CASE STUDY · GOOGLE PAY
Google Pay Split and Request
In Google Pay, the split and the request that follows it are the whole relationship. I made the split fun to do, and gave the request a life after it is sent.
ROLE
Self initiated. I owned research, flows, UI, interface copy and the prototype
YEAR
2026
SCOPE
5 flows, 40+ screens, one clickable prototype
A self initiated redesign of bill splitting in Google Pay. Amounts, names and photos are illustrative.
THE PROBLEM
The app opens only when money has to move
People aged 21 to 30 in India split dinners, trips and flat bills with the same three to twelve friends. For them, Google Pay opens only when money has to move. A nicer split screen improves that moment but never creates a second one, and retention needs a second moment.
Groups already exist, three taps deep behind a chat menu, and an active group never shows up on Home. The split is a total and a rule: evenly, by amounts, by shares or by percentages. No items, no who had what.
WHAT I FOUND
50 million
installs Tez drove with scratch cards, which faded once the rewards thinned
3 taps
how deep groups sit in Google Pay today, behind a chat menu
1 trigger
the only reason to come back is Send reminder, which the app itself calls a private reminder
Venmo showed the other side: people open it multiple times a week just to see what friends are doing, and the note on each split is the content. A Google Pay review from July 2026 named the gap from the user’s side: after a few days, older split payments become hard to find. The record exists and gets buried.
THE HARDEST CALL
Kill the reminder
It was the one return trigger the app had, and I removed it. A reminder is a nag, and it feels bad for both people. You cannot bolt fun onto nagging. So the request becomes a board both people see, and social visibility does the reminder’s job.
A split does not end at send. Give it a lifecycle.
THE SYSTEM
Five flows around one idea: a split that stays alive
Flow 1 is a small group, item by item. Flow 2 is a big group where twelve people claim in parallel. Flow 3 is a bill with no receipt. Flow 4 is the person who owes. Flow 5 is the whole notification loop. Every path ends in the same place: a reason to come back.

Green paths are the happy route, red paths are the fallback. Every box is a screen in the Figma file.
See the Full Flow
FLOW 1, END TO END
Making the split, then everything after it
The first four screens make the split fun to do. The next four are the lifecycle: the reasons to come back after money has moved.
01 · BILL IN
Scan or type
Scan the receipt and every dish comes in on its own, or type the total when there is no bill.

02 · WHO HAD WHAT
Everyone starts at zero
My first version was a checkbox list and it read like a hospital form. The reveal is the fun part: whose plate is heavy, who claimed nothing.

03 · DRAG A DISH ONTO A FACE
Totals move as you go
Drag each dish onto a person and the totals move as you assign. Claimed items stay tappable, so two people on the naan divide it automatically. Nobody gets greyed out of a dish they shared.

POP UP · SHARED DISH
Garlic naan × 6, four people
Tap any dish still on the table and this sheet opens. Everyone who ate some gets their pieces, and the money follows the pieces, ₹120 each here. Done stays locked until the sheet says “All 6 pieces accounted for”.


04 · EVERYTHING CLAIMED
Nothing left on the table
One tap to send. Split evenly stays one tap away on every screen, because fun you cannot opt out of is a tax.

POP UP · SEND REQUEST
One last look before it goes
Tap Send request and everyone’s line shows up with the dishes behind it, so nobody gets a number they have to question. Your own ₹890 stays with you, so the request is ₹2,910, not ₹3,800.


Open
The request becomes a board
It does not end at send. Everyone sees the same board: who paid, who has seen it, who has not.

Paid
Good news, not a receipt
When someone pays, it arrives as good news on the board everyone shares.

Settled
Once it settles, it becomes a memory
The total, how long it took, who paid first, and photos from the group. Photos earn their place here because the memory needs somewhere to live.

A month later
It comes back as a recap
A month later the group gets a recap of where they went and who picked up the most. Four states, four reasons to open the app, and none of them need money to move.

FLOW 2 · BIG GROUP
Twelve people. Fifty items.
Twelve people and fifty items break any assigning screen. Bulk assign was my first attempt and it still took fifty decisions. So the organiser stops assigning, and everyone claims their own, live, on their own phone. Claiming online also means it works after everyone has gone home, or for a trip or a grocery run.

POP UP · LEFTOVERS
Two items nobody claimed
While the group claims, the organiser can pick what is still unclaimed and tap “Give 2 selected to someone”. Choose who had the sangria and nachos and the sheet splits them equally, ₹655 each.


FLOW 4 · THE PERSON WHO OWES
One notification, the same board
The person who owes gets one notification and opens the same board, not a payment sheet that closes. They review their own items and pay from inside it, and the board is still there afterwards.
LOCK SCREEN · SMALL GROUP
“Your bit is ₹350”
Kabir sees his exact share and what it was for, the cold coffee and gulab jamun. One tap opens the same board everyone else sees.

LOCK SCREEN · BIG GROUP
“Tap what you had at Office Lunch”
Fifty items and twelve people, so nobody gets a number yet. Kabir is asked to claim his own, and the message promises it takes a minute.

LOCK SCREEN · NUDGE
“Kabir hasn’t opened the split”
Forty minutes later he gets a nudge, not a nag. If nobody taps their items, the leftovers split evenly, so the bill never waits on one person.


FLOW 3 · NO RECEIPT
Nothing to itemise
No bill, nothing to itemise. Type the total and split evenly, and the item by item option is not shown at all.

SURFACE WHAT EXISTS
The group was already there
Groups already existed three taps deep. They move above People on Home, and the group page leads with the balance and memories, with settings at the bottom. Nothing needed inventing, it needed surfacing.


PROTOTYPE
Click through every flow
Pick a flow and tap through it, from Home to the monthly recap.
DECISIONS
Every call, and what I turned down to make it
Solve the split, not the whole app
Give the request a lifecycle
Kill the reminder
Items onto faces, not a form
Tap to join, never grey out
At scale, everyone claims their own
Online claiming, not pass the phone
One board, same for everyone
Surface the group that already exists
Photos, because the memory needs somewhere to live
Every fun idea passes one test
Itemising is always optional
THE TEST EVERY FUN IDEA HAD TO PASS
If you removed it, would anyone open the app less next month? Confetti, stickers, scratch cards and reactions on requests all failed and were cut. Tez showed why: scratch cards drove 50 million installs and faded once the rewards thinned. Reward fun runs out. People fun does not.
ASSUMPTIONS
Amounts, names and photos are illustrative. Receipt scanning assumes the bill comes in as line items. Everyone in the split is assumed to be on the same app.
WHAT CHANGED
The same split, four more moments
Before
A split ends the moment it is sent, and the only way back is a reminder.
After
The same split returns four times, as a board, good news, a memory and a recap.
HOW I WOULD MEASURE IT, IN THE 21 TO 30 COHORT
1
Weekly opens that no money request triggered
2
Share of splits made item by item
3
Median time for a split to settle
WHAT I LEARNED
The fix was a longer life for an old feature
The biggest lesson: the fix for retention was not a new feature but a longer life for one people already use.
STILL OPEN
Nothing stops someone from removing all their own claims, so a real build needs a confirm step. If two people claim the same item and disagree, I have not designed that moment yet. The big group flow has no monthly recap yet. That is an omission, not a decision.
More work
Check out more work
Twelve · Plans in years, not months
ixigo Trains · The four hours nobody designs for
Basil · Product page set builder concept



