The problem

Ward-level census and household surveys in Nepal were, and largely still are , paper-driven. A field worker walks a route with a clipboard, fills out long Nepali-language forms by hand, then returns to the ward office where staff re-enter everything into a spreadsheet. The result is predictable: lost forms, illegible handwriting, double entry, no real-time visibility of who'd been surveyed where, and an aggregate dataset that took weeks to compile and could never quite be trusted.

Ward 10 wanted to skip that whole pipeline. Capture the same data digitally at the doorstep, tag it to a household with GPS, sync it centrally, and use the same interface every survey worker already understood from paper, in Nepali, with the same field names and the same flow.

Designing inside three hard constraints

This wasn't a greenfield product where I could invent terminology and flows. Three constraints shaped every design decision.

Two forms, one household

Every household visit produced two linked records: a family-level form (पारिबारिक प्रश्न) capturing dwelling and household-wide data, and one or more individual-level forms (व्यक्तिगत खण्ड), one per member. Splitting the data model into two forms, with a clear card-based picker on the home screen , let workers move between them without losing context, and matched the paper survey's existing two-section structure exactly.

Family form: dwelling, ownership, livelihood

The family form captures who lives at an address and how. Across two pages, workers log form number, respondent name, head of household, house and street numbers, toll (locality), phone, migration history (if the family moved from another district), primary and secondary occupations, and dwelling-level details like access road type (motor road / bike-only / walk-only) and ancestral-home status. Each section sits in its own card, so a worker can scroll, edit, and review without losing place.

Individual form: per-member demographics

Each member of the household gets their own record. Name, relation to head of household, gender, parents' names, age, marital status, blood group, mobile, email. An Add new member action lets a worker capture an entire household in one sitting; the form preserves shared context (household name, head of family) across members so workers aren't re-typing the same fields five times.

Routing, coverage, and proof-of-visit

Each worker is assigned a location for their day's route. The attendance screen pins it on a map (heritage landmarks like Patan Durbar Square and Manga Hiti show up, this is real Kathmandu Valley geography), captures whether the owner was available, and lets the worker upload a photo of the house. That photo plus the GPS pin is the audit trail: ward office knows the visit actually happened, where, and to which dwelling.

Close the loop

Every form submission ends with the same calm success screen. Workers see a single clear confirmation, then a Continue action that takes them straight back into their route. No celebration, no extra clicks, this is government data work, and the fastest possible turnaround per household is the actual feature.

Outcome

The app shipped and was used by Ward 10 field workers to survey households in Jwagal and Kopundole, a ward of roughly 1,729 households and 6,554 residents. The design replaced the previous paper-and-spreadsheet pipeline with single-pass digital capture, GPS-tagged proof of visit, and central sync.

The ward's public web presence has since been rebuilt at ward10.lmc.gov.np, which now points residents at external government portals for forms and incident reporting. The app I designed sits behind that change, in the parts of the ward's operations that residents don't see.

What I'd take into the next project

• Sit next to the user before you draw. Designing for workers I could talk to in person changed which fields belonged on the same screen, where validation could relax, and how aggressive the bilingual labelling needed to be. None of that came out of a brief.
• Match existing mental models, then improve. The temptation in a digital-replaces-paper project is to redesign the whole form. Keeping the field order, labels, and Nepali wording was the unglamorous decision that made adoption real.
• Government work is design work. Civic and public-sector tools rarely show up in portfolios because they're bilingual, regulated, and built to a brief. They're also where design has the most direct effect on people's lives, I'd take more of this kind of work.