Abstract
Digital tools are spreading quickly in indoor dairy systems, but they continue to struggle in pasture-based farming. Most research explains this gap by focusing on farmers and their adoption decisions. This article takes a different approach. It shows that many failures originate upstream, in how dairy-tech startups imagine, design and commercialise their products. Drawing on 51 interviews with founders, engineers, product designers, dairy experts/advisors, board members/investors and farmers, the study examines a range of digital tools, including animal health and behaviour monitoring sensors, pasture measurement systems, herd management applications and decision-support applications. It identifies three types of systemic misfit that shape performance in grazing environments: Ecological-Design Misfits, Temporal-Commercial Misfits and Institutional-Prioritisation Misfits. These misfits arise from indoor-dairy design assumptions, year-round business models and investor expectations that favour fast scaling over local adaptation. The article contributes to socio-technical studies of agricultural innovation by showing how design logics and commercial models can misalign technologies with ecological and organisational realities. It concludes with practical recommendations for policymakers, funders, advisors and cooperatives to support ‘pasture-first’ innovation and improve the fit between digital tools and grazing-based dairy systems.
Introduction
Pasture-based dairy systems play a central role in global milk production and rural livelihoods (Delaby et al., 2020; Hanrahan et al., 2018; Rapiya et al., 2025). They support grassland biodiversity and soil carbon sequestration (Rapiya et al., 2025), underpin rural employment in many regions (Wright Morton and Miller, 2007), and they dominate the dairy sectors of countries such as Ireland, New Zealand and parts of Latin America and continental Europe (Fariña et al., 2020; Markiewicz-Keszycka et al., 2023; Pinxterhuis et al., 2015).
Digital tools – sensors, apps and decision-support systems – continue to perform far better in indoor, high-input dairies than on grazing farms (Palma-Molina et al., 2023a, 2023b; Shalloo et al., 2018).
Most research has tried to explain this gap by focusing on farmers: their age, digital skills, herd size, attitudes towards technology, levels of education or willingness to adopt new tools (Edwards et al., 2015; McDonald et al., 2016; Palma-Molina et al., 2023a; Sarma et al., 2025; Yang et al., 2021). While valuable, this approach assumes that the technologies available to farmers are already well-suited to their context. It treats the tool as fixed and places the burden of success on the user. It is also important to note that technology adoption depends on cost and return on investment (French et al., 2015). Farmers are unlikely to adopt tools whose costs exceed their practical benefits. This study does not set that consideration aside, but argues that without fit between a tool and its environment, cost-benefit calculations cannot be meaningfully made.
Linked to this, Shalloo et al. (2018) argue that digital tools are often ‘solution-focused’ rather than ‘problem-focused’, suggesting that development is frequently driven by technological capability rather than the specific management challenges faced by pasture farmers. This misalignment between ‘tool logic’ and ‘farm logic’ is a central theme in this analysis.
This article argues that many digital tools struggle in pasture-based systems not because farmers are reluctant or ‘hard to reach’, but because the technologies themselves are shaped by assumptions that do not match outdoor, mobile, seasonal grazing environments. In other words, the problem lies upstream: in how startups imagine, design and commercialise their products.
To investigate this, the study asks three research questions (RQ):
RQ1: How do dairy-tech startups conceptualise pasture-based farming when designing digital tools? Do they treat it as a core market, or as a variation of indoor systems? RQ2: Which design, engineering and business-model decisions lead to misalignment between digital tools and the realities of pasture-based systems? How do assumptions about data, connectivity, seasonality and user behaviour shape product performance? RQ3: What system-level pressures – such as investor expectations or scaling strategies – reinforce these mismatches? Why do even well-intentioned companies struggle to build technologies that truly fit grazing environments?
Drawing on 51 interviews with founders, engineers, product designers, dairy experts/advisors, board members/investors and farmers, this study shows that digitalisation challenges in grazing-based dairying are not only demand-side issues. They are shaped by how innovation is imagined, funded, engineered and brought to market. By shifting attention from farmer behaviour to startup-side design logics, this article offers a new explanation for why digital tools often fail in pasture systems – and what it would take to build technologies that truly work in these environments.
Conceptual background
Digital technologies have become increasingly visible across the dairy sector, with tools promising to optimise heat detection, health monitoring, grazing management, pasture measurement and whole-farm decision-support (Basunathe et al., 2010; Bewley et al., 2015; Liu et al., 2023; Shalloo et al., 2018). A substantial body of work has validated the technical performance of these tools in controlled settings, examining sensor accuracy, detection algorithms and connectivity performance in precision livestock farming (Aquilani et al., 2022; Bretas et al., 2024). Furthermore, a significant amount of literature around these tools has focused on adoption: whether farmers have the necessary skills, whether they trust the technology, whether advisory networks support digitalisation, and whether the tools are perceived as useful (Edwards et al., 2015; Giua et al., 2022; Gabriel et al., 2023; Martínez-García et al., 2015; McDonald et al., 2016; Palma-Molina et al., 2023a; Rizzo et al., 2024; Yang et al., 2021). These studies have contributed very valuable insights, showing how factors such as labour availability, generational change, farm size, education levels and user confidence shape the likelihood that technologies will be integrated into daily work (Edwards et al., 2015; Martínez-García et al., 2015; McDonald et al., 2016; Palma-Molina et al., 2023a; Yang et al., 2021). Some of this work also shows that low adoption can reflect poor fit between a technology and farmer needs, rather than farmer resistance alone (Eastwood et al., 2012; Kaushik et al., 2024; Shalloo et al., 2018). Nonetheless, the dominant emphasis in adoption research tends to focus on farmer-side variables. This perspective often views digital tools as largely neutral artefacts and assigns most of the responsibility for their uptake to the farmer.
A growing body of socio-technical research challenges this assumption, arguing that technologies function effectively only when they are shaped by the ecological and organisational contexts in which they are used (Eastwood et al., 2019; Higgins et al., 2017; Klerkx et al., 2010). This literature shows that many agricultural tools incorporate design assumptions about what ‘normal’ farming looks like, assumptions that often come from high-input or indoor models of livestock production, where routines are stable, work is regular, and data can flow continuously (Bretas et al., 2024; Burton et al., 2020; Driessen & Heutinck, 2015; Pronti et al., 2025). When such templates are applied to pasture-based dairying, ecological variability and outdoor conditions frequently disrupt the underlying logic (Shalloo et al., 2018; Wilkinson et al., 2020). Mud, uneven terrain, roaming animals, hedgerows, humidity, and unreliable connectivity make continuous data capture difficult, and the seasonal nature of grazing means that farmers engage differently with technologies across the year (Bailey et al., 2018; Deming at al., 2018; Hogan et al., 2022; Nadimi et al., 2008). Research on context-dependent innovation has repeatedly shown that when tools do not fit these rhythms and realities, farmers either adapt them in unintended ways or abandon them entirely (Basunathe et al., 2010; Eastwood et al., 2012; Higgins et al., 2017; Kaushik et al., 2024). These patterns suggest that low adoption cannot be explained by farmer behaviour alone.
There is also increasing recognition that the commercial and institutional environments shaping digital agriculture matter as much as the technical ones (Klerkx & Rose, 2020). Many Agtech startups depend on investment models that favour scalability, subscription revenue, and rapid expansion across markets (Fernandez-Vidal & Alarcon, 2025; Rotz et al., 2019a). These expectations encourage the development of standardised products designed to travel easily between regions, often at the cost of local adaptation (Eastwood et al., 2012; Fernandez-Vidal & Alarcon, 2025; Hamdan-Livramento et al., 2024). Market priorities also play a role (Fernandez-Vidal & Alarcon, 2025; Lajoie-O'Malley et al., 2020; Shepherd et al., 2020). Pasture-based regions are sometimes classified as less attractive because of their smaller herd sizes or lower perceived revenue per farm (French et al., 2015). Such classifications frequently occur early in a company's life, shaping where prototypes are tested, how much engineering attention is dedicated to ecological challenges, and which users are treated as central to product development. Studies of agricultural innovation systems show that once these priorities are set, they become self-reinforcing: shallow testing produces weaker performance, weaker performance confirms perceptions of low-value markets, and investment is further redirected elsewhere (Eastwood et al., 2012).
Together, these strands of literature point towards a broader conceptual gap. While many studies examine farmer adoption (e.g. Palma-Molina et al., 2023a; Yang et al., 2021), far fewer examine the upstream design and organisational choices that shape how technologies reach the market. The focus tends to be on how farmers respond to tools (e.g. Edwards et al. 2015; El-Osta et al., 2000; Kaushik et al., 2024), not on how those tools were imagined and built in the first place. This leaves unexplored the possibility that digitalisation challenges in pasture-based dairy systems result from misfits generated within the innovation process itself. Insights from qualitative research on agricultural technology design support this view, showing that breakdowns often emerge long before a device arrives on a farm, within design imaginaries, engineering assumptions, commercial models, and investor expectations (Bronson, 2019; Carolan, 2018; Fernandez-Vidal & Alarcon, 2025).
Existing socio-technical studies have highlighted the importance of context, advisory networks, and organisational routines in shaping the development and performance of agricultural technologies (Eastwood et al., 2019; Gaitán-Cremaschi & Klerkx, 2026; Higgins et al., 2017; Klerkx et al., 2010). However, this work tends to examine misalignment after technologies reach farms. Very few studies analyse the upstream stages of innovation, where assumptions are formed, architectures are fixed, and commercial models are negotiated. As a result, we know much more about how farmers react to digital tools than about how those tools gain the design features that later affect their use.
This study offers a new perspective from the startup side, showing how misfits can emerge even before a product is tested on farms. Three types of misfit are used analytically throughout this article. An Ecological-Design Misfit refers to the gap between a tool's design assumptions and the physical, environmental, and spatial realities of outdoor grazing systems, including hardware durability, connectivity, and movement patterns of animals. A Temporal-Commercial Misfit refers to the conflict between business models and performance metrics that assume continuous year-round engagement and the seasonal rhythms of pasture-based farming. An Institutional-Prioritisation Misfit refers to the way market narratives, investor expectations, and organisational cultures systematically devalue or marginalise pasture-based contexts during product development.
Material and methods
This study uses a qualitative research design to understand how dairy-tech startups imagine, design and commercialise digital tools for pasture-based systems. Because the aim is to uncover the assumptions, pressures, and decision-making processes that shape technological fit, the research relies on in-depth, open-ended data rather than large surveys (Klein & Myers, 1999).
The core dataset consists of 51 semi-structured interviews conducted between 2022 and 2025 with founders, engineers, product designers, dairy experts/advisors, board members/investors and farmers (see Table 1 for details of participants). These participants came from 26 distinct startup companies. Some companies contributed multiple interviewees across roles, for example, a chief executive officer (CEO) and a head of product, which was intentional to capture different internal perspectives on the same development decisions. Participants were recruited through purposive and snowball sampling to ensure a diverse mix of company sizes, technologies, and regional contexts (Parket et al., 2019; Valerio et al., 2016).
Characteristics of interview participants.
Interview questions differed by participant group. Technology executives, founders, and product designers were asked about how their companies defined target users, how design decisions were made, how ecological and seasonal conditions were considered, and how investor expectations influenced product development. Board members and investors were interviewed to verify these claims and gain complementary insights. Farmers were interviewed separately, with questions focused on how they used digital tools in practice, what adjustments or workarounds they developed, and how tools performed across different seasons and conditions. Farmer accounts were included to cross-check startup characterisations of farmer behaviour and to understand on-farm perceptions. Dairy experts and advisors were asked about their experience working with both farmers and technology companies, and about the barriers and gaps they observed in the field.
The geographic spread of participants was driven by the innovation networks of the startup companies studied, not by a design to achieve national representativeness. Many of the companies operated across multiple regions and tested their products in different markets. Farmers from Ireland, New Zealand, Spain, and Australia were selected because their countries represent major pasture-based dairy systems and because several startups in the sample were actively targeting these markets. Farmers from Canada, and the USA were included as purposive cases: they were early users of the specific technologies under study or had been identified through startup networks as relevant informants. Participants classified as dairy experts or consultants (e.g. extension agents) included both independent farm advisors with experience in pasture-based systems and technical consultants engaged by startup companies.
The interview data were analysed using an abductive coding approach (Vila-Henninger et al., 2024). Following the Gioia methodology's emphasis on researcher – informant centricity, analysis began with open, line-by-line coding of interview transcripts to preserve participants’ vocabulary and allow unexpected patterns to emerge (Gioia et al., 2013). As coding progressed, emerging categories were iteratively compared across companies, regions, and technology types, gradually moving towards second-order theoretical themes through focused coding and constant comparison (Charmaz, 2006).
Throughout the process, theoretical insights from socio-technical systems research (e.g. Geels, 2004, 2005, 2018; Klerkx et al., 2010, 2012) served as ‘sensitising devices’ rather than predetermined codes, helping refine interpretations without constraining inductive discovery (Charmaz, 2006). Coding cycles continued until theoretical saturation was reached, when additional interviews no longer produced substantively new concepts but instead deepened existing patterns (Corbin & Strauss, 2014). Illustrative data excerpts supporting the development of second-order concepts, particularly for the ecological-design misfit, are provided in Appendix A, offering transparency regarding the analytical steps and grounding of the findings.
To support validity, several steps were taken to address potential biases, in line with research best practices (Gioia et al., 2013). Because companies willing to participate may be more reflective or open to critique, the sampling strategy intentionally included firms at different stages (e.g. successful, struggling and defunct) identified through industry networks and public records. Interviewees were asked to describe specific events, failures, and design choices rather than general opinions, which helped reduce retrospective glossing. Where possible, internal documents provided by participants, such as product roadmaps, user-testing reports, investor presentations, and technical notes, were used to triangulate interview claims. These materials offered insight into how companies framed their decisions for different audiences and how assumptions travelled through engineering, commercial, and investor processes.
The overall methodological goal was to trace how design imaginaries and organisational pressures shape the technologies offered to pasture-based farmers. A qualitative approach is particularly well suited to this task because it captures the reasoning, uncertainty, and negotiation involved in building digital tools (Corbin & Strauss, 2014). Rather than measuring adoption outcomes alone, the method reveals how misfits arise inside the innovation process itself, long before technologies reach farms. This provides a foundation for understanding why digital tools often struggle in grazing environments and helps identify leverage points for improving technological fit.
Results
Our findings show that digital monitoring and decision-support tools (i.e. animal health and behaviour monitoring sensors, pasture measurement systems, herd management applications and decision-support applications) struggle in pasture-based dairy due to three systemic misfits: places where the logic of startup innovation clashes with the ecological, temporal, and institutional realities of grazing systems.
What follows in this section is an overview of each misfit.
Category I: Ecological-design misfits
Many design problems in pasture-based dairy arise from how startups form early assumptions about the production environment.
Seventy percent of the companies represented in the sample reported that their teams began building their products using models taken from indoor or high-input dairy systems. Early wireframes assumed stable routines, fixed cow locations, and regular access to computers or Wi-Fi. Algorithms were often trained on indoor datasets because they were easier to access or already available within partner companies.
Several engineers explained that they relied on ‘standard dairy workflows’ without realising these were grounded in indoor settings. As one founder recalled, ‘Our first heat-detection model was trained on data from a partner in California, totally indoor, totally different movement patterns. We didn’t realize how much that shaped the behavior of the algorithm until we tried it on cows that spend 18 hours walking grass’.
Around sixty percent of the startups in our sample developed their minimum viable product – the earliest working version of their product – before engaging in field immersion. Key decisions – such as sensor type, casing materials, connectivity options, and data sampling rates – were based on internal discussions or imagined use cases rather than direct observations of pasture conditions. As one product designer described, ‘We had a working MVP before anyone on the team had actually stood in a muddy paddock. When we finally visited a farm, half of what we built made zero sense in that environment’. This was echoed by a farmer ‘the first version of their herd management program was useless, [COMPANY NAME] had no idea how pasture farming worked’.
Once teams visited grazing farms, they found that several design choices did not match outdoor practices. For example, the placement of sensors assumed short distances between animals and gateways, and interface flows assumed daily app use. Because these architectural decisions were already fixed, in many instances teams had limited flexibility to adjust their products when they later encountered real farm conditions.
Once devices were deployed outdoors, ecological realities exposed weaknesses not anticipated during development. Mud, moisture, hedges, and micro-topography interrupted signal transmission. Hardware housings absorbed water or cracked, and roaming animals damaged sensors or dislodged gateways. Connectivity dropped when terrain blocked line-of-sight.
These issues produced irregular data, inconsistent alerts, or device failure. The study participants that encountered these issues initially interpreted these problems as isolated technical bugs but later recognised that they were routine consequences of working in outdoor, mobile, and weather-exposed environments.
This led to complete redesigns of certain key features. One founder admitted ‘the cows were moving in patterns we had not expected, and the system was picking up too many signals. We had to redesign it so it could pick what actually mattered’. Another one added ‘we had assumed data uploading would be straightforward. In practice, we had to rebuild parts of the communication system to deal with intermittent and delayed uploads’.
Category II: Temporal-commercial misfits
Our research shows that commercial models and internal performance metrics often conflict with the seasonal and irregular rhythms of pasture-based dairy systems.
All companies relied on subscription models that assumed steady, year-round usage. However, farmers’ activity patterns followed the production cycle, with high engagement during calving and mating and much lower engagement in winter or dry periods. As a board member mentioned, ‘we would be puzzled by some of the early data, we believed that farmers were no longer interested in our product or were not engaging with it. The team had just missed that this was normal in pasture farms!’.
Four of the companies in our sample responded by adding features designed to ‘keep users active’, even though farmers had little need for digital support during these periods. In the words of an executive, ‘we ended up changing our pricing because we had assumed farmers would use the system all year. That was not how they used it’.
Companies monitored ‘active use’ through dashboard analytics. These dashboards often flagged seasonal drops in logins or data entries as potential customer loss. A sales manager mentioned ‘would get alerts saying usage was dropping, but often it was just a seasonal dip’.
These signals triggered redesign discussions, internal review meetings, or outreach to farmers asking them to increase activity. As a head of business development said ‘a farmer might not use the product for a few weeks, and it would be reported as a problem. But often there was nothing wrong’ and he added ‘sometimes we sent someone out after an alert, and then realised there was no real problem. It was just how the farm worked at that point in the season’.
Participants from 14 of the firms in our sample said their systems assumed regular, uninterrupted data flows. Yet interviewees explained that data from pasture systems often arrived in irregular bursts because of terrain, weather, and limited connectivity. Of the 14 interviewees who discussed connectivity, 8 said their systems often treated missing data as faults. One engineer explained, ‘the system would flag an error because nothing had come through for a few hours. Often the device was just out of range, or uploading later in a batch’. Another added, ‘our original system design assumed the kind of stable connectivity you get indoors, and that just was not realistic in pasture settings’.
Interviewees also described uneven patterns of farmer engagement. Of the 25 participants who discussed usage, 18 said farmers used tools mainly during key events such as calving, breeding, or feed shortages, rather than every day. One founder said, ‘farmers would use it a lot when they needed it, then barely touch it for weeks because there was nothing happening that required it’.
These patterns often conflicted with commercial metrics based on regular activity. Of the 31 interviewees who discussed retention measures, 17 said lower activity often triggered alerts or concern internally. One founder added, ‘we would see the alert, send someone out to get the farmer using it again, and then realise there was never really a problem in the first place…it was just part of their normal seasonal routine’. In the words of a farmer, customer of this company, ‘there were constant alerts of [COMPANY APP]. I liked the system but I did not need to use it at that time’.
These findings show that seasonal declines in use, delayed data uploads, and episodic engagement were often interpreted internally as signs of weak customer interest or technical failure. In practice, the farmers we interviewed described these patterns as normal features of pasture-based dairy systems rather than indicators of commercial risk.
Category III: Institutional-Prioritisation misfits
The interviews with Dairy Technology executives and board members/investors show that institutional priorities and commercial narratives strongly influence how pasture-based markets are valued within companies. These influences shape where resources flow, how features are prioritised, and how farmer feedback is interpreted.
Eighty percent of the startups in our sample classified pasture-based regions as ‘secondary’ or ‘niche’ markets, often early in the product's development. These classifications were usually based on assumptions about smaller herd sizes, lower revenue per farm, or perceived weaker commercial potential. One founder explained: ‘we always saw grazing farms as a smaller market segment, so most of the early product decisions were built around housed systems first’.
In circa 60% of the startups in our sample, founders and product designers said that pilot farms were selected because they were convenient or prestigious rather than representative of pasture-based systems. As a result, early user testing did not fully expose the challenges of outdoor environments. Interviewees noted that resource allocation – such as engineering time, feature development, and customer support – tended to favour indoor-oriented markets, even in countries where pasture dairying predominated. One product designer said: ‘the first farms we worked with were large indoor farms because they were easier to reach, and it was easier to show results there’.
All startups in our sample faced pressure from investors or boards to expand quickly across multiple regions. These expectations encouraged teams to design products that could be used with minimal adaptation. One investor explained: ‘from a venture capital perspective, the expectation was to scale across markets quickly. Spending too much time adapting to one grazing region was seen as slowing things down’.
This pushed companies to move into new markets before resolving problems identified in earlier pilots. Founders and product designers explained that these rapid timelines limited opportunities to develop a deeper understanding of pasture conditions or to refine the product through iterative, locally grounded testing. A CEO said: ‘we had a lot of pressure from our investors and were already trying to launch in a second country before we had solved the basic problems farmers were having in the first one’.
Dairy advisors who had worked with multiple startup companies reported that their practical knowledge of grazing systems was not always integrated into product decisions. Three product engineers admitted that local dairy expertise was consulted only after major features had already been built. One dairy advisor corroborated this: ‘by the time they asked us about grazing conditions, most of the important design decisions had already been made’. These interviewees described moments when feedback about ecological or seasonal constraints was overridden by internal narratives that framed indoor systems as the ‘standard model’ for dairy technology.
In addition, pasture-based know-how was sometimes viewed as too region-specific or insufficiently scalable to drive roadmap decisions. A farmer mentioned when discussing a herd management application: ‘we asked for features that would make the system work better for grazing farms, but we were told those requests were too regional. The company wanted standard global features and did not want to keep making local adaptations’.
Founders, product designers, and dairy advisors also reported that farmers often adapted tools to match their workflow. These participants said that they initially interpreted these adaptations as incorrect use rather than helpful feedback. One founder said: ‘we expected farmers to use the full dashboard every day, but most only used one or two functions that were useful during the grazing season’.
Furthermore, these deviations from the expected workflow were framed as training issues. One general manager explained: ‘when farmers did not use the product the way we expected, we often treated it as a training issue rather than a product problem’.
Across these four concepts, the findings show how organisational priorities, scaling pressures, and internal narratives shape product development in ways that reduce attention to the needs of pasture-based farmers. These patterns influence which markets receive investment, how feedback is interpreted, and how tools evolve over time.
Synthesis of results
The three forms of misfit identified in this study are closely connected. Early assumptions about indoor dairy shape product design choices. These design choices then influence how companies interpret seasonal use patterns and allocate resources. Over time, these processes reinforce one another and make it more difficult for startups to adapt their tools to pasture-based systems.
Figure 1, developed inductively from the interview findings and coding process, summarises this process. It shows how indoor assumptions influence design, how pasture conditions expose weaknesses, and how commercial and organisational priorities often limit later adaptation
‘The misfit cycle’ – a reinforcing feedback loop of systemic failure.
Table 2 provides an overview of how startup assumptions, pasture conditions, and resulting breakdowns connect across the three misfit types
How systemic misfits arise in startups designing for pasture-based dairy.
Lastly, Table 3 summarises the empirical responses to each research question and links them directly to the study's findings.
Summary of empirical responses to the research questions.
Discussion
This study examined why digital tools tend to underperform in pasture-based dairy systems by shifting analytical attention from farmer behaviour to the upstream design and commercial processes shaping technological form. The findings reveal that many of the difficulties observed on farms – including sensor failures, connectivity problems, inconsistent alerts, and seasonal drops in engagement – originate long before deployment, embedded within engineering assumptions, business models, and institutional priorities inside dairy-tech startups. This section situates those findings in relation to existing scholarship and clarifies the study's theoretical contribution.
A large amount of the existing literature on digital agriculture explains uneven uptake through farmer characteristics such as age, skills, education, herd size, risk attitudes and advisory support (Edwards et al., 2015; McDonald et al., 2016; Palma-Molina et al., 2023a; Yang et al., 2021). This body of work has produced valuable insights into predictors of technology use, but it tends to assume that technologies arrive on farms as technically stable artefacts. When tools fail or remain underused, explanation is therefore typically sought in behavioural resistance, skill deficits, or cultural hesitation.
The findings of this study challenge that framing. They show that digital tools reach pasture-based farms already misaligned with their ecological and organisational context. Design assumptions derived from indoor production systems are embedded into software architecture, hardware specifications, and usage expectations long before farmers encounter the tool. As a result, low uptake and inconsistent performance appear not as problems of farmer reluctance but as symptoms of inadequate technological fit.
This perspective reframes digitalisation failure as a problem of innovation design rather than user behaviour. Instead of asking why farmers do not adopt tools, this study asks why tools are built in ways that make failure more likely in grazing environments. It therefore aligns with critical work in digital agriculture arguing that technological systems embed particular assumptions about what ‘normal’ farming looks like, whose needs matter most, and which production systems are prioritised during design and investment decisions (Bronson, 2022; Carolan, 2018; Miles, 2019). Building on these insights, this article conceptualises digitalisation failure as a problem of innovation fit: the degree to which a technology aligns with the ecological, temporal, and institutional environment in which it is expected to operate.
Furthermore, socio-technical research has long emphasised that technologies only perform as intended when aligned with their social, organisational, and ecological context (Barret et al., 2022; Eastwood et al., 2019; Higgins et al., 2017; Klerkx et al., 2010). Previous studies have shown how misalignment leads to adaptation work, misuse, or abandonment once tools are deployed in farming systems (Aquilani et al., 2022; Tzanidakis et al., 2023). However, most analyses examine misfit at the point of use, after the tools have been designed and produced.
This study contributes to that literature by shifting attention upstream, to the phase where assumptions are formed and technological pathways become locked in. By analysing how startups imagine users, design systems and define markets, it shows that many failures are not simply due to external conditions, but result from early oversimplification of real-world complexity.
The ecological – design misfit highlights this dynamic clearly. While prior research recognises technical challenges such as sensor degradation, connectivity breakdown, and environmental interference in outdoor systems (Bailey et al., 2018; Bretas et al., 2024; Nadimi et al., 2008), the findings show that these problems frequently reflect early design assumptions rather than unpredictable field conditions. Failure in pasture-based environments often arises because products were shaped by engineering templates developed for indoor systems.
The temporal – commercial misfit further extends socio-technical insights of Klerkx et al. (2010) and Eastwood et al. (2019), by revealing how business models themselves function as technical design constraints. Subscription pricing, user-engagement metrics, and performance dashboards embed assumptions about continuous use that conflict with the seasonal rhythms of grazing production. As a result, tools are designed to favour constant interaction rather than episodic, event-driven use. Seasonality is then misread as disengagement, and agricultural rhythms become interpreted as commercial risk.
Finally, the institutional-prioritisation misfit illustrates how organisational narratives shape technological outcomes. Market classifications, investor expectations, and internal growth strategies influence engineering decisions, testing priorities, and resource allocation, extending prior work on AgTech start-up development and investing (Fernandez-Vidal & Alarcon, 2025; Graff et al., 2020). Our study shows that pasture-based systems become marginal within startup and investor decision-making processes because they are often treated as commercially secondary markets early in product development. Over time, this deprioritisation restricts learning and reinforces design choices that favour controlled environments over pasture-based systems.
Together, these findings show that technology is not merely influenced by context at the moment of use, but is progressively shaped by institutional settings, the organisational routines of startups, and financial logics long before it reaches the farm.
Rather than isolated breakdowns, the findings point to a patterned cycle of misalignment in which ecological abstraction, commercial assumptions, and institutional priorities reinforce one another. Indoor assumptions generate fragile products; commercial metrics misinterpret a natural usage seasonality as customer loss or declining user engagement; and strategic narratives reduce investment in local adaptation. Over time, this interaction stabilises failure rather than correcting it.
This challenges linear models of innovation diffusion that position success or failure as post-launch outcomes (Godin, 2006). Instead, the study shows that technological failure is actively produced inside the innovation process. Our findings indicate that institutional and commercial pressures play a significant role in reinforcing misfits. In our interviews, eighty percent of digital technology executives and board members/investors described pasture-based markets as commercially secondary or lower-value segments, usually because of smaller herd sizes or perceived limits to scalability. These labels shape where engineering resources are directed, which features receive attention, and how teams interpret farmers’ feedback. Scaling expectations from investors further encourages the development of generic, transferable products rather than locally adapted tools.
These patterns contribute to the literature on agricultural innovation systems by demonstrating how organisational priorities of startup companies and investment logics influence the evolution of digital tools. Earlier research by Eastwood (2008), Ingram (2008), Klerkx et al. (2010), Rijswijk et al. (2019) and Fernandez-Vidal and Alarcon (2025) emphasised the importance of advisory networks, knowledge flows, and institutional support for adoption. The present study adds to this work by showing how upstream institutional pressures – such as commercial narratives about ‘attractive markets’ or expectations for rapid expansion – affect the technological fit of dairy tools before adoption even becomes an issue.
Seen in this way, malfunction becomes a sociological outcome rather than a technical accident, extending the work of Carolan (2018), Rotz et al. (2019b) and Bronson (2022). It reflects organisational priorities, imagined users, and the valuation regimes shaping product decisions. The tendency of digital tools to perform better in tightly controlled environments and poorly in open landscapes is therefore not accidental, but the outcome of development systems optimised for certainty, speed, and scale.
Importantly, this reframing also explains why digitalisation progresses unevenly across production systems and expands the work of Bronson (2019), Saruchera and Mpunzi (2023) and Gouthon et al. (2024). It is not that pasture-based agriculture resists innovation, but that contemporary innovation ecosystems struggle to absorb ecological complexity.
Taken together, the findings suggest that digitalisation in pasture-based dairy is shaped not only by farmer behaviour but also by an innovation process that embeds specific assumptions into technology design. This supports a shift from adoption-focused explanations towards the idea of innovation fit, understood as the degree to which a tool aligns with ecological, temporal, and organisational realities.
Finally, the identification of ecological, temporal, and institutional misfits provides a useful way to understand how different layers of the innovation process interact to shape technological performance. Together, these contributions highlight the importance of paying closer attention to how digital agriculture tools are imagined and prioritised during development, not only to how farmers respond once the tools are available.
These dynamics can be represented as a multi-level pattern of system – context misalignment. Appendix B summarises how assumptions formed early in the innovation process cascade through design, engineering, commercial planning, and market prioritisation to create the ecological, temporal, and institutional misfits identified in this study.
Conclusions
This study examined why digital tools often struggle in pasture-based dairy systems by looking at how dairy-tech startups imagine, design, and commercialise their products. The findings show that many of the challenges observed on farms begin earlier in the innovation process, before technologies ever reach farmers.
The results also have practical implications. Public agencies could support shared testing sites and demonstration farms so startups can trial products in real grazing conditions before launch – with infrastructure funded by a combination of public and private funds, as leaving it to individual startups creates an underinvestment problem. Dairy cooperatives and advisory organisations could help startups connect with representative pasture-based farms earlier in development and provide feedback on daily farm routines. Industry bodies could also support common standards for connectivity, seasonal use patterns, and data sharing between systems.
Startups may also need to change how they design and sell their products. More flexible pricing models could reflect the fact that many pasture farmers use tools more during calving, breeding, and feed shortages than at other times of the year. Longer testing periods and more local adaptation may also improve product performance. Lastly, investors may consider incorporating ecological-fit milestones into investment terms.
Like all qualitative studies, this research has limitations. It is based on interview data and does not directly measure technical performance on farms. The sample includes 26 startup companies but does not include large multinational firms, whose development processes may differ.
Overall, the study shows that the success of digital tools in pasture-based dairy depends not only on farmer adoption, but also on how technologies are designed, tested, and prioritised before they reach the farm.
Footnotes
Ethical considerations
Not applicable.
Consent to participate
Informed verbal consent received by all participants.
Consent for publication
Not applicable.
Funding
The author received no financial support for the research, authorship and/or publication of this article.
Declaration of conflicting interests
The author declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.
Data availability statement
Not applicable.
Appendix B: Overview of system – Context misalignments in pasture-based technology deployment