The dairy farming sector plays a key role in strengthening national food systems, particularly in developing economies where it contributes to animal protein supply, rural employment and local economic resilience (Mukson et al., Reference Mukson, Roessali and Setiyawan2014; Kurniawati et al., Reference Kurniawati, Ahmad and Dudarwa2016). In Indonesia, Central Java Province is one of the country’s major milk-producing regions, with Boyolali Regency serving as its primary production hub. Boyolali contributes approximately 20% of national milk demand and up to 72% of Central Java’s total output (Sukoco and Fauziah, Reference Sukoco and Fauziah2023). Beyond its economic importance, the sector also supports the province’s strategic ambition to become Indonesia’s ‘National Food Barn’ by 2026, as outlined in the Regional Development Plan (Pangestu et al., Reference Pangestu, Nuswantara and Christiyanto2023).
Despite this strategic role, the sustainability of dairy farming in Central Java faces increasing structural, environmental and socio-economic pressures. Dairy production remains highly dependent on climatic conditions, making farms vulnerable to prolonged dry seasons that reduce forage availability, limit water access and increase production costs (Sukoco and Fauziah, Reference Sukoco and Fauziah2023). These challenges are compounded by fragmented farm structures, limited technological adoption and reliance on conventional management practices. Demographic change further threatens sectoral continuity, as most dairy farmers are over 54 years old, while youth participation remains low (Budiman and Irwandi, Reference Budiman and Irwandi2020; Kurnadi et al., Reference Kurnadi, Nurlaila, Heriyadi and Ramadhan2024; Wianto and Nurmastiti, Reference Wianto and Nurmastiti2025). Together, these factors undermine long-term resilience and intergenerational renewal within the local dairy system.
Recent policy developments have further intensified these challenges. In November 2024, protests by dairy farmers in Central Java highlighted growing dissatisfaction with low milk absorption rates by dairy processing industries, driven largely by import-oriented procurement policies (Cahyaningrum and Zanggi, Reference Cahyaningrum and Zanggi2024; detikJateng, 2024). These conditions have resulted in significant economic losses for farmers and raised broader concerns regarding the sustainability of domestically based dairy production systems.
Addressing these challenges requires sustainability strategies that are both conceptually robust and operationally applicable. While sustainability is widely recognised as a guiding principle for agricultural development (WCED, 1987), its practical implementation remains challenging due to contextual variability and differing stakeholder priorities (Kemp and Martens, Reference Kemp and Martens2007; Seghezzo, Reference Seghezzo2009). In agriculture, sustainability assessments, therefore, depend heavily on the selection of indicators that appropriately capture locally relevant economic, environmental and social dynamics (Hansen, Reference Hansen1996).
In the context of livestock systems, sustainability assessment is further complicated by the involvement of living beings, making animal welfare an essential dimension alongside economic, environmental and social considerations (D'Silva, Reference D'Silva, Behnassi, Pollmann and Kissinger2013). Governance also plays a critical role, influencing system resilience, policy coherence and the capacity to respond to external shocks (Duru and Therond, Reference Duru and Therond2015). Consequently, sustainability indicators for dairy farming must be multidimensional while remaining context-specific, practical for end users and feasible to implement (Lebacq et al., Reference Lebacq, Baret and Stilmant2013).
Various approaches have been proposed for developing sustainability indicators, including theory-driven, stakeholder-driven and expert-driven methods (Binder et al., Reference Binder, Feola and Steinberger2010). Theory-driven approaches ensure conceptual coherence, stakeholder-driven approaches enhance local relevance and expert-driven approaches support feasibility and technical validity. Increasingly, studies suggest that integrating these approaches yields more robust and legitimate indicator systems (Reed et al., Reference Reed, Fraser, Morse and Dougill2005; Domingues et al., Reference Domingues, Lozano and Ramos2018).
The Delphi method provides a structured and iterative framework for such integration, enabling systematic consensus-building among panellists while allowing iterative refinement based on feedback (Novakowski and Wellar, Reference Novakowski and Wellar2008). Delphi has been widely applied in agricultural sustainability research, including dairy farming, to identify and prioritise sustainability indicators (Bélanger et al., Reference Bélanger, Vanasse, Parent, Allard and Pellerin2012; Chopin et al., Reference Chopin, Mubaya, Descheemaeker, Öborn and Bergkvist2021). However, existing studies often focus on limited sustainability dimensions, apply single-round designs or lack sufficient contextual grounding, particularly with respect to governance and local production systems.
Against this background, this study aims to develop a comprehensive and context-sensitive sustainability indicator system for smallholder dairy farming in Boyolali Regency, Indonesia. Using a multi-round Delphi approach, the study integrates theoretical foundations, expert consensus and stakeholder insights to identify indicators that are both scientifically sound and practically applicable. The resulting indicator set is intended to support farm-level decision-making and inform policy interventions aimed at strengthening the long-term sustainability and resilience of local dairy farming systems.
Materials and methods
Research location
This study was conducted in Boyolali Regency, Central Java Province, Indonesia, one of the country’s main centres of smallholder dairy farming. The regency contributes approximately 72% of Central Java’s milk production and around 20% of national output, earning it the nickname ‘New Zealand of Java’ (Sukoco and Fauziah, Reference Sukoco and Fauziah2023). Dairy farming in Boyolali is dominated by small-scale household farms and supported by a relatively high dairy cattle population. According to data from the Central Bureau of Statistics, Boyolali had the highest dairy cattle population in Central Java in 2024, estimated at approximately 55,000 head, compared with 15,145 head in neighbouring Semarang Regency (Fig. 1).
Distribution of livestock population in Boyolali and surrounding regencies, 2024. Source: Central Java provincial statistics agency. Available online at https://jateng.bps.go.id/id (accessed on 20 July 2025).
These production characteristics are closely linked to the region’s agroclimatic conditions. Boyolali is characterised by highland environments, with elevations ranging from approximately 75 to 1,500 m above sea level and an average annual rainfall of about 2,000 mm (Central Bureau of Statistics of Boyolali Regency, 2014). Such conditions support forage availability and contribute to the suitability of the area for dairy production.
Delphi method
The Delphi method is a consensus‐based decision-making technique initially developed by Norman Dalkey and Olaf Helmer in the 1960s at the RAND Corporation for technology forecasting in conflict-related contexts (Ahmad and Wong, Reference Ahmad and Wong2019; Saptono et al., Reference Saptono, Mahmud, Pratiwi, Purwanto, Khozen, Aditama, Khodijah, Wayan, Asmara and Jie2023). Over time, this method has been widely applied to formulate solutions to complex problems, particularly when empirical data are limited or when integrating perspectives from diverse stakeholders is necessary (Novakowski and Wellar, Reference Novakowski and Wellar2008; Carrera and Mack, Reference Carrera and Mack2010). The Delphi process is conducted anonymously and free from direct influence, enabling the synthesis of panellist opinions that is more objective, reflective and not dominated by any particular individual (Cloquell-Ballester et al., Reference Cloquell-Ballester, Cloquell-Ballester, Monterde-Díaz and Santamarina-Siurana2006; Bélanger et al., Reference Bélanger, Vanasse, Parent, Allard and Pellerin2012).
In this study, the Delphi method is employed to construct indicators and formulate sustainability strategies for the dairy farming sector in Boyolali Regency. To achieve this objective, the approach adopted is the regulatory Delphi, which aims to establish consensus regarding ideal future conditions based on current knowledge (Novakowski and Wellar, Reference Novakowski and Wellar2008). This type of Delphi is particularly relevant for ex ante studies involving the formulation of policies or new standards, especially those related to sustainability issues (Saptono et al., Reference Saptono, Mahmud, Pratiwi, Purwanto, Khozen, Aditama, Khodijah, Wayan, Asmara and Jie2023).
The Delphi method is based on three core principles: iteration, anonymity and controlled feedback. In this study, iteration was implemented through successive questionnaire rounds, with results from each round used to refine subsequent instruments until an acceptable level of consensus was achieved (Hsu and Sandford, Reference Hsu and Sandford2007). Anonymity was maintained throughout to minimise dominance effects and social pressure, allowing panellists to express independent judgements (Bélanger et al., Reference Bélanger, Vanasse, Parent, Allard and Pellerin2012; Ahmad and Wong, Reference Ahmad and Wong2019). Controlled feedback was provided between rounds in the form of aggregated responses, enabling panellists to reconsider their assessments in light of collective views (Powell, Reference Powell2003; Bélanger et al., Reference Bélanger, Vanasse, Parent, Allard and Pellerin2012).
These three core principles underscore the importance of designing a careful and systematic process to effectively build consensus based on panellist judgment without disregarding the diversity of perspectives (Nakatsu and Iacovou, Reference Nakatsu and Iacovou2009). To accomplish this, the present study adopts the regulatory Delphi stages formulated by Novakowski and Wellar (Reference Novakowski and Wellar2008). Following a theory-driven approach, the initial Delphi stage begins with the development of a questionnaire based on a literature review to identify and classify sustainability indicators for dairy farming into five key dimensions: economic, environmental, social, governance and animal welfare.
The questionnaire was then distributed online to the panellists in multiple rounds. Although the fundamental principle of Delphi is iterative, practical constraints such as time limitations, budget considerations and potential panellist fatigue must be taken into account (Rowe, Reference Rowe1994; Hasson et al., Reference Hasson, Keeney and McKenna2000). Therefore, this study implemented two Delphi rounds, which have also been deemed sufficient in various previous studies (Bélanger et al., Reference Bélanger, Vanasse, Parent, Allard and Pellerin2012; Trinh Hai et al., Reference Trinh Hai, An Thinh, Anh Tuan, Dinh Cham, The Anh, H, Ha, Bao, Van Huong, Khanh, Mai, Tuan Tong, Hai and Truong2015; Fefer et al., Reference Fefer, De-Urioste Stone, Daigle and Silka2016). The first round combined expert-driven and stakeholder-driven approaches, focusing on the evaluation of initial indicators and the identification of new indicators through open-ended questions, interviews and focus group discussions (FGDs). The second round aimed to build consensus on the set of indicators compiled from both literature and participatory processes.
The entire Delphi process was conducted over approximately 1 month and 23 days, from late May to late July 2025 (see Fig. 2 for the implementation timeline). The stages of this method are further elaborated in the subsequent section.
Delphi method implementation process.
Figure 2 Long description
The flowchart outlines the Delphi method implementation process over several stages from May to July 2025. The first stage, ‘Literature Review’ from 29 May to 13 June 2025, involves the identification of secondary dimensions and detailing indicators by dimension. The second stage, ‘Questionnaire Development’ from 13 to 16 June 2025, includes the development of questionnaire items and duplication through Google Forms. The third stage, ‘Recruitment of Resource Persons’ from 16 to 17 June 2025, involves the identification of resource persons via Google Scholar, review of local government officials’ websites and compilation of names and contact details of resource persons from local government officials. The fourth stage, ‘First Delphi Round’ from 17 June to 4 July 2025, includes the distribution of the questionnaire, quantitative analysis of first-round results, PGD and interview and revision of the questionnaire. The final stage, ‘Second Delphi Round’ from 4 to 18 July 2025, involves the distribution of the revised questionnaire, selection of consensus greater than 80 percent and discussion of research findings.
Questionnaire development
The questionnaire in this study was developed based on five core dimensions of sustainability: economic, environmental, social, governance and animal welfare. These dimensions were selected because they most accurately represent the multifaceted nature of sustainability in dairy farming enterprises, particularly during periods of heightened vulnerability. Each indicator in the questionnaire was assessed using three criteria: (1) practical usefulness for dairy farm management, (2) appropriateness to local conditions and (3) cost efficiency of implementation. This approach aims to generate indicators that are conceptually robust and practically relevant for both industry practitioners and policymakers. In total, the questionnaire comprises 82 indicators: 17 each for the economic, environmental and social dimensions; 15 for governance; and 16 for animal welfare. The operational definitions and literature sources for each indicator are presented in Table A1.
The economic indicators were developed based on empirical references, particularly Sánchez-Hidalgo and Tadich (Reference Sánchez-Hidalgo and Tadich2024) and Munyaneza et al. (Reference Munyaneza, Kurwijila, Mdoe, Baltenweck and Twine2019), who emphasise the importance of accounting indicators, production costs, profitability and operational efficiency. Galioto et al. (Reference Galioto, Paffarini, Chiorri, Torquati and Cecchini2017) further demonstrate the relationship between production costs and milk output per litre as a measure of economic efficiency. Meanwhile, Feil et al. (Reference Feil, Do Amaral, Walter, Bagatini, Schreiber and Maehler2023) highlight operational expenditures, technological adoption, contributions to regional development and the creation of indirect employment along the supply chain. To reflect the local context of Boyolali, additional indicators – such as access to tax incentives and fiscal support – were incorporated, acknowledging that fiscal issues play a critical role in shaping farm-level sustainability.
The environmental dimension was constructed using literature that addresses sustainability indicators in dairy farming systems. Many indicators were adapted from Munyaneza et al. (Reference Munyaneza, Kurwijila, Mdoe, Baltenweck and Twine2019), Sánchez-Hidalgo and Tadich (Reference Sánchez-Hidalgo and Tadich2024), Feil et al. (Reference Feil, Do Amaral, Walter, Bagatini, Schreiber and Maehler2023) and Bélanger et al. (Reference Bélanger, Vanasse, Parent, Allard and Pellerin2012). The core themes include land, water, energy and waste management, as well as biodiversity conservation. Munyaneza et al. (Reference Munyaneza, Kurwijila, Mdoe, Baltenweck and Twine2019) underscore livestock waste management, water conservation and greenhouse gas emissions. Sánchez-Hidalgo and Tadich (Reference Sánchez-Hidalgo and Tadich2024) add biodiversity protection, fertiliser and pesticide use, and habitat preservation. Feil et al. (Reference Feil, Do Amaral, Walter, Bagatini, Schreiber and Maehler2023) emphasise energy efficiency, water recycling and environmental governance across the supply chain. Bélanger et al. (Reference Bélanger, Vanasse, Parent, Allard and Pellerin2012) complement these aspects with soil quality indicators such as nutrient balance, cover cropping and watershed protection.
The social indicators were developed to represent various aspects of social well-being among dairy farmers. Feil et al. (Reference Feil, Do Amaral, Walter, Bagatini, Schreiber and Maehler2023) highlight labour issues, training, social engagement and transparency regarding social and environmental impacts. Munyaneza et al. (Reference Munyaneza, Kurwijila, Mdoe, Baltenweck and Twine2019) add indicators such as workload distribution, gender equity, labour protection and participation in farmer organisations. Sánchez-Hidalgo and Tadich (Reference Sánchez-Hidalgo and Tadich2024) broaden the dimension by including quality of life, generational renewal, labour intensity and cultural diversity in farming practices. Silvestri et al. (Reference Silvestri, Silvestri and Ruggieri A2024) emphasise that the social dimension encompasses health and safety, protection from exploitation and contributions to cultural and regional development.
The governance dimension was formulated based on Duru and Therond (Reference Duru and Therond2015), who emphasise participatory management, experiential learning, farmer networks and connections with agri-food supply chains as essential components of livestock system resilience. Governance indicators include participation in decision-making, adaptive procedures and external collaboration. Several indicators were also drawn from the social and economic dimensions, such as financial record keeping (Munyaneza et al., Reference Munyaneza, Kurwijila, Mdoe, Baltenweck and Twine2019), digitalisation of farm operations (Feil et al., Reference Feil, Do Amaral, Walter, Bagatini, Schreiber and Maehler2023), collective action and community engagement (Sánchez-Hidalgo and Tadich, Reference Sánchez-Hidalgo and Tadich2024). Camanzi et al. (Reference Camanzi, Arba, Rota, Zanasi and Malorgio2018) contribute the concept of relational governance in supply chains, including transparency, contract management, performance evaluation and mitigation of opportunistic behaviour. All indicators were developed with reference to good governance principles outlined by Gisselquist (Reference Gisselquist2012) – transparency, accountability, compliance, participation, evaluation, risk management and legal documentation – adjusted to the context of Indonesia’s dairy farming sector.
The animal welfare dimension was structured using literature that identifies animal treatment as a fundamental pillar of dairy farm sustainability. Sánchez-Hidalgo and Tadich (Reference Sánchez-Hidalgo and Tadich2024) propose indicators such as body condition, water quality, housing hygiene and social behaviour. Bertocchi et al. (Reference Bertocchi, Fusi, Angelucci, Bolzoni, Pongolini, Strano and Lorenzi2018), through a Delphi-based approach, highlight the importance of housing management, hoof health and the use of analgesics to reduce pain. de Mori et al. (Reference de Mori, Mercugliano, Cerizza, Biasetti, Florio, Da Re and Gili2025) add regulatory compliance, staff training and the management of health, safety and nutrition. Silvestri et al. (Reference Silvestri, Silvestri and Ruggieri A2024) link animal welfare to environmental aspects, including veterinary expenditures and the facilitation of natural behaviours. Based on this synthesis, the indicators in this dimension focus on health, comfort, behavioural freedom and positive human–animal interactions.
Identification of potential panellists
The success of the Delphi method largely depends on the quality of the panellists involved (Hasson et al., Reference Hasson, Keeney and McKenna2000). Representativeness of the general population is not required; instead, what matters is the informants’ substantive expertise and their direct relevance to the topic under investigation (Keeney et al., Reference Keeney, Hasson and McKenna2001). Therefore, purposive sampling – not random sampling – is considered the most appropriate technique (Kennedy, Reference Kennedy2004), as it enables the selection of individuals with deep knowledge and a comprehensive understanding of the research topic (Goodman, Reference Goodman1987; Powell, Reference Powell2003). Panellists must also possess adequate communication skills (Rådestad et al., Reference Rådestad, Jirwe, Castrén, Svensson, Gryth and Rüter2013), given that the Delphi technique relies heavily on written responses that must be articulated clearly and logically to allow accurate interpretation by the researchers.
Accordingly, the selection of Delphi panellists must consider criteria that ensure competency. Previous studies recommend indicators such as years of professional experience and functional positions (Hsu et al., Reference Hsu, Chang and Luo2017). Musa et al. (Reference Musa, Mohammad, Thirumoorthi, Moghavvemi and Kasim2015) suggest that panellists should have at least 5 years of relevant experience and hold a managerial or middle-management position. Educational background is another important factor. Fefer et al. (Reference Fefer, De-Urioste Stone, Daigle and Silka2016) emphasise academic qualifications as an indicator of analytical capacity. In line with this, Ahmad and Wong (Reference Ahmad and Wong2019) and Saptono et al. (Reference Saptono, Mahmud, Pratiwi, Purwanto, Khozen, Aditama, Khodijah, Wayan, Asmara and Jie2023) require panellists to possess at least an undergraduate degree.
Diversity among panellists is essential to ensure that the Delphi deliberation process captures a broad range of perspectives and reduces potential bias (Adler and Ziglio, Reference Adler and Ziglio1996). The resulting consensus should, therefore, be inclusive rather than dominated by a single group (Donohoe and Needham, Reference Donohoe and Needham2009; Martins et al., Reference Martins, Kalil and Dalla Rosa2021). For this reason, the Delphi panel should involve not only academic or technical experts but also other relevant stakeholders (Brodbeck et al., Reference Brodbeck, Kerschreiter, Mojzisch, Frey and Schulz‐Hardt2002), whose inputs are crucial for ensuring that the identified indicators are practical and grounded in real-world conditions. Their participation must be supported by a strong understanding of the local context (Han et al., Reference Han, Kalantari and Rajabifard2023; Renthlei and George, Reference Renthlei and George2025). Including potential end-users – as recommended by Linstone and Turoff (Reference Linstone and Turoff1975) – helps strengthen engagement and commitment throughout the consensus process, which is a key success factor in Delphi studies (Rådestad et al., Reference Rådestad, Jirwe, Castrén, Svensson, Gryth and Rüter2013).
Based on these principles, panellists in this study were purposively selected to represent two main groups: experts and key stakeholders in Boyolali Regency. According to Lim and Antony (Reference Lim and Antony2016), experts are individuals with specialised knowledge, professional roles or a strong scientific track record. In this study, the expert group included officials from the Department of Livestock and Fisheries (Disnakkan), veterinary professionals from the Animal Health Center (Puskeswan), as well as academics and researchers in livestock science. Disnakkan officials were included to ensure alignment between indicators and regional policy priorities, while veterinary experts contributed insights into animal health and welfare (Bertocchi et al., Reference Bertocchi, Fusi, Angelucci, Bolzoni, Pongolini, Strano and Lorenzi2018). Academics and researchers grounded the indicators within broader sustainability frameworks. Stakeholders were defined as local actors directly affected by the study outcomes (Renthlei and George, Reference Renthlei and George2025). Here, they were represented by dairy farmers and board members of the Village Unit Cooperative (KUD), whose involvement ensures that practical field-level challenges and operational realities are fully reflected in the consensus-building process (King et al., Reference King, Gunton, Freebairn, Coutts and Webb2000).
A total of 21 panellists were invited to participate, and 20 agreed. Of these, 19 responded in the first round, yielding a 95% response rate. This level is consistent with common practice in Delphi studies. Previous research indicates that 9–13 panellists are sufficient for indicator development (Quyên, Reference Quyên2014; Henning and Jordaan, Reference Henning and Jordaan2016). However, in studies involving heterogeneous groups – such as a combination of experts and stakeholders – up to 20 participants are considered ideal to capture diverse viewpoints (Geist, Reference Geist2010; Hasson and Keene, Reference Hasson and Keene2011).
The complete profile of panellists is presented in Table 1. All panellists had a minimum of 5 years of professional experience. The expert group consisted predominantly of officials from Disnakkan and veterinary professionals from Puskeswan. Their expertise covered livestock policy and programme management, livestock business and production management, animal health regulation and surveillance, animal disease monitoring, as well as veterinary diagnostics and treatment. The academic panellist contributed specialised expertise in livestock science and sustainability research. The stakeholder group comprised dairy farmers and a cooperative board chairperson, whose areas of expertise related to smallholder dairy operations, farm management, milk production, cooperative governance and milk marketing.
Profile of panellists involved in the study
Table 1 Long description
Participant profile data for 19 panellists are listed by ID, affiliation, years of experience, academic qualification, position or role, area of expertise, gender, and age. Affiliations include 2 dairy farmers, 1 cooperative board head, 7 Disnakkan officials, 7 Puskeswan staff, and 1 academic. Experience is predominantly high: 8 participants report more than 15 years, 7 report 10 to 15 years, and 4 report 5 to 10 years. Qualifications are mainly bachelor’s degrees (14), with 4 master’s degrees and 1 doctorate. Roles range from owners and senior department or division heads to coordinators, veterinary experts, an extension officer, and a full professor. Expertise clusters around animal health monitoring and regulation, veterinary diagnostics and treatment, livestock production and business development, cooperative governance and milk marketing, and smallholder dairy operations. Gender includes 10 females and 9 males, and ages span 35 to 59 years, with the youngest at 35 and the oldest at 59. Interpret results cautiously because the table describes participant characteristics only and does not indicate sampling method or representativeness.
Most panellists (74%) held undergraduate degrees, with one holding a doctoral degree and serving as a full professor in livestock science. In terms of gender, 53% were male and 47% female, indicating relatively balanced representation. Based on age, 37% belonged to Generation X (45–60 years), while 63% were Millennials (29–44 years). These characteristics are noteworthy, as age and gender can influence perceptions of sustainability issues (Gifford and Nilsson, Reference Gifford and Nilsson2014).
First round of the Delphi process
In the first round, the researcher distributed a questionnaire consisting of both closed-ended (modified Delphi) and open-ended (traditional Delphi) questions, following the recommendations of Kerlinger (Reference Kerlinger1973) and Custer et al. (Reference Custer, Scarcella and Stewart1999). The closed-ended items were used to assess the relevance of sustainability indicators for dairy farming derived from the literature review, using a 4-point Likert scale (1 = very irrelevant to 4 = very relevant) to encourage decisive responses without a neutral option (Garland, Reference Garland1991). The open-ended questions allowed panellists to propose additional indicators. Panellists were given 1 week to complete the questionnaire. A total of 19 out of 20 panellists returned the questionnaire, resulting in a 95% response rate. The draft questionnaire for this round is available in the Supplementary Materials.
After receiving the responses from the 19 panellists, quantitative analyses were carried out using mean and median values to identify the central tendency of panellists’ positions along the response continuum (Novakowski and Wellar, Reference Novakowski and Wellar2008). A minimum threshold of 2.5 was set for both measures to distinguish positive (3 and 4) from negative (1 and 2) assessments, while anticipating potential negative bias associated with an even-numbered scale (Johns, Reference Johns2005). In addition, the percentage of responses in the two highest categories (3 and 4) and the interquartile range (IQR) were calculated. Following Huang et al. (Reference Huang, Yang, Inderawati and Sukwadi2022), indicators were required to obtain at least 70% of responses in categories 3 and 4. Consistent with Raskin (Reference Raskin1994) and Rayens and Hahn (Reference Rayens and Hahn2000), this positive assessment was considered homogeneous if the IQR ≤ 1. Indicators that failed to meet any of these criteria were eliminated. New indicators proposed through the open-ended items were added and further analysed in the next stage.
Following recommendations by Bélanger et al. (Reference Bélanger, Vanasse, Parent, Allard and Pellerin2012), this study conducted an FGD to consult the preliminary results of the first round questionnaire with a subset of panellists. The FGD served as a complementary technique to the Delphi method, which is often considered insufficient to capture the evolving dynamics of ideas in group interactions (Clément and Madec, Reference Clément and Madec2006). As noted by Rigby et al. (Reference Rigby, Woodhouse, Young and Burton2001), FGD facilitates the identification of implicit issues within the Delphi process and provides an opportunity for panellists to elaborate on and qualify their responses, thereby enriching the interpretation of quantitative data (Krueger, Reference Krueger1988). Of the 19 panellists, 5 individuals from the Disnakkan (DP1 to DP5) agreed to participate in a face-to-face FGD, becoming the only group involved in this session.
Second round of the Delphi process
The second round was conducted to achieve consensus among the panellists (Jairath and Weinstein, Reference Jairath and Weinstein1994). At this stage, panellists were asked to indicate their agreement (Yes/No) regarding the indicators retained from the first round and to assess the relevance of new indicators generated from open-ended responses, the FGDs and interviews, using a 1–4 Likert scale. The scoring procedure for relevance followed the same approach as in the first round. Of the 19 panellists, only 16 returned the questionnaire despite 3 reminders sent over a 2-week period, resulting in a response rate of 84%. The draft questionnaire for this round is available in the Supplementary Materials.
After receiving feedback from the 16 panellists, consensus for each indicator was determined based on a minimum agreement level of 80% (at least 13 out of 16 panellists), following recommendations for panels with more than 5 members (Lynn, Reference Lynn1986; Davis, Reference Davis1992). This approach is one of the most commonly applied in Delphi studies (Jairath and Weinstein, Reference Jairath and Weinstein1994; Von Der Gracht, Reference Von Der Gracht2012). However, the use of percentage agreement as a measure of consensus has been criticised for its susceptibility to chance agreement (Wynd et al., Reference Wynd, Schmidt and Schaefer2003). Therefore, the agreement levels in the present study were interpreted cautiously and adjusted to account for the possibility of chance agreement.
To more rigorously assess consensus, this study adopted the modified kappa statistic (
${k^*}$) developed by Polit et al. (Reference Polit, Beck and Owen2007). This statistic considers only agreement, indicating that an indicator is relevant. Agreement on irrelevance is excluded, as it does not contribute to the evaluation of indicator validity. The calculation of 
${k^*}$ begins by estimating the probability of chance agreement (
${P_{\text{c}}}$) using the binomial distribution formula:
\begin{equation}{P_{\text{c}}} = \left[ {\frac{{n!}}{{r!\left( {n - r} \right)!}}} \right]{0.5^n}\end{equation}where n is the number of panellists in the Delphi round, and r is the number of panellists who responded ‘Yes’ to a given indicator. The value 0.5 reflects the probability of chance agreement in a dichotomous response format. Next, the 
${k^*}$ statistic is computed as follows:
\begin{equation}{k^*} = \frac{{{P_{yes}} - {P_c}{\text{ }}}}{{1 - {P_c}}}\end{equation} In Equation (2), 
${P_{{\text{yes}}}}$ denotes the proportion of ‘Yes’ responses. The numerator, therefore, represents the observed consensus beyond chance agreement, while the denominator represents the maximum achievable consensus in the absence of chance agreement. Interpretation of 
${k^*}$ follows the classification proposed by Fleiss (Reference Fleiss1971) and Cicchetti and Sparrow (Reference Cicchetti and Sparrow1981): ‘fair’ (0.40–0.59), ‘good’ (0.60–0.73) and ‘excellent’ (≥0.74). Following the approach of Molnar et al. (Reference Molnar, Man-Son-Hing, Dalziel, Mitchell, Power, Byszewski and John1999), only indicators with an ‘excellent’ level of agreement (
${k^*} \geqslant 0.74$) were retained, while indicators below this threshold were excluded from further analysis.
Results
First round results
In the first round, panellists assessed 82 preliminary sustainability indicators for dairy cattle farming. All indicators demonstrated acceptable response consistency, with IQR values ≤ 1 (Table 2). However, 35 indicators were eliminated for failing to meet one or more predefined thresholds.
Mean, median, IQR and proportion of positive responses (scores 3 and 4) for each indicator
Table 2 Long description
Panel ratings across five sustainability dimensions report each indicator’s mean, median, interquartile range, and the share of respondents giving positive scores (3 or 4). Animal welfare performs strongest overall: many indicators have medians of 3, tight spread, and very high positive response rates, including pain management, analgesic use, and drinking water quality at 100 percent. Social indicators are also generally strong, with full positive agreement for social participation and contribution to employment, and high support for collaboration among farmers. Economic indicators are mixed: several operational items are high, but multiple finance and support-related indicators have low positive response rates and are marked as not meeting the preset threshold, including farmer income stability, access to tax incentives, and fiscal support by the tax authority. Environmental results are moderate, with high support for demand-based fertiliser use and several water and energy items, but lower support and threshold flags for greenhouse gas management and some soil and biodiversity practices. Governance is weakest overall, with many indicators showing low positive response rates and several flagged, including business management digitalisation and independent business assessment. Indicators marked with an asterisk did not meet the predetermined threshold and were excluded, so comparisons should consider that some low-scoring items were removed from further analysis.
Notes: Number of panellists = 19.
The asterisk (*) indicates indicators that did not meet the predetermined threshold and were, therefore, eliminated from the analysis.
Within the economic dimension, eight indicators were eliminated, namely income surplus over feed cost (110), positive milk price margin over production cost (111), monthly business profit (112), farmer income stability (113), feed price risk mitigation strategy (114), focus of dairy farming activities (115), access to tax incentives (116) and fiscal support by tax authority (117). All exhibited proportions of positive responses below 70%. The two fiscal indicators – fiscal support by tax authority and access to tax incentives – were the least supported, receiving only 15.8% and 21.1% positive responses, respectively. Several indicators also had mean and median values below 2.5, reinforcing the perception that these indicators were deemed less relevant by the panellists.
In the environmental dimension, seven indicators were eliminated: diversity of surrounding vegetation (201), crop rotation for soil fertility (202), soil nutrient management (204), livestock diversification (206), management of liquid waste and feed residue (211), safe disposal of diseased livestock (212) and greenhouse gas emission management (216). Among these, greenhouse gas emission management and crop rotation for soil fertility recorded the lowest scores, with positive-response proportions of 31.6% and 36.8%, respectively. Meanwhile, indicators in the social dimension performed better than those in the economic and environmental dimensions. Only four were eliminated: satisfaction with quality of life (301), farmers’ education level (309), labour rights protection (312) and quality of supporting facilities (316).
The governance dimension recorded the lowest overall level of positive responses. Of the 15 indicators assessed, only external collaboration for governance (415) met the predefined criteria. Independent business assessment (409) received no positive responses from any panellist, while business management digitalisation (408) also demonstrated very low positive responses. Conversely, most indicators in the animal welfare dimension performed strongly. Only two indicators – comfort of milking barn space (510) and safety of waiting and isolation area (513) – were eliminated due to positive-response proportions below the 70% threshold.
Insights from the expert FGD provided contextual clarification for several indicators that recorded low levels of positive responses in the first round. In the economic dimension, fiscal-related indicators – namely, fiscal support by tax authority and access to tax incentives – were perceived as irrelevant due to farmers’ limited awareness of taxation systems and practical difficulties in implementation.
In the environmental dimension, greenhouse gas emission management was considered infeasible under current conditions, primarily due to resource constraints at the farm level. The governance dimension attracted particular attention, as the widespread elimination of governance-related indicators reflected the predominantly traditional character of local farm management, which remains strongly oriented towards production and economic survival. Formal governance practices, such as systematic record-keeping, performance evaluation and reporting mechanisms, were therefore not yet perceived as urgent by most panellists.
Following the closed-ended assessment, panellists proposed additional indicators through the open-ended questionnaire. A total of 19 new indicators were identified across the 5 sustainability dimensions: 6 economic, 1 environmental, 3 social, 2 governance and 7 animal welfare indicators (Table 3). Combined with the indicators retained from the first round, 66 indicators were advanced to the second Delphi round.
Newly proposed sustainability indicators
Table 3 Long description
Proposed sustainability indicators are organised by dimension, listing a proposer code, indicator number, indicator name, and a brief description of expected farm practice. Five dimensions are included: economic, environmental, social, governance, and animal welfare. Animal welfare contains the most indicators, covering drug quality control, protecting productive female cattle, individual health history records, feed adequacy, artificial insemination records for genetic control, health screening of purchased breeding stock, and biosecurity via livestock movement monitoring. Economic indicators focus on marketing partnerships and information access, innovation and digital marketing, milk-processing skills, financial health, and including family labour and homegrown forage in profit calculations. Environmental has one indicator on solid and liquid waste treatment facilities, while social includes educational outreach, a safe social environment, and managing social visits to avoid operational disruption. Governance includes animal identification and tagging plus formal contract-based marketing to improve market certainty. Proposer codes repeat across dimensions, so the same code label does not imply the same indicator, and the numbering appears to be an index rather than a ranking or performance score.
A total of 40 inputs were generated through the open-ended questionnaire. Of these, 17 were excluded due to overlap with existing indicators, while 4 were classified as policy recommendations rather than sustainability indicators (Table 4). The policy recommendations included adjusting taxation policies to reflect real conditions, such as disease outbreaks and price volatility, providing targeted subsidies for small-scale dairy sustainability, strengthening market price control for dairy inputs and outputs, and enforcing stricter import regulations and penalties for non-compliant milk processors. These issues were retained for further consideration in the second round, particularly with respect to their perceived urgency.
Policy recommendations proposed by panellists for urgency assessment
Table 4 Long description
Eight policy recommendations are listed with the proposing panellist and an item number. Proposals cover adjusting taxation to match conditions like disease outbreaks and price volatility, and creating training and incubation programmes to bring younger people into dairy farming. Market and supply chain actions include local content rules to boost domestic milk purchasing, direct processor-cooperative partnerships to shorten supply chains, and stronger price control for dairy inputs and outputs. Support measures include targeted subsidies for small-scale dairy sustainability and technical assistance to implement GFP and GMP for better efficiency and competitiveness. Regulatory emphasis appears in calls for stricter import rules and penalties for non-compliant milk processors. Most recommendations are attributed to named panellists, while two items list no proposer, so attribution is incomplete.
a This informant did not participate in the Delphi method.
Several indicators were reassigned across sustainability dimensions to improve conceptual alignment (Table 3). Formal contract-based marketing (417), initially proposed under the economic dimension, was reassigned to the governance dimension to reflect its institutional and relational characteristics (Camanzi et al., Reference Camanzi, Arba, Rota, Zanasi and Malorgio2018). Similarly, a conducive social environment (319) was reassigned from the environmental to the social dimension due to its relevance to safe and pressure-free operational conditions.
Additional qualitative insights were obtained through the FGD and interviews. These inputs did not result in new indicators but contributed further policy-oriented recommendations (Table 4). Key themes included the need for government-supported coaching programmes for the implementation of Good Farming Practices (GFP) and Good Manufacturing Practices (GMP) to improve efficiency and milk price competitiveness. Interviews with senior farmers and cooperative leaders also emphasised the importance of policies aimed at attracting younger generations to the dairy sector, such as training schemes, apprenticeships and business incubation programmes. All recommendations were carried forward to the second round for further evaluation.
Final results
In the second round, a total of 66 sustainability indicators were compiled into the questionnaire: 15 economic indicators, 11 environmental indicators, 16 social indicators, 3 governance indicators and 21 animal-welfare indicators. Nineteen of these indicators were newly introduced based on open-ended inputs from the first round. All additional indicators met the predefined statistical thresholds (see Table S1) and were retained for the final consensus assessment.
Based on the consensus criteria applied in this round, four indicators failed to reach the required level of agreement and were eliminated (Table 5). Two of these were from the economic dimension – digital technology for marketing (120) and skills in dairy product processing (121). The remaining two were the availability and functioning of waste treatment facilities (218) from the environmental dimension and the educational role of dairy farms (318) from the social dimension. Each of these indicators achieved a consensus level of 75%, below the required threshold.
Final consensus results and relevance classification of sustainability indicators
Table 5 Long description
Final consensus ratings for sustainability indicators are grouped into five dimensions: economic, environmental, social, governance, and animal welfare. Each indicator lists the number of panellists agreeing (out of 16), the agreement percentage, two consensus statistics, and a relevance score with a category. Agreement is generally very high, with most retained indicators at about 94 to 100 percent agreement and classified as moderately relevant. Highly relevant indicators are limited and include intergenerational business sustainability and access to marketing partnership and information in the economic dimension, plus several animal welfare items such as protection of productive female cattle, health screening of breeding stock upon purchase, and biosecurity control through livestock movement monitoring. Indicators marked with an asterisk did not meet the preset threshold and were removed; these include digital technology for marketing, skills in dairy product processing, and the educational role of dairy farms. One environmental item, waste treatment facilities, shows 75 percent agreement yet is still labeled moderately relevant, so threshold application may vary by metric and should be interpreted cautiously.
Notes: Number of panellists = 16.
a Newly added indicator based on panellists’ recommendations in the first round.
The asterisk (*) indicates indicators that did not meet the predetermined threshold and were therefore eliminated from the analysis.
As a result, a total of 62 indicators successfully reached consensus, comprising 13 economic, 10 environmental, 15 social, 3 governance and 21 animal welfare indicators (Table 5). Among these, 29 indicators (47%) achieved unanimous agreement (100% consensus). The highest proportions of unanimously endorsed indicators were observed in the animal welfare (45%) and social (31%) dimensions, indicating particularly strong support for indicators related to ethical considerations and socio-cultural aspects of dairy farming. Notably, none of the retained indicators were positioned exactly at the minimum consensus threshold (80%), reflecting a high level of convergence among panellists regarding indicator relevance.
In addition, as shown in Table 5, indicators with a mean relevance score of 3.25 or higher were classified as ‘highly relevant’. These indicators were considered highly beneficial for farm management practices, efficient to implement and strongly supportive of sustainability objectives. The threshold value of 3.25 was derived using an interval-class calculation based on the difference between the maximum (4) and minimum (1) scale values, divided by the number of scale points (4), resulting in an interval width of 0.75. Indicators with mean scores ranging from 2.50 to 3.24 were classified as ‘moderately relevant’.
Within the economic dimension, only two indicators – intergenerational business sustainability (102) and access to marketing partnerships and information (118) – were classified as highly relevant, with mean scores of 3.26 and 3.25, respectively. The remaining 11 economic indicators (85%) fell into the moderately relevant category. Among these, farmers’ personal financial health (122) recorded the highest mean score (3.06), indicating the perceived importance of individual financial stability in sustaining dairy enterprises.
All environmental indicators were categorised as moderately relevant, with none reaching the highly relevant threshold. Demand-based fertiliser use (214) and pesticide use control (215) recorded the highest mean scores (3.05 and 2.95, respectively), highlighting the emphasis on efficiency and input management. In contrast, carbon absorption through green vegetation (217) received the lowest score (2.68), suggesting limited perceived relevance or feasibility within current smallholder farming practices.
Similarly, all social indicators were classified as moderately relevant, with mean scores ranging from 2.79 to 3.21. Contribution to employment opportunities (308) achieved the highest score, followed by a conducive social environment (319) and social–operational balance (320). Conversely, women’s participation in farming (313) recorded the lowest mean score (2.79), indicating that gender-related considerations remain a relatively low priority in the local context.
In the governance dimension, two indicators proposed by panellists – animal identification and tagging (416) and formal contract-based marketing (417) – achieved relatively high mean scores (above 3), although they did not reach the highly relevant category. External collaboration for governance (415) recorded the lowest mean score (2.74), reflecting limited prioritisation of cross-institutional cooperation among smallholder dairy farms.
The animal welfare dimension exhibited the strongest performance. Three indicators – protection of productive female cattle (518), health screening of breeding stock upon purchase (522) and biosecurity control through livestock movement monitoring (523) – were classified as highly relevant, each achieving a mean score of 3.25. All three were newly proposed by panellists, underscoring strong concern for reproduction, herd health and disease prevention. In contrast, cleanliness of the animal rearing area (512) received the lowest mean score (2.79), indicating lower perceived urgency relative to biosecurity-related measures.
Table 6 summarises the analysis of policy recommendations derived from the open-ended questionnaire, FGD and interview with senior farmers in Boyolali. Unlike the sustainability indicators, this analysis focused on assessing the perceived urgency of each recommendation. All proposed policies received positive evaluations, with mean scores exceeding 3.00. However, the dispersion of responses was relatively high, as reflected by IQR values ranging from 1.00 to 1.25. The highest variability was observed for the recommendation on adjusting taxation policies, indicating divergent views among panellists regarding its immediacy and implementation.
Results of the policy recommendation analysis from panellists
Table 6 Long description
Panellists rated eight dairy policy recommendations using mean scores and an interquartile range measure, alongside urgency category and an implementation remark. Seven recommendations are labelled highly urgent and are advised for immediate implementation, with mean scores from 3.38 to 3.69 and an interquartile range of 1. The highest mean score is for enforcing stricter import regulations and penalties for non-compliant milk processors at 3.69. Other high-scoring items include promoting domestic milk purchasing through local content policies at 3.63 and offering technical assistance for good farming and manufacturing practices at 3.56. Providing targeted subsidies for small-scale dairy sustainability scores 3.5, and encouraging direct partnerships between processors and cooperatives scores 3.44. Two items share the lowest mean among the highly urgent group at 3.38: attracting younger generations into dairy farming and strengthening market price control for inputs and outputs. One recommendation, adjusting taxation policies to reflect outbreaks and price volatility, is rated important rather than highly urgent, has a mean of 3, and shows higher response spread than the others, so interpret that mean with caution.
Notes: Number of panellists = 16.
The asterisk (*) indicates that the IQR is greater than 1; thus, the mean score for this policy recommendation should be interpreted with caution.
Using the same classification framework applied to sustainability indicators, recommendations with mean scores between 2.50 and 3.24 were categorised as ‘important’, while those scoring 3.25 or higher were classified as ‘highly urgent’. Based on this classification, seven of the eight policy recommendations were deemed highly urgent, highlighting the need for immediate interventions to address structural and technical constraints in smallholder dairy farming. These priorities include strengthening market guarantees and distribution partnerships, providing targeted subsidies, stabilising input and output prices, enforcing import regulations, offering technical assistance for the implementation of GFP and GMP and promoting farmer regeneration through youth-oriented programmes.
In contrast, the recommendation to adopt more adaptive taxation policies was categorised as important but not urgent. Although taxation issues remain relevant, panellists perceived policy measures with direct and immediate impacts on farm productivity, market access and technical capacity as more pressing priorities for enhancing dairy-farm sustainability.
Discussion
A total of 62 sustainability indicators for dairy farming were successfully formulated through consensus among the panellists, reflecting 5 key dimensions: economics (13 indicators), environment (10), social (15), governance (3) and animal welfare (21). All finalised indicators exhibit a consensus level above 80%, with most having 
${k^*}$ values exceeding 0.74. This demonstrates a very high degree of agreement among the panellists. The multidimensional nature of these indicators aligns with the conceptual framework of Seghezzo (Reference Seghezzo2009), which emphasises that sustainability must be examined holistically as an interconnected system. Furthermore, these findings reinforce the importance of accountability in the governance of dairy farming systems, as highlighted by Duru and Therond (Reference Duru and Therond2015) and Segerkvist et al. (Reference Segerkvist, Hansson, Sonesson and Gunnarsso2020).
Economic dimension: regeneration and financial resilience
In the economic dimension, intergenerational business sustainability and access to marketing partnership and information were identified as the indicators with the highest relevance in promoting the sustainability of dairy farming in Boyolali. This finding aligns with Camanzi et al. (Reference Camanzi, Arba, Rota, Zanasi and Malorgio2018), who emphasise the importance of ensuring that agri-food systems not only remain viable today but can also be successfully transferred to future generations. In the local context, these results reflect growing concerns over the limited regeneration of dairy farmers due to the low interest of younger generations (Wianto and Nurmastiti, Reference Wianto and Nurmastiti2025), driven by negative perceptions of the sector's economic prospects as well as challenges related to work–life balance (Beecher et al., Reference Beecher, Gorman, Kelly and Horan2019). This condition is further exacerbated by the relocation of factories to nearby areas, offering jobs perceived as more ‘clean’, although not always more rewarding financially (Wianto and Nurmastiti, Reference Wianto and Nurmastiti2025).
Meanwhile, the high relevance of the indicator on access to marketing partnership and information underscores that the economic viability of dairy farming is shaped not only by internal business factors but also by external support systems, including cooperatives, commercial partners and milk processing industries. This is consistent with Duru and Therond (Reference Duru and Therond2015), who argue that agricultural sustainability requires systemic approaches supported by strong social and institutional networks.
Furthermore, the emergence of new indicators – such as financial health of farmers (which achieved full consensus at 100%) and the calculation of implicit production costs (87.5% consensus) – reflects a shift towards a more holistic view of sustainability. These findings indicate that the economic dimension of dairy farming is no longer understood merely at the enterprise level but increasingly incorporates the financial resilience of individual farmers. This perspective is consistent with Feil et al. (Reference Feil, Do Amaral, Walter, Bagatini, Schreiber and Maehler2023) and Camanzi et al. (Reference Camanzi, Arba, Rota, Zanasi and Malorgio2018), who highlight that economic sustainability in agriculture is strongly influenced by farmers’ adaptive capacity, including their ability to identify and manage hidden costs that often escape formal accounting yet directly affect business continuity.
The indicator with the lowest relevance – yet still retained – was feed management. According to Sarwar et al. (Reference Sarwar, Khan and Iqbal2002), such management is a critical strategy for addressing feed shortages, whether seasonal or long term, often driven by prolonged dry seasons or competing land-use pressures such as settlement expansion, industrialisation and urbanisation (Meyer and Turner, Reference Meyer and Turner1992; Reid et al., Reference Reid, Kruska, Muthui, Taye, Wotton, Wilson and Mulatu2000; McCusker, Reference McCusker2004; Luoga et al., Reference Luoga, Witkowski and Balkwill2005; Garedew et al., Reference Garedew, Sandewall, Söderberg and Campbell2009; Lambin and Meyfroidt, Reference Lambin and Meyfroidt2011). Based on the FGD conducted by Mekasha et al. (Reference Mekasha, Gerard, Tesfaye, Nigatu and Duncan2014) with farmers in Ethiopia, several strategies are commonly used to address feed deficits, including purchasing feed, providing supplements (such as salt and grains) and converting forage into silage. Although these strategies are reported to help maintain productive herd levels (Benin et al., Reference Benin, Ehui and Pender2003), their comparatively lower relevance score suggests that this technical aspect has not yet emerged as a strategic priority within the broader economic sustainability outlook. This may be attributed to the still-abundant forage availability in Boyolali, limited technical training and supporting infrastructure for smallholders and restricted access to feed-processing technologies at the village level.
Environmental dimension: efficiency and technical capacity challenges
In the environmental dimension, the indicators with the highest relevance scores were the demand-based fertiliser use and the pesticide use control. The strong assessment of these indicators aligns with Alary et al. (Reference Alary, Frija, Abdeladhim, Sghaier, Leauthaud, Farhat and Sghaier2024), who highlight that livestock systems grounded in local resource use tend to promote greater efficiency in managing environmental inputs, including fertilisers and pesticides. Conversely, the indicator on carbon absorption through green vegetation received comparatively lower relevance scores. In Boyolali, shade trees or fodder vegetation have not yet been widely utilised as a greenhouse gas mitigation strategy (Setyono et al., Reference Setyono, Dan and Himawan2018). This reflects limited knowledge and insufficient technical capacity among farmers to optimise ecosystem functions such as carbon sequestration (Torres et al., Reference Torres, Jacovine, Nolasco de Olivera Neto, Fraisse, Soares, de Castro Neto and Lemes2017).
Nonetheless, the high level of consensus (93.75%) among panellists confirms that environmental sustainability in dairy farming extends beyond input efficiency to encompass waste management and pollution control. This finding is consistent with the Sustainability Assessment of Food and Agriculture Systems (SAFA) framework by FAO (2014) and the multidimensional model proposed by Duru et al. (Reference Duru, Therond and Fares2015), both of which emphasise the need for comprehensive environmental assessments – from efficiency considerations to long-term impacts on soil quality, water resources and biodiversity.
Thus, although basic practices such as fertiliser and water management are relatively well adopted, more complex and long-term strategies – such as leveraging vegetation for emission absorption – require targeted interventions in the form of training, extension services and policy incentives. As noted by Mosquera-Losada et al. (Reference Mosquera-Losada, Santiago-Freijanes, Rois-Díaz, Moreno, den Herder M, Aldrey-Vázquez and Rigueiro-Rodríguez2018) and Moreno and Rolo (Reference Moreno and Rolo2019), sustainable farming systems such as agroforestry and silvopastoral systems have gained significant traction in countries with strong climate policy support. In this context, enhancing environmental sustainability in dairy farming in Boyolali must prioritise capacity building and the provision of infrastructure that enables more comprehensive climate-friendly practices.
Social dimension: cohesion, social responsibility and gender inequality
In the social dimension, the findings indicate that the indicator contribution to employment opportunities ranks highest in terms of relevance. This result aligns with Munyaneza et al. (Reference Munyaneza, Kurwijila, Mdoe, Baltenweck and Twine2019), who emphasise that social sustainability in agricultural and livestock systems is shaped not only by social protection mechanisms but also by the capacity of enterprises to generate meaningful employment. In the context of Boyolali – particularly among small- and medium-scale dairy farms – job creation holds substantial importance given the limited ability of the formal sector to absorb rural labour (Fauziah et al., Reference Fauziah, Sukoco, Suwandi and Kurnia2021). Furthermore, Suarsa et al. (Reference Suarsa, Sugiartiningsih, Kusumawati, Fitriani, Pratiwi and Fadilah2024) highlight that employment generation contributes not only to economic outcomes but also strengthens social cohesion. Such cohesion can enhance enterprise resilience through information exchange, mutual support, strengthened bargaining power within the supply chain and sustained generational renewal within the livestock sector.
In addition, the indicators conducive social environment and social–operational balance – both proposed by the panellists – jointly ranked second in relevance. A conducive social environment refers to the absence of informal pressures on livestock enterprises, such as illicit levies imposed by non-state actors or unregistered informal groups. This condition is crucial for maintaining social stability and operational comfort, as such pressures can erode farmers’ sense of security, undermine government legitimacy and ultimately threaten enterprise sustainability (Wrage, Reference Wrage2007). Meanwhile, the social–operational balance reflects a distinct socio-cultural challenge in the local context, where farmers often receive visits from academics, government officials and other stakeholders. In Indonesian – particularly Javanese – culture, hosting guests constitutes an embedded social obligation (Wiryomartono, Reference Wiryomartono2020). However, without proportional management, the frequency of these social interactions may disrupt time-sensitive milking routines. Such disruptions reduce operational efficiency and affect animal welfare, as prolonged pre-milking waiting times restrict natural behaviours such as feeding, drinking, lying and free movement (Dijkstra et al., Reference Dijkstra, Veermäe, Praks, Poikalainen and Arney2012). As noted by Rustinsyah (Reference Rustinsyah2019), imbalances between social obligations and operational efficiency can weaken group cohesion and reduce the effectiveness of livestock enterprises.
Conversely, the indicator with the lowest relevance score – yet still retained – was women’s participation in farming. Notably, this indicator achieved full consensus (100%) among the panellists. This suggests that although women play an essential role in dairy production, particularly in routine hygiene-related tasks involving cows and milk (Wijers, Reference Wijers2019), structural challenges remain in achieving gender equity within the livestock sector. Previous studies consistently show that women’s involvement in agribusiness, including livestock production, is often constrained by patriarchal social norms (Hanum, Reference Hanum2024), domestic workload burdens (Kristjanson et al., Reference Kristjanson, Waters-Bayer, Johnson, Tipilda, Njuki, Baltenweck and MacMillan2014) and limited access to productive resources (Quisumbing and Pandolfelli, Reference Quisumbing and Pandolfelli2010).
Governance dimension: contextual adaptation and the importance of partnerships
The governance dimension in this study reveals a particularly distinctive pattern of findings. Unlike governance frameworks in large corporations – which typically emphasise principles of transparency, accountability and formal risk management (OECD, 2015; Tricker, Reference Tricker2015) – most governance indicators commonly found in the literature were eliminated. This outcome aligns with the principles of adaptive governance (Dietz et al., Reference Dietz, Ostrom and Stern2003), which posit that governance indicators must be tailored to local socio-economic conditions and contextual capacities (Cosens et al., Reference Cosens, Craig, Hirsch, Arnold, Benson, DeCaro and Schlager2017; Andrijevic et al., Reference Andrijevic, Crespo Cuaresma, Muttarak and Schleussner2020). In the case of small-scale and predominantly traditional dairy farming systems in Boyolali, governance practices cannot simply replicate corporate models requiring standard operating procedures (SOPs), formal accounting systems or independent audits. Instead, governance efforts must prioritise a limited set of indicators that are both highly relevant and practically implementable.
Accordingly, out of the 15 initial indicators proposed in the first round, only 1 was retained: external collaboration for governance. This indicator received a relatively high relevance score and achieved full consensus (100%) among all panellists. This finding suggests that partnerships with external actors – such as cooperatives, local government agencies, extension institutions or private-sector partners – constitute the most tangible and strategic mechanism for strengthening governance within smallholder dairy enterprises. This aligns with existing literature emphasising the importance of collective action and multi-actor partnerships in enhancing governance among small agricultural enterprises, especially in developing countries (Bitzer et al., Reference Bitzer, Glasbergen and Arts2013; Kilelu et al., Reference Kilelu, Klerkx, Omore, Baltenweck, Leeuwis and Githinji2017). In the Indonesian context, dairy cooperatives and local agribusiness institutions function as substitutes for internal governance mechanisms that remain underdeveloped at the smallholder level (Jahroh et al., Reference Jahroh, Atmakusuma and Fadillah2020; Fadillah and Indrawan, Reference Fadillah and Indrawan2023).
In interpreting these results, it is important to acknowledge the potential influence of panel composition on the prioritisation of governance indicators. A substantial proportion of the Delphi panel consisted of veterinarians and animal health coordinators, whose professional expertise is primarily oriented towards animal health management, biosecurity and welfare supervision, rather than formal business management or organisational governance (Ruston et al., Reference Ruston, Shortall, Green, Brennan, Wapenaar and Kaler2016). While this composition was highly appropriate for assessing animal welfare-related indicators, it may have contributed to the relatively lower prioritisation of formal governance indicators associated with administrative procedures, financial auditing or independent business assessment. Consequently, the limited retention of governance indicators should be understood not only as a reflection of the traditional management characteristics of smallholder dairy farms in Boyolali but also in relation to the disciplinary perspectives represented within the panel (Powell, Reference Powell2003; Geist, Reference Geist2010).
Meanwhile, two newly emerging indicators within the governance dimension – animal identification and tagging and formal contract-based marketing – recorded higher relevance scores than the initially retained indicator. This indicates that local governance is not shaped primarily by internal managerial structures but rather by the need to integrate smallholder dairy enterprises into broader production and value-chain systems. This finding is consistent with previous studies highlighting the critical importance of linking efficient input systems with reliable output markets for enterprise sustainability (McDermott et al., Reference McDermott, Staal, Freeman, Herrero and de Steeg Ja2010; Rich et al., Reference Rich, Ross, Baker and Negassa2011). Animal identification and tagging are considered crucial as mechanisms for control and traceability, both of which are vital for facilitating market access and enabling higher income potential for producers and other actors within the value chain.
Additionally, formal contract-based marketing reflects an emerging hybrid governance structure that combines formal coordination through contractual arrangements with trust-based relationships to reduce market uncertainty. As demonstrated by Zhang et al. (Reference Zhang, Fu, Lai and Chen2024), synergy between contractual mechanisms and social trust is more effective than an overreliance on formal enforcement of contracts without strong relational foundations.
Animal welfare dimension: practical and ethical foundations of sustainability
Within the animal welfare dimension, the indicators with the highest relevance scores were those proposed directly by the panellists – namely the protection of productive female cattle, health screening of breeding stock upon purchase and biosecurity control through livestock movement monitoring. These findings suggest that reproductive management, the quality of replacement stock and cross-animal disease control are perceived as the core elements for ensuring farm sustainability from an animal welfare perspective. Protecting productive female cattle is critically important, as health disturbances, stress or improper handling can reduce fertility and milk yield, with long-term consequences for reproductive efficiency (Crowe et al., Reference Crowe, Hostens and Opsomer2018; Silva et al., Reference Silva, Jard, Camargo-Júnior, Ébrd, Mrpd, Viana, Silva, Silva and Lourenço-Júnior2023).
Health screening of replacement animals at the point of purchase serves as a preventive measure to ensure that incoming cattle do not carry subclinical diseases or genetic vulnerabilities that may lead to significant reproductive and economic losses (Sanderson et al., Reference Sanderson, Dargatz and Garry2000; Sarrazin et al., Reference Sarrazin, Cay, Laureyns and Dewulf2014). Meanwhile, the monitoring of animal movements as part of a biosecurity system is essential for preventing pathogen transmission between individuals and across farms, which could threaten collective herd health and productivity (Thakur, Reference Thakur2022). These findings reinforce the view in the literature that animal welfare encompasses not only direct treatment of animals but also systemic strategies for reducing reproductive and disease-related risks.
Indicators with the lowest relevance scores – yet still retained – were cleanliness of animal rear area and monitoring of livestock body condition. Although not eliminated, the relatively low ratings indicate that indicators requiring regular visual monitoring or more subtle assessments have not yet become a primary focus in smallholder settings. This is likely due to limited technical capacity and insufficient labour to perform these practices consistently. Similar challenges were reported by Suryanti et al. (Reference Suryanti, Sumardjo, Syahyuti and Tjitropranoto2019) in small-scale poultry farms, where constraints in knowledge and labour hindered the adoption of welfare standards. As emphasised by Ritter et al. (Reference Ritter, Jansen, Roche, Kelton, Adams, Orsel and Barkema2017), the successful implementation of best-practice management depends heavily on farmers’ ability to understand, accept and adapt technical recommendations to their specific circumstances.
In terms of panel consensus, the animal welfare dimension emerged as the most dominant compared with the other four dimensions (economic, environmental, social and governance), with 13 indicators receiving full consensus (100%). This finding highlights that animal welfare is regarded as both a fundamental and practical pillar of livestock sustainability in Boyolali. Scientifically, several factors justify this centrality. First, animal welfare is directly linked to productivity and profitability, as stress and poor health are proven to reduce reproductive performance and milk production (Spigarelli et al., Reference Spigarelli, Berton, Corazzin, Gallo, Pinterits, Ramanzin, Ressi, Sturaro, Zuliani and Bovolenta2021; Fraser, Reference Fraser2023). Second, many welfare indicators involve simple technical practices that do not require advanced technologies, making them accessible to smallholder farmers (Tuyttens et al., Reference Tuyttens, De Graaf, Andreasen, de Boyer Des Roches, Van Eerdenburg, Haskell, Kirchner, Mounier, Kjosevski, Bijttebier and Ampe2021). Third, indicators such as udder hygiene, drinking water quality and cleanliness of animal rear area are highly visible and embedded in farmers’ daily routines, making them easier to understand and internalise (EFSA Panel on Animal Health and Welfare (AHAW), 2012). Moreover, ethical considerations and cultural values strengthen the social legitimacy of animal welfare practices (Doerfler and Peters, Reference Doerfler and Peters2006), particularly in agrarian communities such as rural Indonesia (Zali, Reference Zali2018).
Strategic recommendations from panellists to enhance sustainability
Beyond identifying sustainability indicators, this study also elicited policy recommendations through open-ended questionnaires, FGDs and in-depth interviews. The analysis indicates that the dairy farming sector faces multidimensional challenges at both local and national levels (Akzar et al., Reference Akzar, Umberger and Peralta2023). The main structural barriers identified include low youth participation, high dependence on intermediaries within the supply chain and the absence of market guarantees and price stability. These findings align with Delgado et al. (Reference Delgado, Narrod, Tiongco and de Camargo Barros2008), who highlight that such constraints significantly hinder improvements in productivity and competitiveness in livestock sectors across developing countries.
To address these challenges, the panellists emphasised the need for more targeted and context-sensitive policy interventions. Consistent with recommendations by Latruffe et al. (Reference Latruffe, Bravo‐Ureta, Carpentier, Desjeux and Moreira2017), Yaseen et al. (Reference Yaseen, Somogyi and Bryceson2018) and Merlino et al. (Reference Merlino, Renna, Nery, Muresu, Ricci, Maggiolino, Celano, De Ruggieri and Tarantola2022), technical training and entrepreneurship incubation programmes for young farmers are considered essential for attracting new generations into the sector. Additionally, the implementation of local content policies to ensure market access for domestic dairy products, the provision of well-targeted subsidies and stronger regulations on dairy imports were deemed strategic for creating a more stable and inclusive business environment.
Direct partnerships between dairy processing industries and dairy cooperatives also emerged as a highly urgent recommendation. Such vertical coordination models – as noted by Zhong et al. (Reference Zhong, Zhang, Jia and Bijman2018) – have the potential to shorten distribution chains, reduce production costs and strengthen the bargaining power of smallholder farmers in the marketplace. Moreover, capacity-building programmes for implementing GFP and GMP were identified as critically needed. As demonstrated by Bihn and Reiners (Reference Bihn, Reiners, Siddiq and Uebersax2018), these programmes can substantially enhance product value addition and improve overall farmer welfare.
Conclusion
This study formulated consensus-based sustainability indicators to support sustainable dairy farming in Boyolali Regency using a two-round Delphi-based approach informed by theory-driven, expert-driven and stakeholder-driven inputs. The resulting indicator set captures locally relevant economic, environmental, social, governance and animal welfare dimensions, demonstrating the value of participatory approaches in aligning scientific knowledge with stakeholder perspectives in smallholder dairy systems.
The findings indicate that sustainability in smallholder dairy farming extends beyond production efficiency to include intergenerational continuity, access to market partnerships, social conditions that enable fair business practices and strong attention to animal welfare. The high level of consensus achieved across multiple dimensions suggests that the identified indicators are both contextually appropriate and broadly acceptable to key actors within the local dairy sector.
From a policy perspective, the results underscore the importance of integrated interventions combining capacity building, improved market linkages and targeted institutional support. In particular, programmes that strengthen technical competencies, facilitate collaboration across the value chain and enhance farmers’ adaptive capacity are critical for improving long-term sustainability and resilience.
Despite these contributions, this study is subject to contextual limitations. The focus on Boyolali Regency as a representation of rural smallholder dairy systems limits the generalisability of the findings, especially when compared with urban dairy systems characterised by different scales of operation, cost structures, market access and levels of technological adoption. In addition, the disciplinary composition of the Delphi panel – dominated by veterinarians and animal health professionals – may have influenced the relative prominence of animal welfare indicators in the consensus outcomes, although this expertise was appropriate for assessing welfare-related aspects. Future research should therefore adopt geographically comparative designs and more diverse panel compositions to ensure that sustainability strategies are adaptable across diverse socio-economic and spatial contexts.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0022029926102301.
Acknowledgements
The authors would like to extend their sincere appreciation to all experts and stakeholders who participated in the Delphi process, interviews and focus group discussions. We are especially grateful to the officials of the Disnakkan of Boyolali Regency, the veterinary professionals from the Puskeswan, the academic representative, the dairy farmers and the board members of the KUD for their valuable contributions throughout the study. Their insights and thoughtful reflections were essential in developing a comprehensive and context-sensitive sustainability indicator system for dairy farming in Boyolali. The authors also express their gratitude to the panellists of Forum Pusaka Jateng 2025, organised by the Representative Office of Bank Indonesia for Central Java Province, whose constructive feedback helped refine the conceptual framing of this paper.
Funding statement
This research received no external funding.
Competing interests
The authors declare that they have no conflicts of interest related to the research, authorship or publication of this article.
Data statement
The data that support the findings of this study are available on reasonable request from the corresponding author.
Declaration of generative AI use
During the preparation of this manuscript, the authors used ChatGPT solely for language editing and polishing. All content generated by the tool was reviewed, revised and verified by the authors. The authors take full responsibility for the accuracy and integrity of the manuscript’s content.
Appendix
Sustainability indicators for dairy farming
Table A1 Long description
Sustainability indicators for dairy farming are organised into five dimensions: economic, environmental, social, governance, and animal welfare. The economic section lists 17 indicators, ranging from climate resilience and intergenerational continuity to productivity, feed management, profitability, and reduced reliance on subsidies or debt. The environmental section lists 17 indicators focused on vegetation diversity, soil and nutrient practices, water and energy efficiency, waste handling, habitat conservation, fertiliser and pesticide control, and greenhouse-gas management. The social section lists 17 indicators covering quality of life, recognition, working hours, diversification, community participation, education and training, labour rights, inclusion, women’s participation, and occupational health and safety. The governance section lists 15 indicators addressing record keeping, information security, transparency, participation and role clarity, performance evaluation, procedures, digital tools, compliance, risk management, legal documents, planning, and external collaboration. The animal-welfare section lists 16 indicators spanning body condition, disease prevention, udder care and hygiene, safe handling, locomotion and pain management, water quality, housing and flooring safety, shade and shelter, social behaviour, and calm handling. Indicator numbers are grouped by dimension using the first digit, and sources are provided per indicator; the table defines what to measure but does not report scores or performance levels.
Open access
