Beyond the Hype: Ten Lessons from Co-Creating and Implementing Digital Innovation in a Rwandan Smallholder Banana Farming System

Abstract

The fourth agricultural revolution (or Agriculture 4.0) promises to lead the way to an agricultural sector that is smarter, more efficient, and more environmentally and socially responsible. Digital and data generating tools are seen as critical enablers for this transformation and are expected to make farming more planned, predictive, productive, and efficient. To make this vision a reality, agricultural producers will first adopt and use the technologies, but this is easier said than done. Barriers such as limited digital infrastructure, low (digital) literacy, low incomes, and socio-cultural norms are major factors causing sub-optimal access to and use of digital technologies among smallholder farmers. Beyond these use challenges of access and usage, limited evidence exists to support the notion that extant digital technologies add enough value to provide substantial benefits for targeted farmers. In this paper, we unravel insights from a six-year digital agriculture innovation project which was implemented to develop and deploy multi-modal digital tools for the control of a major banana disease. By reaching over 272,200 smallholder farmers in Rwanda through a smartphone app, unstructured supplementary service data, a chatbot, and other ancillary channels, we assessed various assumptions regarding intrinsic motivation, incentives, and skills retention among the target digital tool users. These insights suggest that embedding digital innovation requires intentional user-engagement, proper incentivization of next-users, and targeted communication to foster adoption. We present ten (10) salient, but non-exhaustive, lessons to showcase the realities of developing and delivering digital tools to farmers over an extended period, spanning from ideation, development, and testing to scaling stages. The lessons are relevant for a broad audience, including stakeholders across the digital innovation space who can utilize our experiential notes to guide the development and deployment of similar digital innovations for improved outcomes in smallholder farming systems.

1. Introduction

Within the past decade, there has been a significant growth in interest among various stakeholders in the digitalization of agriculture, with major promises to modernize and transform the agricultural sector. These include an array of digital innovations that encompass infrastructure (e.g., databases and networks), hardware (e.g., sensors, IoT devices, mobile phones), software (e.g., apps and web solutions), and services (e.g., advisory and financial solutions) that have been developed to meet the needs of producers and other actors within the agricultural sector. This unprecedented emergence and proliferation of digital technologies in the agriculture sector has been referred to as Agriculture 4.0 [1,2], and there are expectations that a digitally driven agri-food system will address longstanding and emerging sector challenges, fostering an inclusive and positive impact on smallholder farming households and rural communities. Common value propositions about digital technologies in agri-food systems suggest that digital innovations will help to optimize farm productivity, and create more traceable, responsive, and resilient agri-food systems [3]. In the context of low- and middle-income countries (LMICs), digital agriculture often comes with objectives that match the definition of digital social innovation, encompassing the use of digital innovations with the aim of improving the well-being of marginalized groups and addressing the complex (social) problems affecting these groups, including those listed in the Sustainable Development Goals—SGDs [4].

Notably, the agricultural sector is traditionally known to be quite conservative, and the introduction of digital technologies in LMICs takes place in contexts where anticipated users face barriers such as limited financial resources and (digital) literacy, poor electricity and mobile and internet network connectivity, and limiting social and gender norms [5,6]. For example, most of the world’s 500 million smallholder farmers initially faced challenges with accessing digital innovations offered through mobile-based platforms, mainly because of limited availability and affordability [1,7,8]. Fortunately, such access-related issues are becoming less problematic as basic phones are becoming cheaper and affordable to low-income farming households, and with the anticipation of a similar trend for smartphones [9]. This proliferation of mobile phones and the technological features and services that they offer has also positively affected farmers’ access to, e.g., agricultural information and advice, markets, and financial services [10]. Not surprisingly, there has been a boom in digital agriculture services. For instance, nearly 50% Digital for Agriculture (D4Ag) solutions targeting LMICs started in the past 5 years, and 50% of the current solutions are active in Sub-Saharan Africa (SSA) [11]. To date, this translates into 1200–1400 agricultural services and close to 600 deployed in SSA alone [11,12].

Generally, digital tools and services are conceptualized to serve the needs of (smallholder) farmers, with promises of accelerated delivery and impact at scale [13]; however, the deployment is typically punctuated by contextual realities [6], while the envisioned outcomes can be elusive or unattainable for most smallholder farmers [14]. This disconnect between conceptual and contextual realities can potentially eventually translate to the poor adoption of technologies by farmers, lead to loss of investments, and erode the aspiration for improved productivity and livelihoods among farmers [15]. However, prior consideration of core requirements for embedding digital innovations in smallholder farming contexts by practitioners can significantly influence overall outcomes for farmers and stakeholders. Exercising due diligence is important in successfully introducing digital solutions that can address challenges to agricultural production and empower smallholder farmers at a national scale. A case of interest is the banana Xanthomonas wilt (BXW) disease which has been reported as a leading cause of over 50% loss in banana land area and productivity in Rwanda between 1990 and 2020 [16,17,18]. The disease poses a major threat to banana production and can cause up to 100% yield loss per stand [16]. However, banana (Musa spp.) plays a critical role in the food security of East and Central Africa (ECA), it is one of the most important staple food crops in the region [19]. They are mainly grown by smallholder farmers for subsistence and income generation, and nearly one-third of global production occurs in Sub-Saharan Africa, with more than 50% of this derived from the ECA region, including Rwanda, Burundi, the Democratic Republic of Congo (DRC), Uganda, Kenya, and Tanzania [19,20]. National government extension agencies in these countries have limited capacity to support farmers for controlling BXW due to (a) high farmer/extensionist ratio, and (b) lack of access to up-to-date, science-based knowledge. The overall lack of participatory and demand-driven approaches from non-governmental institutions (including international research organizations) are major underlying factors behind the low adoption of control measures by farmers [21,22], and the low buy-in of governments to deploy early warning systems for the proactive mitigation of disease spread.

The multidimensional factors that underpin the spread and prevalence of BXW diseases are generally related to lack of tools and actions in i. monitoring and understanding incidence dynamics; ii. strengthening institutions to effectively collaborate and target resources towards the most affected areas; and iii. empowering farmers for the timely control of the disease. Previous efforts have shown that mobile-based ICT platforms were achieving success in addressing similar problems in other contexts due to their versatility [8,11,23], thus suggesting that digital innovation, including systems and tools, could help to address the prevalent challenges for banana production in Rwanda. Therefore, we embarked on an ambitious project to develop and deploy a digital innovation that can enable real-time information access, diagnosis of BXW, reporting its incidence, and the production of healthy bananas, in collaboration with different stakeholders.

Over a six (6) year period (2018–2023), and within two project implementation phases, a mere conceptual idea evolved into a national intervention that achieved critical milestones, including improved banana health and production in Rwanda. This created a major opportunity to reflect on the salient lessons learned that may help future development actors and researchers to navigate the nuances of embedding digital innovations in similar contexts that involve smallholder farmers. Therefore, the objective of this paper is to share experiential lessons on the process of co-creating and implementing a multi-modal digital intervention for smallholder banana farmers in a Sub-Saharan African country. We reflect on lessons learned from the development, testing, and scaling of digital innovations, and provide suggestions that may enhance future outcomes and impact digital agriculture innovations in similar LMIC contexts.

2. Materials and Methods

The collation of lessons learned presented in this paper was based on documentations from the ICT4BXW project in Rwanda, supported with survey data and direct assessments. Below, we briefly present the general implementation process of the project to guide the understanding and contextualization of the lessons and related insights.

2.1. Study Area

The digital innovation was implemented in Rwanda, a country with a population of 14 million people, out of which approximately 76% are smallholder farmers [24]. Generally, 99% of the farmers cultivate banana at different scales and for different purposes, ranging from homestead cultivation for household consumption to larger farmlands for commercial production, while the typical farmland size per household is 0.5 ha [25]. Banana is often cultivated under a mixed cropping system alongside other crops such as maize, cassava, sweet potatoes, cowpea, and others. The implementation initially focused on eight (8) major banana-producing districts (Figure 1), during the first three years (pilot stage), while the rest of the districts in Rwanda were later included during the final three years (scaling stage). Also, for the pilot stage, 138 villages were selected through a stratified random sampling technique, and these were randomly split (50:50 ratio) into intervention and non-intervention (control) villages to assess the impact of the digital tool deployment. In the intervention villages, selected farmers and farmer promoters were engaged in the development, testing, and scaling of the digital innovation tools, whereas there was no direct effort to reach or engage with those in the non-intervention villages. However, beyond our control, we acknowledge that some knowledge exchange and diffusion of information may have passively occurred between the intervention and control villages.

2.2. [Re]-Defining Digital Tool Users Among Smallholder Farmers

Our approach to co-developing and deploying the digital innovation started from the mere intention to directly engage farmers in the tool development. However, as we gained a better understanding of relevant the entry points to gain traction, our approach evolved with strategic collaboration to accommodate user inputs, stakeholder requests, and salient project team observations. At the development stage, we recognized the inherent operational hierarchy of the national agricultural extension delivery agency in Rwanda, and assessed the advantage for the introduction of new innovations to farmers. In summary, farmers rely on so-called “farmer promoters”, while farmer promoters rely on field extension officers (including sector and district agronomists), and the extension officers in turn rely on principal researchers and technical leads for guidance and direction on priority needs and activities within the agricultural sector. This understanding informed our decision to pivot from our initial conceptual idea towards a more practical approach that leveraged the current extension hierarchy to improve the chances of adoption of the envisioned digital innovation (Figure 2).

Furthermore, during the scaling phase, we conducted a survey to gather basic profile data on 12,653 farmer promoters who were on the official records of RAB. The survey data revealed differing educational levels, competency to use mobile-based tools for extension support, and readiness to engage peer farmers with digital tools. The data insight guided our approach to select top-ranked farmer promoters and categorized them into two (2) groups, namely the Scaling Champions (SCs) and Scaling Enablers (SEs). Specifically, a total of 1069 of the most promising farmer promoters were selected from the pool of prospective 12,653 farmer promoters, based on tallied scores for their self-reported profile, including age, extension experience, village-level leadership, and ability to use digital tools. These selected cohorts were further ranked based on the same criteria, and assigned as SCs (n = 134) or SEs (n = 935). While both SCs and SEs are farmer promoters, the required commitment of each group to support innovation uptake and further dissemination was different. Specifically, the SCs were directly engaged, facilitated with incentives, and tasked to actively support other farmers in their villages while mentoring peer farmer promoters. In contrast, the SEs were not directly engaged or facilitated with any incentives, so they were only offered access to the tools with the understanding that they can use and support other farmers at will, but without any defined obligations. More details about the relevance of the differentiated roles and experiential outcomes of SCs and SEs are presented under the lessons learned.

2.3. Development and Deployment Process of Digital Tools for BXW Control

The digital innovation for BXW disease control was implemented in two phases. During the first three years (phase 1), the focus was on the development and testing of a single digital solution, but the focus shifted to diversifying from the single solution to scaling as a multi-modal solution during the latter three-year period (phase 2). In phase 1, we adopted and implemented a stepwise co-creation process, called a participatory technology design approach (participatory iterative technology design, PITD), and progressively gathered user inputs, which strengthened user engagement and improved the experience. The PITD co-creation methodology was grounded in human-centered design (HCD) and design thinking theories, guided by the notion that such an approach can increase the likelihood for adoption and sustainable outcomes and impact [26,27,28,29], while fostering social inclusion [15]. Theoretically, PITD leverages the various existing models for HCD, encompassing design orientation, research orientation, and bi-directional inclination between users and researchers and/or designers [26]. Our implementation process included cohesive engagement between stakeholders, mainly anticipated users, researchers, value chain actors, technology developers, a government agency, and non-governmental organizations, prior to the full deployment of the digital tool. As shown in Figure 3, the PITD flow involved four progressive stages, which included quarterly contact sessions, data gathering, and reviews, over a 12-month period (August 2018–August 2019). During the early stages of the process, we conducted immersive co-creation workshops with selected representative end-users to map persona, define user journeys (including fail and pain points), create mock-ups and prototypes, review and iterate minimal viable product versions of the digital tool, and pre-test the beta version. In the last stage of the PITD, we conducted field testing of the co-developed digital innovation, and this supported the transition from phase 1 to phase 2 of the project, driven by the positive user experience and diverse feedback from the on-farm interaction of farmers with the core tool (i.e., a mobile-based app, v2.0.101). The transition focused on scaling the core co-developed tool, enabled by training and facilitating farmer promoters, with desk-based support through phone calls and dedicated WhatsApp groups for the timely resolving of technical issues. Performance benchmarks were preset for field activities of the next users, primarily based on an agreed target for the farmers (end-users) to reach in their respective villages within an 8-month period. Progress towards the target was assessed through a self-reporting activity logbook and follow-up (phone-based) surveys to tally the number of farmers reached by each next user.

The value proposition for a BXW-focused digital innovation is centered on the proactive prevention and timely control of BXW disease through best practices for banana crop and disease management. Therefore, the tools provided information that included i. actionable guidance on how to avoid infestation and the spread of BXW; ii. step-by-step guides to implement the cost-effective on-farm control of BXW incidences (mainly the removal of the diseased stem); and iii. how to maintain healthy banana mats. Generally, the management practices to reduce the vulnerability of banana farms included the removal of male buds, the thinning of banana suckers to optimize mat density, the cleaning and disinfecting of cutting tools, and the cutting of any diseased stem at soil level. This is overlaid on broader recommendations regarding the sourcing of banana stocks, the optimal spacing of banana stands, when to plant or harvest, fertilizer types and application rates, and the staking of stems.

2.4. Scaling Approach for Broader Use and Impact

The scaling of the tool during phase 2 was guided by a re-imagination of the digital solution (BXW app v2.0.101), which was the major digital tool output from phase 1. In phase 2, the smartphone-based app was reassessed and redefined as an innovation package composed of multiple core innovation elements, which included i. basic information on BXW for enhanced awareness; ii. remote extension training for capacity building; iii. stepwise BXW diagnosis for incidence surveillance data; iv. agronomic and BXW control advisory for decision support on banana health; and v. early warning alerts for timely control action. Each of the elements served different functions in the delivery of the overarching goal of empowering users for BXW surveillance and control. The relevance or direct use of each element was differentiated by the profile of target users and their role within the agricultural system; therefore, the content and resources were deployed through different digital channels to cater to different user contexts. For instance, the elements for “BXW Information” were deployed through a smartphone app (BXW app v2.0.101), short message service (SMS), and unstructured supplementary service data (USSD) channels, while the disease surveillance was deployed mainly through the BXW app since this allowed for the reporting of BXW incidence with georeferenced coordinates and linkage to a functional back-end dashboard to ensure near-real-time dataflow and access (Figure 4). Generally, text-based and audio-based contents were accessible through both smartphones and basic phones, while the capability for the reporting of georeferenced BXW disease incidence, videos, graphics, and pictures was restricted to the smartphone-based app.

It should be noted that the scaling phase focused on empowering farmers nationally to improve the health and productivity of their banana farms, but the majority of the end-users were constrained by poor or a lack of smartphone access. This reality informed the decision for multi-platform deployment in the innovation package, which in turn fostered the strategic delivery of the target solution to farmers either through smartphones or basic phones. The SCs and SEs were mainly nudged to leverage their personal or project-provisioned smartphones to provide direct advice to farmers within their respective villages. The broader population of farmers, who predominantly relied on basic phones, individually (and independently) accessed the relevant resources through USSD and SMS.

2.5. Data Gathering and Synthesis for Lessons Learned

We reviewed major documents and knowledge resources that were generated during the implementation of the project in Rwanda, including peer-reviewed articles, newsletters, blogposts, activity reports, and meeting minutes (

Table 1. Published resources that underlie the synthesis of lessons learned on embedding of digital innovation for healthy banana production in Rwanda.

Year	Area of Focus	Implementation Stage	Reference
2018	Introduction of ICT-based approach for BXW control	Development	[20]
2021	Reality check on extension tool development	Testing	[6]
2021	Assessment of farmers’ readiness to use digital tools	Development, Testing	[31]
2021	Use of BXW surveillance data for banana land area mapping	Development, Scaling	[32]
2021	Evidence to support peer-based targeting of digital tool and extension for banana farmers	Scaling	[33]
2022	Digital rights, access, and inclusiveness of farmers in tool development	Testing, Scaling	[34]
2021	Cost and time efficiency for control of BXW disease	Testing	[18]
2022	Heterogeneity of mobile phone usage among banana farmers	Testing, Scaling	[33]
2023	BXW risk modeling for climate-related early warning alert	Scaling	[35]
2023	Website for relevant project information and non-peer-reviewed publications	Development, Testing, Scaling	www.ict4bxw.com (accessed on 28 Dember 2024)
2024	Relevance of farm(er) typology for BXW management	Scaling	[36]

). Additionally, we synthesized feedback notes from the tool users and convened semi-formal reflection sessions with project stakeholders. A total of 19 major lessons were initially collated, with supporting data from baseline, mid-line, and end-line surveys of users. Following the innovation framework approach in the existing literature [26,30], we organized the initial list of lessons into the three major stages of innovation, including development, testing, and scaling. A second round of review was conducted to refine and consolidate the initial list of lessons into final ten lessons that are considered important for a broader research and development audience.

3. Results

Below, we present the specific lessons from project-level experience and outcomes, supporting each lesson with related data where applicable. Unequivocally, the highlighted lessons are inexhaustive, but we consider that these lessons are generally important for the innovation delivery journey. We reiterate that the lessons are organized and presented progressively to follow the development, testing, and scaling stages of the project implementation. As the project progressed along the innovation continuum, we transitioned from the development stage, where we only offered various solutions within a single mobile-based app for BXW control, to the scaling stage, where we redefined each solution as standalone tools or elements within an innovation package (Figure 4). Considering that each element served different functions and catered to different user contexts, it was clear that they can either be bundled into one tool (such as a mobile app) or they can be deployed separately to address specific needs, including extension training, information dissemination, data collection and reporting, etc. More details about the rationale and implication of the approach are presented in later sections.

3.1. Development Stage

3.1.1. Lesson 1: Define the Innovation Challenge but Co-Define the Solution

The project adopted broad principles of design thinking, human-centered design, and inclusive design from the onset. This resulted in a tailored approach and consideration of nuanced contextual dimensions while developing the first digital solution: the BXW app. Our adoption of the PITD process illustrates how (research for) a development project can co-design a digital intervention together with anticipated users. Furthermore, the inclusion of diverse users was valuable in fostering new ideas and creativity, exploring complex issues from diverse angles, and cultivating equity in influence and decision-making power [37]. The inclusion of farmers, with their local and indigenous knowledge and skills, in innovation design and implementation also legitimized the digital intervention approach [38].

Inclusiveness at the design stage often portrays that each actor or stakeholder has a leading or supporting role to play for the effective delivery of the target digital tools, notwithstanding the initial visioning of the problem scope or dimensions. Prior to the PITD process, the project team conceptually predefined a digital solution that was intended to provide basic information to farmers and retrieve data on disease incidence. Yet, there was no clarity about how this solution would operate, how users would engage with the envisioned tool, what enablers or blockers were inherent within the target innovation space, and what were the preferences or perceptions of the stakeholders in relation to the use of digital technologies for banana farm management. The PITD process unraveled these important considerations and fostered agile implementation. For instance, at the initial stage, 28 stakeholders (including farmers, developers, governmental and non-governmental organizations, market actors, and academic researchers), who were invited from 25 organizations, collaboratively accomplished the initial scoping tasks. In summary, they (1). identified five major challenges within each of their respective stakeholder groups in relation to the banana value chain; (2). identified the viable entry points for a digitally driven approach to solve the BXW disease problem; (3). highlighted and confirmed their demands for the envisioned innovation; and (4). challenged assumptions regarding the willingness and readiness of farmers to adopt digital tools. At the end of the PITD, the stakeholders’ input sharpened focus on the need to at least deliver a tool that can demystify the BXW problem for farmers, foster their access to control measures in real time, empower them to identify the presence of the disease in their farm, and foster transparency of the trend of BXW spread to drive collective action.

3.1.2. Lesson 2: Stay Contextual but Also Aspirational

After co-defining the target digital solution, it is important to assess and align with the innovation context. The first step should include cursory horizon scanning to generate rich information on contextual realities. This can include a baseline survey, focus group discussions, or thematic desktop research to establish a baseline and curate data-driven insights regarding the innovation contexts such as geography, demography, agricultural production system, target problem, and targeted innovation user personas, etc. Through the ICT4BXW baseline survey, we generated data which highlighted the knowledge, preferences, and persuasions of banana farmers in Rwanda regarding the potential use of a digital tool to control BXW. For instance, we unraveled the dependence of most farmers on radio (73% accessed information this channel for news/information), documented that four out of five farmers have access to basic phones, and recorded that most farmers (72%) are willing to pay for access to information through digital tools.

Staying true to contextual realities does not directly connote an over-riding of relevant aspirations that may include the utilization of advanced or new technologies, or expanding the (digital) competencies of the target users. For example, while it was evident that most of the farmers were incompetent with (and lacked access to) smartphones, we proceeded to train a few farmers and equipped them with smartphones. Simultaneously, we continued developing a smartphone application with a guided expectation that smartphone ownership and digital literacy will improve among farmers over time. Our final project-level assessment justified this outlook because the data showed that 13.7% of the sampled banana farmers reported ownership of a smartphone during the end-line survey, which indicates over two-fold increase when compared to the baseline survey where 4.8% owned or had access to smartphones. Interestingly, current mobile phone access hovers around 97%, with about four out of every five farmers owning or having access to a basic phone.

In the (near) future, newer technologies and more advanced functions will become available in the agriculture sector, as already confirmed by the rapid development and growth of artificial intelligence and natural language processing [39]. Users who possess higher education and higher incomes generally have more positive attitudes toward these emerging technologies and are in the best position to become early adopters [7]. While low (digital) literacy levels among farmers persist as a barrier to the adoption of digital tools, the majority (93%) of our surveyed smallholder farmers indicated that they are highly motivated to use digital tools to support their farming objectives and increase productivity. Therefore, the process of encouraging and empowering late tool adopters must be intentional, guided by a conviction that the incremental engagement of the less-educated and less-endowed users with mobile-based digital tools can result in collateral advantage, accelerating the delivery and adoption of other digital solutions at scale.

3.1.3. Lesson 3: Define and Co-Create Target Solution with Users

User engagement and the progressive integration of user inputs into the product can significantly impact uptake, yet it is often one of the least prioritized aspects of innovation design for smallholder agricultural systems. Developers and researchers often hold the notion that farmers (as end-users) are not adequately sophisticated or literate to provide useful inputs at the onset of innovation design [8,15,40]. In contrast, gathering relevant user inputs from the onset of the design enables researchers and developers to become cognizant of the nuanced user requirements which may shape the acceptability and trustworthiness of the product down the line and foster equity and social inclusion in digital agriculture. Under the ICT4BXW project, we recognized the importance of segmenting the users based on the understanding of extant relationships between the different user groups, so we leveraged social network and ego-network analyses [36]. Our key consideration was to strategically identify and engage users who can bridge the gap between the high-level technical process and ground-level engagement process with less-educated farmers. Based on our baseline survey and interactions with the extension agency in Rwanda, we defined our target users as “next users” and “end-users”. Concomitantly, the next users are relatively better positioned than end-users to embrace, adopt, and promote tools among peer farmers. While the farmers are designated as end-users, the village-level extension agents (i.e., farmer promoters) were considered as the next users of our digital solutions because of their knowledge and influence within the community, combined with their government-assigned duty to support peer farmers within their respective villages. The engagement of both next users and end-users in the co-creation process fostered a mutual exchange of perspectives on the digital solution, which resulted in the decision to create different user journeys and permission levels within the BXW app, mainly based on user roles as farmer, farmer promoter, or core government extension staff.

As we unpacked the village-level interactions, influence, and power dynamics [31,34,36], it was clear that the cohesive engagement of the next users along with the end-users will help to incorporate rich user perspectives into the design process. Therefore, the selected farmer promoters, who were involved in the PITD process, provided valuable inputs that helped to define and refine the design elements (including language, color, illustrations, gender representation, and cultural norms) of the initial digital tool output. Also, the co-creation process revealed that the next users required short-structured training to effectively understand and navigate the tool functions. Therefore, self-paced training modules were developed and deployed through an interactive voice response (IVR) channel. We hypothesized that the deployment of the ancillary training could foster adoption by enhancing the capacity of next users to use and articulate functions of the tool while strengthening overall extension delivery. The training modules were developed with minimal expectations, considering that the target users possessed mid- to low-literacy levels. Therefore, it was surprising that 98% of the selected next users independently completed the assigned three modules through a virtual delivery mode (i.e., through IVR), and cumulatively achieved a 35% increase in knowledge on core topics related to digital tools, BXW prevention and control, and banana agronomy.

3.1.4. Lesson 4: Think and Rethink User Incentives

Users are driven to adopt innovations based on specific motivations, which can be both intrinsic and extrinsic [15,41]. The agricultural innovation system is often skewed with the erroneous assumption that farmers are generally inclined to access and use new tools because the tools are pitched as useful or promising for their farming operations or related activities. Although this may be partly true (especially for innovative farmers who are not risk-averse), rightly structured user incentives are important to attract and sustain the engagement of users. This applies in both the short (in the piloting and testing phase), and long term (in the deployment stages), and particularly for digital solutions that do not lead to an immediate offline benefit.

We realized that without compelling incentives and nudges, most farmers were unlikely to use the BXW app or provide feedback, especially since the value proposition was yet to be proven. This consideration guided ICT4BXW implementation as we sought to answer two fundamental questions: 1. What are the add-on incentives for farmers to use and reuse the solution beyond the initial surveillance of BXW? 2. How can farmer promoters (as next users) be effectively nudged to consistently engage farmers with the BXW app and offer valuable support to their peer farmers at the village level? Regarding the first question, the selected next users (farmer promoters) confirmed interest in supporting the development and testing of the BXW app so they provided input into the development process, including advising the project team on viable approaches to generally incentivize farmers to use the tools. For instance, since the envisioned solution was targeted towards the surveillance and control of BXW disease, the farmers requested the inclusion of ancillary information contents as a compelling incentive for tool (re)usage. Their input informed the project team’s consensus to embed a module that provides the best agronomic practices for banana production in the BXW app and later in the USSD channel during the scaling phase.

The second question relates to appropriate incentive structures for motivating the next users in their line of duty towards engaging their peer farmers at the village level. By conducting an initial mini-survey on the next users, we assessed their access to basic infrastructure (including smartphone ownership and internet access) and gathered data on their preferred incentive modality (including bank transfer, airtime vouchers, or mobile money). The responses guided the project teams to equip selected next users with smartphones and provide them with financial micro-incentives, including transport fares to engage in user sessions and internet vouchers for testing the tool. In effect, the offered incentive contributed to the attainment of high engagement on the part of the next users (97% actively engaged and provided feedback), with a flow of timely and useful data from over 120 villages across Rwanda.

3.2. Testing Stage

3.2.1. Lesson 5: Facilitate User Champions, Not Only User Interaction

In the agriculture innovation space, the potential for the uptake of promising innovations can be limited by recurring disconnections between conceptualized solutions and the reality of end-users (mainly farmers) who are expected to adopt specific innovations [5,6,42]. This disconnect often occurs due to various factors, including differing understandings of the problem space, non-aligned perspectives in the solution space, and end-users’ lack of competency to test or use multi-functional digital tools. Therefore, we considered that the targeting and engagement of champion tool users as intermediate bridges (for testing at the village level) can significantly enhance the reach of peer farmers and subsequent uptake of BXW solutions.

Notably, as technology end-users, farmers require frequent interactions with such champion users to establish a good understanding of new innovations and implement relevant actions in their farm operations. However, in practice, practitioners fall short of this requirement and merely convene a few demo or interaction sessions (at best), with a presumption that the target user base will attain a near-instant grasp of the proffered solution or deployed tool, followed by rapid adoption and scaling. In contrast, our experience illustrates a need to foster deeper and organic connections with (and among) farmers as end-users of digital tools. To transcend mere interaction between researchers or developers and end-users, we selected and facilitated 35 farmer promoters, as champion users, who continuously engaged and re-visited the end-users from 25 villages during the testing phase. The interactions between farmers and the champion users was based on pre-existing trust and enabled the gradual transfer of knowledge and skills through social exchanges while using the BXW app. This proved to be beneficial in terms of scaling the innovation package, as evidence from the project end-line survey showed that 9 out of every 10 farmers (94.9% of respondents) indicated that they felt more comfortable contacting farmer promoters in their village for support on the use of BXW app tools and the deployment of control measures.

3.2.2. Lesson 6: Create Space for Real-Time Support to Document and Resolve Fail and Pain Points

The assumption that new technologies will work seamlessly without glitches is far from the reality. Moreover, with farmer-facing tools, there is a high likelihood that a combination of the potentially steep learning curve and inherent bugs in new digital tools can create significant pain points for the next or end-users. This creates a major need for a responsive support system that provides a safe space for the users to lodge and resolve issues. During ICT4BXW implementation, the initial four (4) weeks after deployment were most critical because some of the farmer promoters had never independently owned or managed a smartphone. They struggled with basic issues related to literacy about mobile-based digital tools, such as confusion regarding icons, overloading of the phone storage memory with various personal contents leading to the crashing of the application, forgetting the navigation steps, device care, etc. (Figure 5). By creating a user support desk during the first four (4) months of deployment, we progressively and effectively documented issues and bolstered user confidence by either providing solutions directly or advising that the issues are noted for resolution in the next iteration of the tool. Based on the follow-up surveys that were conducted, 95% of the flagged problems were resolved through real-time support, which enhanced the confidence of the next users as they eased into smartphone usage and working with the mobile-based digital tool (BXW app v2.0.101).

3.2.3. Lesson 7: Assess Perspectives and Experiences of Next and End-Users to Guide Further Iteration

An assessment of user experience and sentiment during or after a validation process can generally indicate the potential for the sustainable adoption of a (digital) innovation at scale. This may include focus group discussions (FGDs) and user surveys that allow the developers, researchers, and administrators to gather new information that may [in]validate assumptions about the potential uptake and readiness to scale the innovation. For instance, the insights from our post-validation FGDs and surveys showed that the farmer promoters felt empowered (in competence and confidence) to engage the farmers in their village with the audio-visual contents that were embedded in the BXW app, but the experience was initially marred by poor internet connectivity in many locations, leading to the continuous buffering or poor resolution of the video. This informed the developers that they had to compress the size and optimize the resolution of the audio-visual content for enhanced compatibility with low-bandwidth conditions, coupled with a reconfiguration of the android package kit (APK) file so that some content can be downloaded locally into the phone during the installation of the app to improve users’ experience.

Beyond the initial inputs of next users during the development stage, their perspectives and experience at the testing stage became valuable, because they were based on the actual use of the innovation within farmers’ fields and enriched by their one-on-one interactions with peer farmers. At this stage, expected and unexpected context-specific conditions (including mobile broadband connectivity, terrain navigation, phone conditions, farmers’ disposition, or availability, etc.) shaped the experiences of the next users and their feedback. Therefore, the gathered insights from the field-level testing were taken forward into further iterations to fine-tune both the features of the BXW app and the deployment process for scaling.

3.3. Scaling Stage

3.3.1. Lesson 8: Re-Assess the Innovation and Define the Scaling Strategy

During the early stages of developing the ICT4BXW innovation, it became clear that reaching diverse users with a smartphone-based digital intervention, which required ownership and use of smartphones, was not realistic. This was due to the generally low ownership and use of smartphones in Rwanda and digital literacy limitations among smallholder farmers, more so among certain user groups (such as older female farmers with little or no education). Therefore, our first step towards scaling involved a reassessment of the opportunities and limitations associated with BXW app usage among the target farmers, followed by a re-conceptualization of the tool as an innovation package or toolkit. We convened a 2-day scaling workshop with stakeholders, and agreed that the individual elements of BXW innovation can be scaled through different channels to effectively reach the diverse target users and circumvent known bottlenecks such as disparate access to internet or smartphones. As noted earlier, these elements include information for awareness, capacity development, BXW diagnosis for surveillance, BXW alerts for early warning, and banana agronomy advisory. Also, this scaling strategy was layered on an improved understanding of the user base (i.e., farmers and farmer promoters), characterized by different levels of resource access, influence, and competencies [36]. For instance, the remote extension training tool (for capacity development) was offered to SCs and SEs through IVR, while banana agronomy and BXW control tools were delivered directly to farmers through SMS and USSD channels.

The approach of defining which elements of the innovation package are appropriate for specific user groups became a basis for the targeted scaling of the innovation, which resulted in an increased number of engaged farmer promoters from 65 to 1069 and accelerated delivery to empower more farmers, from 492 farmers in the pilot phase to 272,567 farmers in the scaling phase. A similar approach can be replicated for digital innovation projects that are focused on smallholder farmers within global south contexts for cost efficiency and the potential ease of reaching prospective innovation users at scale, with focus on offering relevant elements that match the needs and statuses of specific user groups.

3.3.2. Lesson 9: Diversify Communication and Incentivize Access and Use

One of the major aims of scaling an innovation is to facilitate the dissemination and adoption of a successfully piloted solution. In simple terms, this means moving from a limited pool (sample) of next and end-users to reach a broader population of end-users, with the goal of expanding adoption and impact. Effective communication and the use of appropriate incentives are important in this phase.

During ICT4BXW implementation, it was clear that these two factors were indispensable to nudge SCs (and SEs) towards providing hands-on support to prospective end-users (i.e., peer farmers) of the digital innovation at the village level. In the context of smallholder farmers, the process of communicating is not trivial. Issues encountered include language barriers (mainly farmers’ limited use of the English language), the messaging platform, poor signal reception, phone damage, and prior reliance on face-to-face interactions for extension information, among other factors. Therefore, our strategic communication approach was diversified to include WhatsApp-based group chats, follow-up phone calls, SMS texts, printed brochures, web-based blogs, radio jingles, TV panel discussions, and quarterly newsletters. Although each channel served different needs, together, they supported the major goal of fostering continuous interaction with the SC cohort, creating awareness among village-level farmers (and SEs) and keeping stakeholders abreast of progress in embedding digital innovation in Rwanda’s banana production system.

Financial incentives were considered inevitable for the SCs to cover transport costs between farms or households and internet bundle subscriptions for timely data auto-upload from their devices to the back-end database. Moreover, our choice of performance-based incentives, which emerged from deliberations with the technical and non-technical collaborators, proved to be effective in nudging SCs to use the BXW app, disseminate information about other tools, and support their peers. The pre-requisite for each SC to receive earmarked incentives was based on pre-defined benchmarks, which included the completion of BXW diagnosis in four farms (or more) per month, with at least one visit to farmers’ field per week, and interaction with at least three SEs to provide peer-to-peer mentorship on tools and general extension delivery. The SCs who attained the performance thresholds received micro-incentives in the form of internet vouchers on a bi-weekly basis, with additional mobile money transfers to offset transport costs that they may have incurred. Performance assessment results showed that up to 63% of incentivized farmer promoters (i.e., SCs) stayed active monthly and supported other farmers with the BXW app during the final year of the project, compared to less than 5% of non-incentivized farmer promoters (SEs). It should be noted that the SCs had the choice to stay active or inactive, so while a SC may be active in a month, they may opt to be inactive in the succeeding month, depending on their availability, persuasion, or other influencing factors. Notwithstanding, the contrast in percentage of active SCs compared to SEs suggested that incentives influenced the adoption and continuous use of the digital tools among the incentivized farmer promoters, and consequently the overall reach and interaction of farmers with the digital tools.

As project outcomes point towards evidence regarding the impact of incentives on the engagement of farmer promoters who are positioned as next users of the deployed digital innovation, it is equally relevant to unravel the inherent disincentives that may have caused absolute non-engagement or sub-par engagement in some of the mobilized SCs. Inactive SCs were occasionally nudged with messages and phone calls, resulting in few turnarounds, but there was no ancillary data collection to assess their reasons for inactivity. In retrospect, this could be valuable in shaping future directions in terms of scaling efforts, but it was a missed opportunity for the project.

3.3.3. Lesson 10: Adopt a Reflexive Learning Approach: Monitor, Evaluate, and Do It Again

Evidence of impact is critical to successfully iterate digital tools and provide solutions that meet user needs. Deployment of digital tools should be guided by relevant hypotheses and evaluated based on clear and empirically sound evidence. At the onset of the ICT4BXW project, the stratification of intervention and non-intervention villages, and subsequent gathering of baseline, mid-line, and end-line survey data, supported the collection of data to test various hypotheses, including the influence of access to digital tools and information on farmers’ perspective and persuasion for farm-level action on BXW and banana agronomy. For instance, 45% of farmers in the intervention villages took action to control BXW disease during the project implementation period compared to 15% in non-intervention villages, notwithstanding their age and gender. Also, by conducting periodic assessment surveys during the scaling period, we generated progressive user-reported evidence regarding the impact of the innovation elements on specific outcome criteria, including at least three out of four farmers reporting a reduction in BXW prevalence, improved banana health, improved nutrition, and enhanced livelihood. Additionally, farmers’ response to our impact survey showed that 83% of farmer respondents reported that BXW incidence had reduced in their farms, while 54% of these farmers attributed the lower BXW disease incidence to access and use of the deployed digital tools for BXW control. Interestingly, we observed a gender disparity in the impact and attribution of change in disease incidence. Most female farmers (88%) reported a reduced incidence of BXW, compared to a lower percentage for male farmers (74%). At least one out of every two female farmers attributed the positive change to the BXW app, in contrast to one out of every three male farmers.

From this perspective, there is still a lack of disaggregated evidence about the use of digital tools, beyond a focus on the gender and age of users. Therefore, the relevance of digital solutions to diverse sub-populations remains unknown [11,43,44]. Similar limitations accrued to the ICT4BXW project because the focus was mainly on the gender (male/female), age, and socio-economic status of the next and end-users. Yet, previous research in Kenya indicates that other factors, such as cognitive and emotional burdens associated with divisions of household and farm activities, can influence decision-making and the adoption of novel technologies [45].

The agriculture research for development (AR4D) sector is traditionally strong in conducting research on agricultural innovations but not necessarily in evaluating, marketing, or scaling these innovations. Lessons can be learned from (industrial) product design, particularly in relation to mechanisms for the continuous eliciting of feedback from users, and subsequently iterating the design of digital products and interventions. For instance, CGIAR’s User Research Toolkit [46] and USAID’s Inclusive Digital Design Toolkit [11] both provide recommendations for suitable methods to gather feedback, evaluate progress, and communicate impact. This is not a trivial task, because the digital agriculture sector suffers from a general lack of quality impact assessments, thereby creating challenges regarding evidence and transparency about the true impact of digital technologies, including unintended or possible negative consequences [8,11]. In summary, more research is necessary to collect (non-anecdotal) evidence of the positive and negative impacts of digital technologies, and particularly research that disaggregates the impact on social inclusion and equity among user groups. Researchers for development organizations are in a good position, and presumably have the mandate, to become front-runners in collecting solid evidence of the real impact of digital technologies on the agricultural sector and for diverse groups of technology users and non-users.

4. Discussion and Conclusions

In this paper, we presented lessons learned from the implementation of a digital innovation project that targeted crop disease surveillance and control within the context of the smallholder banana farming system in Rwanda. These experiential learnings can be useful in supporting further transformative journeys in digital agricultural innovation. Obviously, the process of ideating, testing, and scaling digital tools for and with farmers is not linear or trivial, so the initial drive of technology enthusiasts to develop and deploy tools that solve the so-called “wicked problems” of farmers can be punctuated (at best), or fully stalled (at worst), as they navigate the twists and turns of delivery along the innovation continuum. Therefore, it is important to note that the lessons presented are non-exhaustive, and their relative importance may also be subject to the context of the digital innovation and the targeted agricultural problem. We refrained from prioritizing or ranking the lessons; rather, we intentionally followed the order of the implementation stages. We expect that such documentation of insights about best practices for digital innovation development and deployment within smallholder farming systems will grow as data and systems become more open for cross-institutional learning and scaled impact.

The implementation of the ICT4BXW project in Rwanda over an extended period (6 years) allowed for an uncommon breadth of evidence-gathering in a nascent field where quick wins and big transformation promises are often elevated above incremental learning, multi-modal testing, and the progressive engagement of users. Arguably, digital agriculture is now transcending the initial hype, with the increasing accrual of more knowledge and evidence about the practicalities of developing, deploying, scaling, and sustaining innovations. As presented in this case study, hybrid approaches are emerging, with less focus on introducing new sophisticated solutions, and more focus on improvements, adaptations, and the integration of existing solutions, underlined by collaborations or mergers between diverse service providers and practitioners. We anticipate that this will unlock more opportunities to test and [in]validate more assumptions regarding user and technology readiness, with an outlook for demonstrable impact and incentives for sustainability at scale. It is important that designers, researchers, and implementers align with a unified perspective, seeing digital solutions through a lens of providing long-term contributions to agri-food system transformation, climate change adaptation and mitigation, and biodiversity [47]. This perspective can foster a better perception around “successful” or “failed” technologies, nudging stakeholders to embrace lessons learnt with each iteration or intervention, and to intentionally incorporate design changes for next-generation tools to empower farmers for timely decision support and inclusive agricultural development.

Further, digital agriculture innovation practitioners are frequently confronted with questions about the business cases, business models, and sustainability pathways for their tools or solutions. These are all pertinent for donors and stakeholders, who are generally enthusiastic about the introduction of novel solutions at scale and equally hard-pressed to ensure that achieved impacts are not short-lived. Notwithstanding, donor funding and impact investments have a limited lifespan, and profitable business models are difficult to achieve when also aiming to be inclusive and socially responsible. A major challenge in the Digitalization for Agriculture (D4Ag) space is the financial sustainability of digital interventions/solutions and their long-term chance of survival. Considering that the use of digital technologies in SSA’s agricultural sector is nascent, it is logical that there has been a rapid expansion of solutions and services that are offered to farmers over the past few years. Many organizations have developed proprietary digital tools or collaborated with AgriFoodTech start-ups [48]. Very often, the development, piloting, and implementation of these tools and services is highly subsidized and development-oriented [13], and many of these solutions are unable to reach full envisioned maturity in terms of adoption by a large user base and evolving toward a sustainable financial model. So, these solutions inevitably fail to demonstrate sustainability.

There are several theories regarding the scaling of innovations; however, practitioners coalesce around the notion that scaling strategies should invariably result in a broader impact for more end-users [49]. At the ideation and piloting stages of innovations, an improved understanding of stakeholders’ demand and the problem dimensions can lead to a more robust offer of solutions that cater to the pertinent needs of end-users. Researchers’ recognition that the solution space is likely to evolve along the innovation continuum is important in guiding the choice of the scaling strategy and pathway. While there is typically a pervasive tendency to create an out-of-the-box or standalone solution, it is arguable that practitioners can improve the efficiency of digital innovation delivery by embracing a (frugal) mix of existing digital technologies to reach more and more diverse users. This approach has been instrumental for our embedding of a digital innovation within Rwanda’s agricultural system, and the lessons shared in this article are instructive for (similar) future efforts. Finally, this article is intended for a broader public and written to engage stakeholders along the innovation delivery continuum who can relate, interpret, and apply the experiential nuggets. We anticipate that the insights will support future endeavors and practitioners who are seeking relevant knowledge to guide their innovation efforts within smallholder agricultural systems.