A digital platform where credit card customers could visualize their spending patterns, raise transaction disputes, and receive dynamic credit-limit suggestions based on their salary. Simultaneously, it gave banks a dashboard to analyze customer spending behavior, identify cross-sell opportunities, and suggest credit-limit adjustments.

Customers had no clear visibility into where their money was going across multiple cards, and raising disputes was a slow, opaque process. Banks, on the other hand, had raw transaction data but no easy way to turn it into actionable insights for customer retention or targeted product offerings.

A full-stack system with customer-facing spending visualizations, a dispute-raising workflow, and a bank-side analytics dashboard for segmentation, spending-pattern analysis, and automated credit-limit recommendations tied to income data.

I built a customer-retention tool for an organization that was structurally designed to acquire merchants. Credit card marketing was merged with POS marketing, so the team’s KPIs and budget were tied to merchant sign-ups and transaction volume — not customer lifecycle, cross-sell, or credit-limit optimization. Worse, the bank’s actual operating policy was force-selling: pushing credit products onto the same merchants who already held POS terminals or loans, using the fake promise of a "better relationship" rather than acquiring quality customers backed by data and sustained behavior change. The bank had no dedicated team with the authority, budget, or bonus structure to own customer analytics. The data my platform generated was valuable, but there was no organizational receptor to absorb it — because the institution wasn’t trying to understand customer behavior; it was trying to squeeze existing relationships. Without a champion whose performance review depended on the output, the product became a cost with no clear owner.

A real-time reconciliation system that processed SWIFT transactions and ATM general ledger entries end-to-end — automatically identifying ambiguous transactions, executing reversals, and flagging errors with contextual routing for operator review.

Banks were running legacy reconciliation platforms that ingested massive transaction volumes but still choked on edge cases, leaving teams to manually resolve ambiguities, perform GL reversals, and chase down unmatched entries. The result was operational drag, settlement delays, and a quiet accumulation of financial risk hidden inside spreadsheets.

A reconciliation engine designed for the full lifecycle — from SWIFT message ingestion and ATM transaction matching through to GL posting. It could spot ambiguous transactions in real time, trigger automated reversals where logic was clear, and surface only the exceptions that genuinely needed human eyes, complete with contextual flags to make those decisions fast.

After multiple rounds of negotiation, the commercial reality became impossible to ignore. The banks wanted enterprise-grade reliability — a system that would handle their most sensitive financial flows without sleep — but they wanted it priced like a one-time software purchase. They saw the invoice as the cost. They didn't see the support calls at 2 AM when a SWIFT message format changed. They didn't see the six-month knowledge transfer. They didn't see that reconciliation systems don't just process transactions; they process exceptions, and exceptions require judgment, context, and sustained partnership. Then came the familiar promise: "Keep the price low on this one, and the next project will be properly scoped." I had learned that lesson the hard way — a discounted first project doesn't buy loyalty; it buys a reputation for being cheap. It trains the client to expect subsidies and traps your team in an underfunded maintenance cycle that consumes the very time you need to pursue better opportunities. I walked away because saying yes would have meant becoming their outsourced operations team at a software vendor's rate, and closing the door on every project that might have actually mattered.