Projects

This page collects selected applied research projects on technology for underserved users in financial services, mobile, and health. Each was funded by an organization invested in those communities — the Gates Foundation, Women's World Banking, Google, My Oral Village, and UNICEF — and each was built around qualitative fieldwork, mostly in rural and peri-urban Pakistan. Two case studies are written up in detail below. Another one on My Oral Village on numeracy and mobile wallets will be added soon.

Digital ROSCA: Human-Centered Design for Informal Savings Groups

A study to digitize informal rotating savings groups for low-literacy Pakistani users, working from 80 user interviews through prototype testing.

Methods:

Semi-structured interviews, iterative design testing, two-phase usability testing

The challenge

In Pakistan, 36% of people save money but only 4% save through formal institutions. A third save through informal Kamaitis — rotating savings and credit associations where members pool fixed contributions and take turns receiving the pot. Kamaitis are deeply embedded in everyday financial life, especially for women in low-income communities. The question we were asked was whether digitizing this familiar behavior could be a path for introducing the unbanked to digital financial services.

Approach

We followed a three-phase Human-Centered Design process. In the inspiration phase, we conducted 80 semi-structured interviews with ROSCA organizers, members, and non-members, varied across gender, locality, and group size. Two-thirds of respondents were women. Interviews ran 60 to 75 minutes in the participants' preferred language.

In the ideation phase, we mapped findings to a digital ROSCA flow and made the modality decision: smartphone over USSD, given the complexity of ROSCA management and the smartphone ownership rate of 66% among organizers. We designed application icons through iterative in-context and out-of-context testing, redesigning them based on user interpretation failures rather than designer assumption.

In the implementation phase, we ran two-phase usability testing with two complete ROSCA groups. A facilitated first phase, followed by unsupervised completion of a full ROSCA cycle.

Key finding: the organizer is the design pivot

The most consequential finding reframed how we approached the entire application. Conventional wisdom assumed ROSCA trust is distributed across the group. Our interviews showed something different. Members usually do not know each other. Trust resides between each member and the organizer, and the organizer absorbs the administrative work of recruitment, record-keeping, collection, and disbursement. This made the organizer the right entry point for the application, the most likely adopter of digital tools, and the design center of gravity.

Other findings shaped specific features. Women organizers face acute mobility constraints during pot collection and often rely on male intermediaries, which compromises group privacy. Women keep records more consistently but face literacy challenges. Lucky draws and need-based assignment are the two dominant turn allocation methods. Cash remains the preferred payment medium even where mobile money is technically available, because users' familiarity with mobile money is limited to brand recognition.

Design recommendations

We built the application around the organizer rather than the group. The interface is bilingual in English and Urdu with audio help throughout, designed for low-literacy users. Icons went through multiple rounds of redesign against user interpretation, with three icons substantially changed after testing failures. The application supports both cash and mobile wallet payments to reduce adoption friction. Turn exchange, notifications, and record automation match the way organizers actually manage groups today.

Outcome

A working bilingual prototype was built and tested. The work was published with Hamid Mehmood as co-first authors at CSCW in 2019, with co-authors from ITU, the University of Washington. The full public report was released jointly with the Gates Foundation and Karandaaz is available here.

Guddi Baji: Evaluating a Women-Agent Mobile Money Program in Rural Pakistan

A stakeholder evaluation of JazzCash's women-agent program in rural Punjab, surfacing structural selection and training failures that were undermining its gender goals.

Methods:

Stakeholder interviews, activity-based segmentation, field observation

The challenge

Mobile financial services in Pakistan are heavily skewed toward male agents in male-dominated public spaces, which limits women customers' access. JazzCash's Guddi Baji program (the name means "good elder sister") aimed to onboard women as branchless banking agents in rural areas, expanding both financial inclusion for women customers and economic opportunity for the agents themselves. Women's World Banking funded a pilot evaluation in Kasur and Sialkot. The question was whether the program was working, and if not, why.

Approach

We conducted 45 stakeholder interviews across multiple roles within the pilot. Women agents segmented by activity level (low, mid, high). The male counterparts who often facilitated transactions on their behalf. Account holders, OTC users, and customers. BRC field supervisors. JazzCash field staff. This stakeholder breadth was deliberate. A pilot evaluation that focused only on agents would have missed structural problems in the surrounding ecosystem.

Key finding: a selection-criteria failure

The most consequential finding was structural. The pilot was recruiting women as agents based on availability rather than on the realistic preconditions for being able to transact. Many recruited agents did not personally own phones, lacked the basic literacy needed to operate the agent application, or had no viable local customer base for branchless banking services. The program was creating "agent" labels without the conditions women needed to actually do the work. When we segmented agents by activity level, the high-activity agents shared a set of pre-existing conditions, including phone ownership, literacy, social standing, and an existing community network. The low-activity agents did not. The selection process had not screened for any of these.

Related findings

In a meaningful share of cases, the named woman agent was acting as a front for transactions that were actually being run by a male relative. This raised questions about both the program's gender goals and the integrity of the activity data the pilot was generating.

Field operations on the JazzCash side were predominantly male, which limited their ability to support women agents on training, privacy-sensitive coaching, and culturally appropriate troubleshooting. There was no female field-staff layer to mirror the agent layer.

Training and marketing materials were generic. They did not engage with the specific cultural and social conditions women face in becoming visible financial intermediaries in conservative rural communities.

Recommendations

We delivered four recommendations to Women's World Banking and JazzCash. Rebuild selection criteria around realistic preconditions for women's agent activity, with phone ownership, literacy, and customer base as gating factors rather than nice-to-haves. Introduce a female field-staff layer for training, in-home support, and check-ins. Develop training and marketing materials that engage with the gendered visibility and social acceptance issues women agents actually face. Account for and reduce male counterpart involvement so the program's gender outcomes match its gender claims.

Coming soon

Page updated

Google Sites

Report abuse