GHANTI rings only when it matters.

"Where is the bus?" was never the parent's question — it's the sensor's answer. The parent's real questions are about the child, the day, and the walk to the stop. India runs on shared drop points, not doorstep drops: timing that walk is the highest-frequency job, and none of the audited products served it directly.

Primary research was autoethnographic — I logged my own daily use of a live tracking app as structured pain points. Then a desk audit: nine competitor products, every capability claim verified against vendor pages and store listings, and — more importantly — against documented field behavior. The gap between the two is the insight.

I audited nine products against the capabilities that matter to a parent — every cell verified from the vendor's own pages, official store listings, or documented user reviews. Proximity alerts exist (fixed radius). ETA exists (bus-level). RFID boarding exists (hardware add-ons). Nobody wires them into a pipeline gated on the child. The green rows are GHANTI's territory — no competitor documents a single one.

Both personas are built from my own household and the parents at a shared drop point (small n — treated as proto-personas, not validated archetypes) — composites, not fabrications, and flagged as such in the assumption register. The child is deliberately not a persona: Aarav is the subject of every signal, but the phone in hand belongs to a parent.

Each one is a direct answer to a research insight — and each one costs something. A principle that costs nothing decides nothing.

GHANTI inverts the category's architecture. Competitors broadcast vehicle state and let parents filter; GHANTI filters first and treats silence as a designed state. Different signals fire on different rules — not every notification passes through the same conditions. Three inputs govern what each signal knows: context, timing, and presence. One honesty layer sits above all of them.

The IA is deliberately shallow — a parent mid-walk gets one thumb and three seconds. Everything orbits Today, where the live map lives; the other three tabs exist so Today can stay almost empty.

Mapped from my own mornings before designing a single screen. The old-world emotional line dips at the same two points every day: the guess (when do I leave?) and the gap (did he actually board?). GHANTI exists to flatten exactly those two dips.

Two flows define GHANTI. The morning flow shows how context, timing and presence govern different moments; the exception flow shows why presence had to come first in the architecture — without it, the "not aboard" alert is undetectable.

Three layouts for Today, evaluated against one scenario: it's 7:38, the parent has one thumb free and three seconds. The map — every competitor's hero — lost immediately. A map is a question; the parent needed an answer.

A parent walking to a stop on an Indian street is holding a bag, a child's hand, or both — screen-first design fails them. So the signal layer is designed for ears and skin first, screen second: each signal type gets its own tone and haptic pattern, and the walk tone encodes proximity in its tempo. You learn the language in a week and stop looking at the phone.

Thirteen screens across four flows, assembled from a component library built on themeable variables. What follows is the final prototype — assembled from a component library, not drawn screen by screen.

The screens run as a prototype, not a gallery: sheet drags smart-animate between states, screens push, tabs dissolve, and loading advances on its own timer.