Where they started

The programme was eleven years old and, until three years ago, had been growing every year. Then something changed. Enrolment stayed healthy. Attendance looked fine on paper. But the women who joined weren't coming back for the second year, and no one at headquarters could explain why.

The annual evaluation was a 42-question survey. It measured "satisfaction" and "perceived economic outcome." It didn't ask what was hard. It didn't ask what was missing. It didn't leave room for someone to say the sessions ran at the wrong time of day.

What we did

  • Spent five months on the ground in six of the collectives the programme served — three districts, two states. Not doing formal academic research. Doing the kind of listening a good consultant does when the client will pay them to actually show up: drinking chai, running long conversations, asking questions that would have been embarrassing to write on a form.
  • Ran 84 semi-structured conversations and 6 group sessions. Mapped what members did on a normal day, where the programme fit into it, and where it didn't.
  • Facilitated co-design workshops with members from each collective. Not asking for validation — asking members to redesign the programme themselves. What to keep. What to drop. What was missing.
  • Helped the team build a simpler reporting framework — one that could carry the qualitative signal the survey had been erasing. Members chose the indicators. Not us. Not funders.

What we heard

  • The sessions ran at 2pm. Members had to be home to feed their children by 3. This one finding accounted for most of the drop-off. Nobody at HQ had ever asked when the sessions ran.
  • The credit product had been designed by someone who had never taken a loan of that size. Members had quietly rewritten the repayment structure in two of the six collectives; their version worked better than the official one.
  • The literacy component was patronising. Members wanted a phone-first version. HQ had a five-year contract with a printing vendor and hadn't asked.

What it moved

By end of engagement: 84 community conversations completed. 2× the previous uptake metric in the pilot cohort after redesign. 6 communities consulted deeply, not surveyed. 1 flagship programme redesigned — by its members, written up by us.
"They didn't just hand us a report. They handed us a programme our members helped write."

What we'd do differently

We tried to keep our team consistent across all six communities and it stretched the schedule by a month. In hindsight, we should have run pairs of listeners simultaneously in different districts. Consistency of voice matters less than getting there while the door is still open.