Dr Rebecca Sarku

A story from Western Kenya

Two years ago, while doing fieldwork in Western Kenya, I sat with a woman who farms a small plot of maize. I had my notebook ready and was asking the usual questions: How much did you harvest? What inputs did you use? Who decides what gets planted?

Halfway through our talk, she pulled out a basic phone and mentioned, almost casually, that an agent from an agri-tech company had signed her up for an SMS advice service the season before. She had shared details like her farm size, planting dates, and expected harvest so the app could “calculate” what she should do next.

When I asked if the advice actually helped, she paused. Then she said something I still think about: she wasn’t sure who those numbers were really for. The company seemed to know more about her farm on a screen somewhere than she did, yet the tips never quite matched what she saw with her own eyes in the field. She wondered who truly benefited from all that data—her, or the people who owned the platform.

That moment stayed with me as I headed to the Development Studies Association’s 2026 summer conference in Dublin. There, I served as a discussant on a panel titled: “Tension? Competing visions for digital agriculture and rural development: Smallholder agency vs profitable business models at scale.” Across three sessions and nine papers, the panel wrestled with the very tension that farmer had named—without using any academic language at all.

Is digital agriculture giving smallholder farmers more power and choice? Or is it quietly concentrating that power in the hands of platforms, investors, and middlemen?

Three key messages from the panel

1. “Sustainability” and “impact” mean different things to different people

Several papers showed that what counts as a “good outcome” in digital agriculture depends on who gets to define it.

• In Kenya, researchers studying agri-tech start-ups found that “sustainability” was often treated as “can this business make money?” Environmental and social goals only mattered when they lined up with revenue.
• In Uruguay, newspaper stories and interviews with stakeholders painted an overwhelmingly positive picture: “digitalisation is inevitable.” There were almost no critical voices in public debate.

In both cases, the success story being told about digital agriculture was shaped more by companies and institutions building the tools than by farmers’ real-life experiences.

2. The real picture is messier than the hype—or the fear

The evidence on the ground is more limited, uneven, and ambiguous than either the techno-optimist or techno-critical stories suggest.

• In Ghana, studies of agricultural platforms found that most start-ups simply aren’t big enough yet to deliver on the grand promises of “big data” farming—or to trigger the dystopian fears of mass data extraction. Both narratives are, for now, ahead of reality.
• In Malawi, data showed that digital access did improve welfare, and smallholders actually gained proportionally more than large commercial farmers, with lower inequality among adopters. But access itself was uneven: education, phone ownership, and social networks still determined who could benefit.
• Another paper from Ghana on mechanisation platforms added nuance: women’s exclusion from digital matchmaking services wasn’t just about gender. Age, marital status, plot size, and location all played a part—problems a single digital “fix” can’t solve.

3. Agency depends on governance, design, and people—not just technology

Whether digital agriculture expands or limits farmer agency comes down to governance, institutional design, and human intermediaries—not the technology itself.

• In India, the Commons Connect project showed that when farmers contest and localise algorithmic advice through community-centred data systems, outcomes were more equitable.
• In Africa, ILRI’s work on climate-data digital public infrastructure warned that without deliberate inclusion, digitising livestock data can deepen existing divides.
• In Zambia, the same climate app produced very different results depending on group leadership and whether extension workers were present as intermediaries.
• In Kenya, research on M-Shamba highlighted a temporal dimension: digital tools can produce “scaling intelligence,” “mis-intelligence,” and “dis-intelligence” over time. Agency isn’t a fixed feature of a tool—it evolves based on how the tool is bundled with people, organisations, and communication practices.

Linking the messages to three connected questions on “farming by metrics”:

1. What do metrics actually measure, and can they capture the complexity of farming?
2. Do sustainability metrics actually lead to more sustainable outcomes?
3. What role do metrics play in transforming food systems?

On complexity (Question 1)

The Kenyan and Uruguayan cases show that when “sustainability” is defined by those who build the metric, often commercial actors, it collapses farming’s complexity into whatever is easy to track and monetise: input use, repayment rates, yield forecasts. The Ghana mechanisation study makes a similar point from another angle: a metric or algorithm built around a single axis (like gender) can’t capture the layered, intersectional reality of exclusion. Metrics measure what designers can see and sell not necessarily what matters to the farmer.

On change (Question 2)

Here, the evidence is genuinely mixed

• Malawi’s data suggest that digital access, as a metric, does correlate with real welfare gains and reduced inequality among users.
• But the Ghana platform-economy paper cautions that, at current scale, most platforms aren’t generating enough data flow to produce either the positive or negative changes their narratives promise. The “change” implied by adoption metrics may be more aspirational than actual.
• Zambia’s mixed results reinforce that whether a metric-driven intervention improves outcomes depends on complementary social infrastructure—group leadership, intermediaries—not the metric or tool alone.

On transformation (Question 3)

This is where the governance message lands hardest.

• Commons Connect and ILRI’s digital public infrastructure work suggest metrics and data can support food system transformation only when governance is participatory, when communities can contest, reframe, and own the numbers rather than simply supplying them.
• Where data governance remains extractive or opaque as the Kenyan agri-tech ethnography and the “opposing discourses” paper both suggest, metrics risk entrenching a narrow, market-centric vision of transformation rather than an inclusive one.

What is the implications for research on agricultural innovations and food systems

1. Treat metric design as political and relational—not just technical

Instead of asking only whether a digital tool “works,” researchers should investigate:

• Who defines the indicators embedded in agri-tech platforms?
• Whose knowledge is made visible—or invisible—in the process?
• How do commercial incentives shape what gets measured?

This calls for more co-design and participatory methods following the Commons Connect model where farmers help determine what “success” actually means in their own context, rather than having sustainability or productivity metrics imposed from outside.

2. Use longitudinal and intersectional research designs

The “scaling intelligence/mis-intelligence” framework and the Zambia and Ghana findings both suggest that a single snapshot metric like adoption or welfare at one point in time will miss how inclusion and exclusion unfold over time. It will also flatten differences among women, youth, and marginalised groups into a single average effect.

Future research should combine mixed methods, repeated measurement, and intersectional analysis, so we capture not just whether digital tools help “smallholders” as a category, but which smallholders, under what conditions, and for how long.

3. Prioritise comparative, governance-focused research

Since several papers converge on the finding that governance arrangements—not technology—determine whether digital agriculture expands or constrains agency, future work should systematically compare cases of participatory versus extractive data governance.

This would build a stronger evidence base on which institutional models actually translate metrics into durable, equitable transformation rather than treating “digital agriculture” as a single, undifferentiated intervention.

Coming back to the farmer in Western Kenya

That farmer’s instinct that the numbers being collected about her farm might not be “for” her, turns out to be a live, unresolved question running through an entire field of research.

It’s not a naïve misunderstanding. It’s an accurate reading of a system whose metrics, and whose benefits, are still being contested.

And until we centre farmers’ voices in how those metrics are designed, governed, and used, we risk building digital futures that serve platforms more than people.