A contextual spending review system that helps users pause and reflect before high-risk financial decisions.
Routine payments should be fast. But some transactions — unusually large amounts, unfamiliar payees, late-night spending — carry higher risk of regret or fraud. SafeSpend is a design exploration into contextual spending review: a system that surfaces relevant financial context at the moment of decision, without blocking or moralizing.
Most fintech products are optimized for transaction completion, not decision quality. Two patterns illustrate this clearly.
A user wants to purchase a ₹78,000 smartphone. The checkout screen anchors their attention on one number: ₹3,249/month. Everything else — total repayment, processing fees, stacked debt — is either hidden or minimized.
Users frequently approve fraudulent payment requests because the interfaces feel familiar, approval flows are fast, and warnings are generic. Current systems verify authentication — but rarely evaluate behavioral risk. The interface looks legitimate, so the user assumes it is.
The research goal was to understand why users ignore financial warnings, how impulsive spending happens, what makes scam interfaces believable, and where financial regret originates.
Most participants evaluated affordability through monthly payment amount and short-term convenience. Very few spontaneously calculated total repayment, stacked debt impact, or projected future financial stress. The monthly frame is cognitively cheaper — and interfaces exploit this.
Alerts like "Are you sure?" were dismissed almost instantly. Users responded meaningfully only when systems explained why the action was risky, what impact it might create, and how confident the system was in its assessment. Context transforms a warning from noise into signal.
Impulsive spending was not random. It concentrated around late-night hours (10 PM – 1 AM), immediately after stressful events, during urgency sales, and in the days following salary credits. These patterns are detectable — and therefore interruptible.
SafeSpend evaluates transaction timing, spending patterns, merchant context, and affordability signals — then determines whether a review prompt is warranted, and at what intensity. Low-risk transactions proceed uninterrupted. The system intervenes only when multiple signals indicate a user may benefit from pausing to reconsider.
The system uses a combination of behavioral and transactional signals to decide whether a review prompt is warranted. No single signal triggers intervention — the logic requires multiple signals to align. Recommendations are probabilistic, not definitive. Users always retain full control.
A user attempts an ₹18,000 sneaker purchase at 12:14 AM — outside typical spending hours, and after already exceeding their monthly discretionary average. The system detects a combination of timing, amount, and budget signals, and surfaces a contextual review prompt rather than proceeding silently.
The review logic analyzes UPI handle patterns, transaction urgency signals, payment amount anomalies, and known scam pattern signatures. When multiple signals align, the interface explains the specific reasons — not just triggers a generic warning.
Two additional flows were sketched but kept out of the primary design focus: a full EMI repayment breakdown (surfacing total cost, not just monthly amount) and a recurring subscription review screen (flagging inactive or duplicate subscriptions). Both are coherent extensions of the same review logic — but including them as full features risks diluting the core use case. They're documented here as directions, not deliverables.
Any system that adds friction to payments carries real risks. These constraints shaped every design decision — and would need to be explicitly managed in a production implementation.
No risk detection system is error-free. Understanding what breaks — and how the system should respond — is as important as designing the happy path.
Real product work means navigating competing priorities. These tradeoffs shaped every design decision in the system.
Aggressive warnings reduce transaction completion, EMI adoption, and platform revenue.
Too many alerts erode user trust and train users to dismiss everything.
Overprotective systems feel paternalistic and manipulative, driving disengagement.
The project scope was deliberately narrowed to one primary problem: high-risk spending interruption and review. A concept exploring four features simultaneously risks credibility — so secondary flows were documented but excluded from the primary design focus.
Without deployment or user testing at scale, these represent intended design goals — the measurable outcomes a functional version of this system would need to achieve to be considered successful. They are conceptual KPIs, not validated results.
SafeSpend is a design exploration into contextual spending review — not a deployed product. The core question it explores: can a fintech product add friction selectively and helpfully, without damaging user trust or blocking legitimate payments?
The most important constraint discovered: the system's credibility depends entirely on how rarely it fires, and how clearly it explains itself when it does. An over-eager review system becomes noise. A well-calibrated one becomes a tool users trust.