Mastercard Farmers Network. why a working payments platform still wasn't being adopted

Define & Plan

The Mastercard Farmers Network (MFN) — branded e-Kilimo in Tanzania and e-Rythu in India — is a digital platform meant to let buyers, aggregators, and farmers register, transact, and get paid for produce without relying on cash and middlemen. By the time Mastercard Labs brought in Digital Disruptions, MFN had already been built and piloted for roughly two years. The question on the table wasn't "what should we build" — it was a harder one: why isn't it being used, and what should we fix first?

This made the engagement an evaluative research project rather than a generative one. We weren't designing a product from a blank page; we were diagnosing a live deployment, separating problems that came from strategy and implementation (within Mastercard's control) from problems rooted in the surrounding environment (largely outside it).

Three goals shaped how every insight was framed: increase confidence by validating or invalidating existing assumptions, add nuance by explaining why, and cut through noise with a coherent storyline rather than a long list of disconnected observations.

Research Approach

The wider engagement was structured as two phases — Design the Right Thing, then Designing Things Right — moving from secondary research and user interviews through synthesis, ideation, prototyping, and in-market testing, with a target of practical, user-tested solutions within roughly 14 weeks.

FIG. 01 — Two-phase framework: Phase 1 (Research) moved from kick-off to key hypotheses in ~6 weeks

This case study covers Phase 1 — the deep-dive research and synthesis that produced the insights and problem statements below. The work combined three lenses: light transactional data analysis, qualitative fieldwork, and the team's accumulated experience diagnosing low adoption across other digital financial services deployments.

Fieldwork & Data Collection

Fieldwork ran across two very different markets, chosen because MFN was operating two distinct go-to-market models within them: a commercially-driven trader aggregator in Tanzania, and a community-driven farmer-producer organization (FPO) model in India.

FIG. 02 — Fieldwork sites: Mbeya Region, Tanzania, and Araku Valley, Andhra Pradesh, India

31 individual, in-depth contextual interviews were conducted by two teams (each with a translator) over nine days. The sample deliberately spanned the full value chain rather than just end-users, since adoption in this kind of network depends on every link — farmer, agent, aggregator, and buyer — pulling in the same direction.

FIG. 03 — Interview coverage: aggregator admins, trading agents, lead farmers, and farmers in Tanzania (left); buyers, FPO CEOs, procurement agents, lead farmers, and farmers in India (right)

FIG. 04 — Contextual interviews were conducted in homes, collection points, and farms — not in offices

Secondary & Market Context

Before drawing conclusions from interviews, the team grounded itself in the surrounding agricultural and digital-financial-services context in both countries — agriculture employs roughly 80% of Tanzania's labor force and is the livelihood for about 58% of India's population, but smallholders in both markets face long-standing problems with fragmented price information, weak market linkages, and poor post-harvest infrastructure.

FIG. 05 — Secondary research: 67% of Tanzanian farmers cite low open-market prices as their top complaint (CGAP–McKinsey, 2015)

This context mattered for interpreting the fieldwork: MFN's core value proposition — better prices and access to more buyers through digital transparency — was aimed squarely at a real, well-documented problem. That made it more important to understand precisely why uptake was still low, rather than to question the underlying premise of the product.

FIG. 06 — The two MFN models: a commercial trader aggregator in Tanzania, an FPO/community aggregator in India

Data Analysis

A light transactional database review ran in parallel to interviews, and it set the tone for everything that followed: the volume of real activity on MFN was small, concentrated, and seasonal. Tanzania accounted for the vast majority of recorded collections, but in practice only one aggregator — PCT — recorded any meaningful volume, with sharp spikes tied to crop seasonality rather than steady usage.

This database signal — low, concentrated, seasonal usage — directly shaped where we focused fieldwork: not on farmers in the abstract, but on the specific agents and aggregators who sat at the bottleneck between a working back-end system and the people who were supposed to use it daily.

Synthesis: Key Insights

Interview data was clustered into themes, then tested against the database signal and secondary research to arrive at insights with three layers of evidence behind them. Seven insights emerged; the four below carried the most weight for the eventual problem statements.

Taken together, the insights pointed to a value proposition that resonated conceptually but broke down in delivery — particularly for agents, the intermediaries every farmer's experience of MFN actually depended on. Agents had low literacy and technology comfort, no compelling incentive to use the app, and a "status quo bias" that meant every UI change cost real retraining effort rather than being received as an improvement.

Channel partner involvement also surfaced as a double-edged signal: in India, Tanager's years of pre-existing trust with FPOs transferred directly onto MFN's reputation; in Tanzania, NMB's intermittent presence and conflation of MFN with broader financial-inclusion messaging confused farmers and, in at least one case, created false expectations about access to loans.

Personas

Mastercard already had baseline personas from MFN's initial, pre-launch research. Rather than duplicate that work, we built five additional personas with deeper context — three covering the agent intermediary role that the data analysis had flagged as the critical bottleneck, and two covering buyer businesses at different scales.

FIG. 07 — Agent personas (Buyer Agent, Lead Farmer, Procurement Agent) and Buyer Business personas (Large, Small)

The distinction mattered for design: a Buyer Agent is effectively his "own boss" and needs to be persuaded; a Procurement Agent will follow instructions from an employer with much higher ability to comply, but still needs infrastructure and motivation addressed independently.

Journey Mapping

A consolidated agent journey traced every touchpoint with MFN — onboarding, training, farmer registration, inventory updates, the sell-purchase transaction itself, and support — capturing actions, challenges, mindset, and a simple plan-vs-reality assessment at each step.

FIG. 08 — Agent journey (1st half): onboarding through the start of the sell-purchase transaction

FIG. 09 — Agent journey (2nd half): the highest-friction moment was recording the transaction on the smartphone

The journey made one thing vivid that interview quotes alone couldn't: the "Record of Transaction" step accumulated nearly every category of problem at once — technical (pairing with the digital scale), usability (error messages agents didn't understand), and environmental (poor network) — which is precisely the moment most likely to cause drop-out.

Problem Prioritization

To separate what Mastercard could realistically influence from what it couldn't, every issue was mapped against two axes: the four classic adoption levers (value proposition, pricing & incentives, user experience, go-to-market) crossed against whether the root cause was internal (within the team's control) or environmental (requiring a workaround or different target segment).

FIG. 10 — Internal problems (value proposition, pricing, UX, GTM) sat above environmental constraints (literacy, power, network, mobile money infrastructure)

This framing did real work in the room: it let the team say plainly that poor rural network coverage wasn't something a redesign could fix, while a confusing onboarding flow or an unincentivized agent absolutely was — and therefore deserved first claim on the next phase's design effort.

Problem Statements

Every stakeholder group — buyers, aggregators, agents, and farmers — was scored across the same four levers using a simple traffic-light system, making it possible to see at a glance where the most severe, addressable problems clustered.

FIG. 11 — Agents scored "many issues observed" across value proposition, UX, and go-to-market — the clearest signal in the research

From that comparison, three candidate problem statements were proposed — one per primary user group — for the project team to select two from, ahead of the December ideation session:

FIG. 12 — Final recommendation: the project team selects 2 of these 3 problem statements to carry into ideation

Outcome & Next Steps

The research reframed the conversation inside Mastercard Labs from "the app isn't being adopted" to a specific, evidence-backed account of where and why — concentrated almost entirely in the agent layer, compounded by inconsistent channel-partner support, and only partly explained by environmental constraints the team couldn't control.

Just as importantly, the evaluative framework drew a clear line between problems worth solving in the next design phase and constraints that would require a workaround or a different target segment entirely — preventing the team from spending the next 6-week design sprint chasing root causes outside their control.