I approached this by anchoring on the reality of how money actually moves in India: groups frequently mix cash and UPI, balances linger because nudges feel awkward, and people want clean closure without spreadsheets or back-and-forth screenshots. My goal was simple—make Splitwise the most frictionless way to close the loop in India, not just log it.
First, I clarified the outcome with the interviewer: improve the Indian user experience in the next few months, with success measured by settlement completion rate and time-to-settle. With that, I focused on the most active cohort—young professionals living with roommates and going out in groups—because they create the highest frequency of small, mixed-mode expenses.
From user journeys, three pain points stood out. One, cash is still common (rent splits, house help, cabs), and recording it is clunky. Two, settlements stall because asking friends to pay feels uncomfortable. Three, closing balances outside Splitwise (UPI/Paytm/GPay) breaks the loop—money moves, but the app doesn’t always reflect it cleanly.
To solve this, I’d start with four improvements.
1) Cash Tracking That Feels Native: when adding an expense or settling, let users tag “paid in cash” or “part cash, part UPI” in one flow. Auto-attach who handed cash to whom and reconcile the graph instantly—no manual notes.
2) UPI-First Settlements (Deep Links + Autoclose): generate one-tap UPI intents (GPay/PhonePe/Paytm) with memo and amount prefilled; once the user returns to Splitwise, show a lightweight confirmation (“Mark as settled?”) or read the matching UPI SMS to auto-close, with consent.
3) Gentle, Socially Aware Nudges: periodic, polite reminders that feel neutral (“Looks like your April dinner split is pending—settle when convenient”) with cadence controls, group summaries, and the option to nudge yourself privately.
4) Flexible Splits by Reality, Not Ideals: quick presets for unequal shares (designates, fixed amounts, percentages), and a “round-off smartly” toggle that minimizes tiny coins while keeping totals exact.
In prioritization, I’d ship UPI deep links + autoclose and native cash tracking first—they address over 70% of stalled settlements and reduce drop-offs at the moment that matters. Nudges and flexible splits follow; they boost completion and reduce social friction, but only after the monetary rails feel effortless.
Trade-offs are manageable. Cash support adds UI surface area—solve with contextual fields that only appear when needed. UPI autoclose must be opt-in and transparent; pair it with clear privacy copy and an activity log. Reminders risk notification fatigue, so default to low cadence and give users control.
How I’d measure success:
North Star: settlement completion rate within seven days of expense creation.
Supporting: median time-to-settle, % of settlements via UPI deep links, % of “cash recorded” events that fully reconcile the group graph, repeat usage for settlements per month.
Counter: increase in support tickets about confusion or privacy, reminder muting rates.
In short, I wouldn’t try to “reimagine” Splitwise for India—I’d finish the last mile. Make cash first-class, make UPI one-tap and self-closing, remove social awkwardness with considerate nudges, and respect the messy reality of unequal splits. If we do that well, balances won’t just be tracked—they’ll actually get settled.