D4Ag solutions
D4Ag solutions will increasingly use cutting-edge technologies – fuelled by new sources of data and improved analytical capabilities – to increase their value proposition. This will enhance the
precision and relevance of D4Ag solutions, even as they become easier for farmers to access and use. We have seen signs of this trend in our research; over one-third of the respondents to the CTA-Dalberg survey already use at least one form of advanced technology – defined here as drones, augmented/virtual reality, blockchains, machine learning, the internet of things (IoT), big data, artificial intelligence/machine learning, and voice activated technology.325
Nearly 60% of respondents expect to
integrate new technologies over the next three years, the most popular of
which are IoT, blockchains and machine learning.326
As discussed in depth in our overview of emerging D4Ag solutions in Chapter 2, we already saw many examples of how sector actors are making use of these data to enable more tailored, precise, real-time recommendations for farmers; give financial service providers the ability to better assess and control risks; and provide valuable insights into smallholder supply chain needs and opportunities for agribusinesses.
While we are excited about the promise of advanced technologies and the growth in data, many technologies (e.g., drones, field sensors) will likely remain in the experimentation phase in the African smallholder farming context for years to come and do not yet have fully settled business models, or at least not yet at scale. It is therefore important, as D4Ag actors experiment with these technologies, that they continue to capture the evidence needed to build the business and impact cases such as technology investments.
We are already seeing an explosion in raw data capture from a range of sources, yet the agriculture data ecosystem remains fragmented. The
sheer amount of data collected has increased exponentially.327 This includes farmer data,
them to generate powerful insights. On a more institutional and policy level, it is becoming increasingly clear that data aggregation is only possible with better defined data regulations and innovative data-sharing business models; progress on both of these fronts is at an early stage.
Strong data analytics capacity – essential in deriving insightful recommendations for farmers from increased data volumes – is developing rapidly but currently lags behind the pace of data generation and capture. Data analytics and machine learning
– two methods by which to leverage these raw data – are in more experimental stages but are quickly improving. There are many forms of data analytics, each of which serves a distinct purpose: descriptive, diagnostic, predictive, prescriptive or cognitive.329 A handful of
agriculture sector actors have begun to experiment with integrating those capabilities into their businesses. The most common models to date have involved specialist agriculture data analytics vendors who collect, analyse, and sell data to interested parties, or in-house teams that accumulate data from other places.330
The focus for many players over the next three years will be on continuing to improve the quality of data capture and then developing meaningful, actionable insights from these data sets. Big data has
data. The trend is explained in part by the ubiquity of mobile phones (e.g., mobile surveys), but a number of other technologies facilitate agriculture-specific data capture at even greater scale and lower cost – namely, drones, sensors, and satellites.
The data capture from these sources continues to get better, faster, and cheaper, which has led to a growing wealth of available information for both D4Ag intermediaries and farmer end-users. However, despite the growing volume and promise of data, we are still seeing a very fragmented data ecosystem, with many valuable datasets – including much of the data from the public agronomy research community at national and regional levels – locked in organisational silos, not fully digitised, or embedded in proprietary systems owned by financial institutions and agribusinesses.
Sector actors have started to recognise the importance of aggregating data.
These is a growing focus in the sector, led by open agriculture data initiatives from organisations like GODAN and the Open Data Institute (ODI), on ways to ensure that whatever data are captured are stored in an accessible, usable format, and are employed by a broad range of players to improve farmers lives.328 On a technical level, cloud storage and
big data analytics tools facilitate the low-cost storage and aggregation of data in ways that allow others to easily access them and use
an important role to play here; we expect the ‘winners’ to be those who are able to combine the various datasets in the most meaningful way and package the insights so that they resonate with farmers.331 Machine learning
will be an important tool for accelerating this process. As algorithms learn and improve, they can have increased relevance and power for specific enterprises and farmers.
However, not every organisation will have the financial and human resources to follow this path. The use of data – and
especially the more advanced technologies around data – requires specific skill sets and sufficient resources to invest. Many players today lack one or both of these. We expect that many D4Ag organisations will try to embrace the potential of data, but only a small percentage (though impossible to quantify) will be able to take advantage of it. Thus, in the coming years, we may also see some greater consolidation within the sector as data analytics leaders outcompete their slower-moving rivals.
Successful solutions will be those that can ‘crack the code’ on how best to use data.332 These solutions will be able to
integrate the many sources and types of data in a compelling way to best deliver value to the farmer. The data-informed output must be insightful, precise, simple to use, and – most importantly – truly address the pain points that farmers care about most. This ‘data revolution’
should lead to markedly better products for B2B and B2C users, as they will be specifically and precisely designed to meet these users’ needs.
This data-driven approach will push past some of the limitations of today’s solutions in order to target what people want. Data-informed solutions will be
designed around a deep understanding of their users’ behaviours and needs; as such, they should encourage higher uptake and create real impact for farmers. Eighty per cent of survey respondents indicated that they have tailored or plan to further tailor their products for smallholder farmers. Moreover, the ongoing collection of data and use of pattern- recognition and machine learning tools should enable D4Ag solution providers to recalibrate their solutions based on user results and the ability to diagnose what is and is not working.
This ‘data revolution’ will not only allow for improved user information and feedback loops but will also extend the offerings that solutions can provide smallholders. For example,
chatbots that share photos with farmers and voice-based solutions that allow farmers to hear advice rather than read it have begun to overcome the challenges of illiteracy and low connectivity. Additionally, data-driven solutions can provide smallholders with critical farm guidance with an unprecedented level
into these systems would be solutions skewed towards men, their outputs would reflect the same biases. These technologies also come with other important risks and concerns around data governance and consumer protection (including privacy and informed consent). In Chapter 5, we discuss how governments, donors and investors can ensure that these technologies are adopted in an effective and appropriate manner.
User design, experience, and understanding must also go hand in hand with such data-based insights. One
commonly cited benefit of data analytics is that it “can reduce the amount of direct input needed from the farmer”.336 But by distancing
themselves from farmers, solutions may more easily misrepresent their desires and needs. The balance between data and ground-level knowledge is an important one to strike and will be discussed more later.
Longer time horizons are the key to managing these and other risks. It is
critical that players take time to think through the consequences of the models and methods they design before implementation and follow up with rigorous evaluation and adjustment – even if doing so slows down the pace of of precision, localisation, and customisation.
Similarly, drone technology is being used to create highly accurate maps that can be used for mapping land boundaries with a range of possible uses, such as land titling and clarifying land ownership.333 These and other methods
should further bridge the gap between reach and impact.
The increased use of data in agriculture is not, however, without risks. To begin
with, many of the technologies in question (e.g., machine learning, data analytics) leverage similarities. In other words, they rigorously use data from one case to predict another. This reliance on commonality could present a challenge in a sector as massive and varied as agriculture.334 The agricultural
sector in Africa comprises nearly 70% of the workforce and differs widely from place to place in crop, climate, human context, farmer characteristics, etc.335
Moreover, when it comes to data analytics, and artificial intelligence especially, there is a danger of reinforcing existing biases. As one illustration, today’s solutions currently reach very few women or other marginalised groups. The algorithms in question are based on inputs of historical data. Since all inputs
transformation. Moreover, by its very nature the agriculture sector moves more slowly than the technology sector; tech players will need to practice patience and re-orient themselves toward a more long-term approach. Failing to do so will risk entrenching existing issues in the design of new solutions, creating new and unanticipated consequences, and veering away from an inclusive agricultural transformation.