address this, we focused on clustering users based on their sustainable intentions and derived five distinct personas and conceptually related formalized clusters. Both descriptive personas and formal clusters can inform product development and communication strategies to promote sustainable behavior. Therefore, our findings provide a practical framework for translating user insights into actionable steps.

Overview of key findings

Critical factors of sustainable behavior emerged from the conducted qualitative semi-structured interviews. More precisely, responses highlight values, social norms, belief in climate change, and barriers such as price sensitivity, convenience, and tangibility as primary contributing factors. Based on these insights, we created five personas with unique perspectives on sustainability based on attitudes, behaviors, beliefs, and values: the Activist, Interested, Indifferent, Hedonist, and Dismissive. While the Activist actively promotes a more sustainable world and engages in sustainable behaviors, the Interested may invest less time in sustainable behavior but still practice it. By contrast, the Indifferent is not interested in the issue of sustainability, and the Hedonist is primarily interested in financial gains from their investments. Finally, the Dismissive is less likely to believe in human-caused climate change and may be resistant to adopting sustainable behaviors. By understanding these personas, efforts to engage and educate individuals about sustainability can be designed more effectively. Although these personas may appear stereotypical, they still hold descriptive value and provide directions for more thorough quantitative validation. It is vital to acknowledge that the personas presented serve as one potential cluster representation. As such, variables like age and gender serve as exemplary indicators and are based on the median of the clusters grounded in actual participant data.

We systematically assessed differentiating factors underneath the described personas using a quantitative survey. Therefore, we analyzed the obtained data for critical factors, including environmental concern, trust and skepticism for sustainable labels, economic benefit, and social status. These variables explained the highest variance in the data set, consequently serving as influential clustering factors. Building on interview insights and additional related research, we opted to use five clusters. Therefore, each data-driven cluster represents one persona, which differs regarding configuration and characteristics of influencing factors. In summary, the Unconcerned Spenders Cluster is more hedonistic and has higher scores on the economic benefit scale than the other clusters. By contrast, the Skeptical Consumer Cluster is less likely to believe in climate change and is skeptical about sustainable labels. In addition, the Socially Sustainable Cluster places a high value on its social image and is motivated to be sustainable as a result. The Responsible Saver Cluster is motivated by environmental awareness and has a deeper understanding of sustainable products. Finally, the primary motivators of the Comfort-Oriented Cluster are money and convenience.

From a methodological viewpoint, combining quantitative and qualitative data in mixed-methods research strengthens the robustness of the inference compared to using either approach alone. While qualitative research can add an in-depth understanding of the results, quantitative research can provide numerical data, offering possibilities for statistical analyses and generalization (Creswell, 2009). During the interviews, we noted that products require more personalized communication and product design strategies. Recognizing the emerging potential of individualized approaches, we aimed to validate the obtained insights through a subsequent structured survey. We carefully selected scales that had undergone rigorous validation to ensure the validity and robustness of our data. Upon inspection, we found a significant difference in the broad subtopics, such as values or hindering factors, among various users. To explore these subtopics further, we identified literature-based reliable measures that could be used in the second part.

Limitations

Taking on a critical perspective, our research faces several limitations. First, sample sizes were determined by reviewing the relevant literature, ensuring our results were scientifically rigorous and reliable (Dolnicar et al., 2013); Guest et al., 2006). However, it is essential to note that the observed outcomes may require further generalization to account for potential bias resulting from the sample characteristics. While our interviews aimed to identify and describe the factors influencing sustainable behavior, we faced capacity limits on the number of factors we could validate quantitatively. We also recognized a potential bias towards apparent factors and plan to explore subordinate factors in future studies. It is worth mentioning that the sample size for the validation part was appropriate, but having more samples could continuously improve the robustness and generalizability of the findings.

Moreover, the overall applicability of the current evidence might be slightly limited due to the use of an online access panel, the exclusion of participants, and the fact that the study was solely conducted in a German context. Given the chosen recruitment means, some participants may have only participated in our research for monetary reasons. To ensure the quality of the data, we set a shorter survey time limit during preregistration and excluded faster participants. Information about agency-specific recruitment methods for bias mitigation is available in online resources (Bidou, 2020; Lämmer, 2020). To exclude international dependencies, only German-speaking participants were accepted for the study, limiting the emerging personas to a particular set of national characteristics. Hence, other populations and contexts could be researched in the future.

While participating in the survey part, respondent fatigue is possible due to the survey’s length and matrix questionnaire format. To overcome this problem in future studies, we will utilize a short screening instrument to determine individual cluster assignments. In the development process, items with particularly high value for the explained variance were chosen. The underlying goal relates to assessing the effects of individualization measures on different user groups more easily by splitting samples in a construct-oriented manner. However, relying solely on self-reported intentions may not accurately predict actual behavior. Future research could measure user behavior by employing a broader methodological spectrum, including direct behavioral observations. On this base, personalized design strategies promoting sustainability could be implemented and tested. Randomized controlled experiments comparing individualized solutions could be an effective way to determine the effectiveness of selected strategies in promoting sustainable behavior change.

Regarding data exclusion, it is essential to note that excluding certain participants from a study is always prone to introducing selection bias into the data and lowering the sample’s representativeness. For example, excluding longer trials may systematically exclude parents who need to take breaks in the survey to supervise their children. Therefore, data exclusions may result in a biased depiction of the study’s target population, leading to distorted clusters. Identifying and removing participants with response patterns that do not provide valuable information but instead add noise and distortions to the dataset is essential. One such pattern is straight liners, where participants answer with the same response in every question, resulting in significantly shorter survey times than the sample average. To tackle this issue, we employed meaningful time limits and strategic analyses to identify and remove participants identified as straight-liners.

Lastly, it is essential to note that the variables do not necessarily capture the full complexity of the cluster. Nevertheless, the HDBSCAN algorithm identifies clusters with varying densities, regulates noise, and detects clusters with arbitrary shapes. In contrast, other algorithms, such as K-Means or DBSCAN, cannot handle noise or detect clusters with varying densities (McInnes et al., 2017).

Implications

The emerging personas gave user-centered insights into the characteristics of user groups. We have shown in qualitative and quantitative data that people possess diverse factors, including motivations, values, and the significance of social status, which could promote the adoption of sustainable behaviors. However, the described clusters are based on self-reported intentions, not actual behavior. Since the attitude-behavior gap is a recognized factor in psychological research (Claudy et al., 2013), these clusters should be validated based on behavior. In general, the attitude-behavior gap may be most prevalent in all sustainably oriented groups. Thereby, a lower awareness of sustainability might result in a smaller attitude-behavior gap. In particular, the Socially Sustainable Cluster may be prone to the attitude-behavior gap because of the higher influence of their peer group. Hence, it is essential to consider that other socially based biases can also influence whether individuals adopt sustainable strategies in everyday behavior. These insights ease the understanding and addressing of users and direct the focus to further differences in cognition, which could play a significant role.

We observed individuals making error-prone assumptions regarding high-impact behavior. High-impact behaviors concerning greenhouse gas emissions involve mobility, buildings (insulation and heating), and nutrition (Antony et al., 2020; Tukker & Jansen, 2006). Most participants wrongly assumed plastic bags consume more CO2 than high-impact behavior, such as insulated living spaces. Furthermore, participants wrongly assumed that more energy is used in the eco-program than in the standard washing program because it takes twice the time. Individuals with sustainable attitudes had those incorrect assumptions, too. However, they were less prevalent compared to participants who were less concerned about sustainability. Increasing environmental knowledge can be beneficial for behavior change (Frick et al., 2004) and is a possible behavioral intervention. Including education about behaviors with high CO2 emissions is crucial, even though knowledge alone is no significant predictor for sustainable behavior (Heeren et al., 2016). Hence, it is crucial to integrate knowledge with other effective strategies, such as modifications in choice architecture or socio-cultural transformations like cultivating an eco-surplus culture. This might entail promoting a shared set of pro-environmental attitudes, values, beliefs, and behaviors among a community to minimize adverse human-induced effects on the environment and preserve and rehabilitate nature (Nguyen & Jones, 2022). The extent of observed individual differences suggests that targeting different clusters with tailored strategies is promising. Therefore, individualized approaches according to the user’s key motivators must be developed and translated into practice, for example, through product design. An initial approach would involve the consideration of diverse group needs while providing products tailored to specific clusters. It might not be necessary to tailor the products entirely to each user if the requirements of diverse user groups are considered. Within this scope, mass individualization could play a role since it is also predicted to be the next paradigm in product design (Koren et al., 2015). Climate communication should also be targeted and tailored to the individual needs of the population (Bostrom et al., 2013). In communication, the focus could be on milieu-dependent role models to promote them. For some clusters, ecotainment could be used to address such purpose, as Reisch et al. (2016) have shown, for the early stages of consumer behavior adaptation.

Furthermore, the saturation of a specific behavior within the peer group is essential for individual behavior change, as known from diffusion research (Rogers, 1995). The importance of saturation could be especially true for the Socially Sustainable Cluster, as social norms influence this cluster in particular. Implementing small behavioral changes is crucial to enhance saturation. One possible approach is to change the behavior in small steps. The less an individual must change their behavior, the more likely the change will be adopted (Fogg, 2020). For instance, automated heating systems allow a straightforward adaption of the temperature on all radiators simultaneously, thus simplifying the regulation considerably.

Changing the behavior with small steps could be promoted by individualized product designs and communication. One way to analyze individualized behavior intervention relates to the COM-B model (Michie et al., 2011). Since the underlying approach allows to identify barriers and develop behavior-change interventions, it could be used to target specific clusters. Therefore, more hedonistic clusters could be addressed with interventions from the incentives category. Future research could investigate in detail which clusters respond particularly well to different interventions. On a related note, the change in the choice architecture relates to the behavior change wheel (Leal & Oliveira, 2020). In general, the choice architecture presents options or choices to the individual in a modified form to stimulate a specific behavior. Modifications include changing the default option, providing information or feedback, or simplifying complex choices. By changing the choice architecture, individuals can be encouraged into more sustainable choices or behaviors (Thaler & Sunstein, 2008). Choice architecture can also impact behavior change on a large scale (Marteau et al., 2021).

Individual differences provide insights for user-centered product development, services, and communication strategies that target segments on an individual level. Building on the obtained results, communication strategies and product development should prioritize individual factors like sociality or hedonism to promote sustainability. For example, a persona derived from the Socially Sustainable Cluster ranks high on social status and social norms scales. Simplified usage and feedback mechanisms can motivate Responsible Savers to behave more sustainably. While simplified usage benefits all clusters, the Comfort-Oriented Cluster would particularly benefit from simplification approaches as a predominant emphasis on comfort and money hinders them. Unconcerned Spenders may benefit from receiving feedback on monetary savings related to choosing sustainable strategies, considering their high levels of hedonism and spending. Similarly, Skeptical Consumers may respond positively to a product design that includes monetary feedback, allowing them to spend more money on enjoyable activities rather than unsustainable behavior like washing clothes at unnecessarily high temperatures. Engaging in dialog with Skeptical Consumers is especially important, as they may be open to change but put lower emphasis on mitigating climate change or trusting sustainable product labels. Furthermore, it may be promising to encourage perceived consumer effectiveness in less sustainable groups and then evaluate the impact.

While individualization may offer advantages in meeting specific user needs, it is essential to consider potentially related drawbacks. For example, individualization can create privacy issues regarding data collection, higher expenses, and restricting resources for both the user and the organizations involved. Creating individualized products and services can be more expensive than standardized ones, leading to higher costs for both the company and the consumer. Some companies may require more resources to create personalized products and services, limiting their ability to compete in the market. Additionally, creating customized products and services can take time and effort, slowing production and hindering innovation. Collecting and using individual data to create personalized products and services can raise consumer privacy concerns. When developing products and communication strategies, it is essential to keep these challenges in mind. Consequently, exploring ways to encourage sustainable behavior by individualization raises the necessity and offers promising avenues for further related research.

Future research

Addressing the previously outlined demands, future research should explore more thoroughly to what extent personalized interventions can promote sustainable behavior. Follow-up studies are necessary to validate the effectiveness of individualization for behavior change. Thereby, measuring actual behavior while robustly controlling for various factors is crucial. In addition, subsequent research should expand its scope to encompass other cultural contexts and larger sample sizes, enabling a more comprehensive understanding of the viability of personalized interventions. Given the dynamic nature of user behavior, regularly reviewing and updating segmentation variables holds essential value. Furthermore, to maintain accurate representations of the envisioned target audience (Brickey et al., 2012; Pruitt & Grudin, 2003), it is important to periodically reassess and refine the derived personas and related sustainability clusters in research and practice.

While individual behavior change is essential, it cannot foster the transition to a more sustainable world on its own. Governments, the private sector, civil society, and individuals must jointly create beneficial conditions for change. Examining derived strategies for product development, design, and communication for each cluster holds valuable potential for practical application. Communicators, product managers, and designers may evaluate the practical usefulness of the results. This evaluation could be done in a focus group, which applies the knowledge base related to the personas to a concrete product.

Conclusion

Climate change and related threats require immediate actions, not only by economic and political stakeholders but also by individuals in private households. Consequently, the outlined research attempted a holistic approach to systematically consider a broad range of individual factors, such as needs related to sustainable consumption, motivations, values, intentions, and social norms, allowing the development of more individualized interventions for behavior change. Our main contribution is the integration of individual needs into easily understandable and usable personas, which are backed up by data-driven clusters. The identified clusters and personas provide a starting point to bridge the gap between theory and practice, allowing to address the needs and characteristics of different user groups more systematically. Thereby, critical areas, such as high-impact sustainable behavior, could become a fundamental part of user-centered product development, design, and communication of products and services. In conclusion, by tailoring interventions to individuals, we can increase their motivation and engagement in sustainable practices, ultimately leading to a more environmentally conscious society.