- Original authors / creators
- Natalie Jane de Vries; Rodrigo Reis; Pablo Moscato
- Original publication
- PLoS ONE · PLOS · 2015-04-07
- License & retained notices
- CC BY 4.0
© 2015 de Vries et al
Article-specific license evidence ↗Exact license notice
https://creativecommons.org/licenses/by/4.0/This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited. Article-specific XML license URL: http://creativecommons.org/licenses/by/4.0/ - Edition changes
- Narrative text edition with normalized whitespace. Figures, tables, display equations and separately licensed material are omitted; consult the preserved XML and original publisher for definitive content.
- Evidence limits
- Adjacent research, not a directly sampled family-office study. Applicability to a family office requires separate assessment. Although this study provides literature and charity and not-for-profit marketing managers several implications and contributions, there are some limitations. Firstly, the data set used came with some limitations. As it was an online administered questionnaire, the system was able to direct respondents to various parts of the survey as well as skip questions that did not apply to some of them. This left several items uncompleted where some or the majority of respondents did not answer all questions. In order to be able to still use the entire dataset, we omitted those questions (variables) that not every respondent answered. This reduced the number of variables we were able to use in our clustering method and subsequent cluster analysis. Another limitation we have found with the dataset is the way the ‘Involvement Class’ is constructed. The ACNC assigned respondents to either the ‘low involvement class’ or the ‘high involvement class’ based on their responses in question 3. There were several questions that, if respondents answered ‘yes’, it would automatically classify them as ‘high involvement’. For instance, whether they had made regular monthly donations to a charity or sponsored an animal or a child via a charity. Although we do not argue that it is logical to assume certain activities are of a higher involvement level with a charity than others, we do argue that a respondent could feel like they are highly involved with a charity even though they had not answered ‘yes’ to these questions. One example of an activity classified as ‘low involvement’ by the ACNC is “Sold raffle tickets or conducted other fundraising on behalf of a charity”. We note, that if an individual supports a charity by selling raffle tickets once, we resonate with this claim, however, if a person spends a considerable amount of time doing so (for example, several days a week) it could also be reasonable to assume that this person is ‘highly involved’ with the charity they are supporting. For our study this means that the results found to define the involvement class could in the future be better defined if a better framework for classifying the ‘Involvement Class’ is designed. Instead of merging activity types into the dichotomy provided by ‘Low/High Involvement’ an alternative would be to seek mathematical models that could predict each one of the activities separately. The last limitation we highlight is a conscious “self-imposed” limitation during the analysis of the symbolic regression modeling. As stated, we only selected those models for analysis that included ‘simple’ functions used by Eureqa (the argument for the logistic model is a linear function). We have also explained that more complex models have found better results in terms of the Area-Under-the-Curve, which means that in future studies more effort could be exerted in analysing these findings. Our aim in this study was to provide the literature and the NFP sector with an easily-interpretable method for segmentation and subsequent analysis. It is for this reason we have imposed this limitation and suggest future research directions to explore this path further. Furthermore, future research could consider a mixed-method approach utilising a qualitative method such as in [2] combined with a quantitative method as outlined in this study for greater depth of results and to provide a more detailed analysis of each cluster or group of consumers. For example, in their qualitative research, Polonsky, Shelley and Voola [2] state that, “the strength of the stimulus related to personalized appeals and may create a link between the individual and the cause, which would increase donor’s support” [p.74]. Another future research suggestion is to conduct another study of a similar kind with a different questionnaire tool. In the current study, no questions about either actual monetary donations or time spent volunteering were included. These questions could help to more accurately gauge consumers’ behaviours and their attitudes towards charities and, therefore, help charities to understand their donor base even better by considering their individualities. Future research could also extend the study across multiple countries and/or cultures in order to find more generalisable results which can then be applied in more countries and areas. As stated, we have only used the best most simple Eureqa models. In the future, a new way could be constructed to automatically analyse all results found by Eureqa’s symbolic regression modelling method to assist the user of the clustering results in interpreting their findings. In this process, other Artificial Intelligence and Machine Learning methodologies could be employed to truly find the best underlying structure in the data. Eureqa employs a genetic programming technology which, combined with other types of algorithms to “learn from the data”, would lead to a new generation of more accurate models which, in turn, could provide more refined donor behaviour models. Taking this study even further, the novel concept of “functional construct” as introduced in an earlier publication [24], could be applied in the context of NFP sector research. For instance, the data in this study comprised of separate questions, or “items”, which could be analysed by a process as outlined in de Vries et al. [24] to find functional constructs for further behavioural modelling. This would provide literature and the charity and NFP sector with an even deeper understanding of the relationship that exists between consumers’ attitudes, trust and confidence in and the behaviours such as donating they actually display towards charities. These constructs could even form the basis for other methods such as Structural Equation Modelling (SEM) as in [26] to find the structural relationships between ‘constructs’ of items. In a study investigating the charity and NFP sector in particular, a SEM could investigate the relationships between trust and donating behaviour or confidence and trust in a charity or the relationship between trust and volunteering. These future research directions are bountiful and based on the results in this study we highly recommend the charity and NFP sector to continue their trend in this way to become more successful, less wasteful and ultimately, serve their true cause better and more efficiently. The clustering method in this study can lead to a data-driven marketing strategy as it allows charities to segment their donor market and then investigate their salient features with a relatively simple method that eases interpretation (the CM1 score). The technique is also highly scalable and can easily be applied to problems involving several millions of consumers [16]. Obviously, such large datasets are unlikely to come from questionnaire-based studies but with the increasing use of online communication channels, this is a realistic scenario and a research area worth pursuing. The findings related to the cluster analysis collectively provide guidance and a better understanding of a possible donor base for the NFP and charity sector. If charities take actions to drive donor involvement with a data-driven approach, higher levels of monetary donations and an increase in volunteering efforts will ensue. By following for-profit organisations and businesses in being more informed about the environment and market they operate in, charities can hope to thrive, be more successful in obtaining funds and ultimately, achieve greater success in the causes they are working towards.
- Retraction / correction check
- No linked notice identified in retrieved XML or Crossref relation/update-to fields. Article-specific repository XML related-article links, Crossref DOI relation/update-to fields, and a bounded publisher-page notice search where accessible; unavailable pages and unregistered/unlinked notices remain coverage limits. Verified 2026-11-05. Automated/source checks are not human peer review.
- Preserved record
- Rights and full text checked 2026-11-05. Included in this volume on 2026-11-05. Original source dates remain unchanged.
Download the preserved source
SHA-256 (retrieved artifact): 31bd7ec8b21f484babfe91a166f4ec6b6059352c7b32917c198465f1ad3b6f14
SHA-256 (preserved text edition): 9b39a9d8bc9d0c8ddf8a47bc6f36883ee0d077062e16126e8c260d6efccd0516