Decision 01 · Insight 03
Hick's Law
Duplicates get two buttons. Skip, or add anyway. One tap, never a menu. A person with thirty seconds of patience will not open a menu.
The SprintMart
shared cart.
Fig.01 · The shared cart · join · live cart · split and pay
01 · How a flat of three orders today
Problem.
Requests trickle into WhatsApp. One person collects them from memory, orders alone and pays the whole bill, then spends two days chasing everyone for their ₹600. Every quick commerce app treats all of that as one shopper.
The cart, the payment, and the awareness of both live in three different places. I put them on one screen.
02 · 12 interviews, 60 survey responses
Research.
Six of us split into tracks for the sprint. Mine was household ordering: 12 interviews, a 60-person survey, and five real group order threads read line by line. The others covered store operations, delivery trust and the base build. Four findings shaped the feature.
Insight 01
61%
Orders are group decisions paid by one person. money must be visible to every member, before checkout.
Insight 02
80%
One "default orderer" compiles everyone's items from memory. everyone adds their own, so joining has to cost nothing, not even a sign-up.
Insight 03
2in 5
A duplicate lands weekly; 78% forget items. guard the moment of adding, with a two-button check, instead of cleaning up later.
Insight 04
₹600×3
Chasing money is social friction; being chased feels like surveillance. settlement has to be built in, using caps, quiet approvals and per-member payment.
“I don't mind ordering. I hate being the house bank.”Interview 07 · a household's default orderer
03 · Two people, one system
Flow.
Hover any step to see its actual screen
Organiser · control without chasing
01Creates the cart
02Picks the payment model: admin pays or each pays
03Shares the invite code
06Reviews the joint cart as items arrive
10Locks the cart & starts checkout
13Pays the total, or per-member splits go out
Contributor · freedom without being watched
04Joins with the code, no sign-up
05Sees the live cart, members and budget
07Adds items, marked common or personal
12Pays their own share via UPI
Done here. The organiser carries it to checkout
System · carries the friction
05aSyncs the cart live; logs every change
08Catches duplicates: skip, or add anyway
09Updates the household budget bar
11Opens a 10-minute payment window
13Confirms one order, tracked by everyone
04 · Four calls that made it work
Decisions.
Decision 02 · Insight 01
Public cap
The budget is public. One cap that every member can see, editable until the cart locks, so nobody is surprised at checkout.
Decision 03 · Insight 04
Quiet approval
Over-cap items wait, silently. The Organiser approves or removes; the contributor never sees a call-out.
Decision 04 · Insight 04
Split & pay
Everyone pays their own share inside a ten-minute window. If someone misses it, the cart does not die. The organiser covers that share and the groceries still arrive.
05 · What I claim and what I don't
Outcome.
Straight answer
This is a college project. There is no usage data and I am not going to pretend there is. What I will claim is a working multi-user feature that Blinkit, Zepto and Instamart do not ship. I checked all three myself.
If it shipped, I'd measure:
Duplicate-guard acceptance
Settlement time in the window
Orders with 2+ contributors
Repeat carts per household
01
Design and build should move together. Two screens got designed before they were wired in. I fixed it, and I noted it.
06 · Who did what
Credit.
My role
Household research track, Shared Cart feature design, design system tokens shipped as CSS variables used across the React build, and direction on the AI-agent build
Team
Five teammates ran the store-operations, delivery-trust and base-app build tracks
Tools
Figma, Stitch, React, Claude-based agentic build pipeline
Program
IIT Madras, UI/UX with Agentic AI and GenAI, 2026