Coast FinTech
Rebuilt their reconciliation engine. Monthly close: 4 days → 4 hours.
Coast FinTech processes payments for mid-market B2B companies. At month end, their finance team spent 4 days matching transactions across 6 payment rails, 3 banking APIs, and a legacy ERP system. They had tried two vendors. Nine months with another agency had produced nothing shippable.
Nine months. Nothing shipped.
Ravi had signed with a larger agency the year before. After nine months of meetings, design sprints, and status updates, the agency had produced a Figma file and an architecture document. No code. Coast's finance team was still doing month-end manually. The tolerance for another long engagement was zero — Ravi needed a working system, not another proposal.
Working system in 11 weeks. No theatrics.
We started with the hardest part: the data. Six payment rails (Razorpay, PayU, HDFC, ICICI, Stripe, UPI aggregator) each had different API schemas, different settlement timing, and different edge cases for refunds and chargebacks. We built a normalisation layer that ingests from all six, maps to a canonical transaction schema, and flags exceptions for human review. The matching engine uses a rule-based system for 94% of transactions and a ML-assisted classifier for the long tail.
- →Transaction normalisation layer across 6 payment rails
- →Rule-based matching engine for standard reconciliation (94% of volume)
- →ML-assisted classifier for exception handling (remaining 6%)
- →Legacy ERP integration via custom middleware (read + write)
- →Exception queue: only unmatched transactions reach a human
- →Audit trail: every match decision is logged with reasoning
- →Finance team dashboard with drill-down from summary → transaction level
- →Automated month-end report generation
Month-end is now a Tuesday afternoon.
The first live month-end ran in 4 hours and 12 minutes. The finance team had blocked out the usual four days. They finished before lunch on day one. Ravi told us the second month-end, two members of the finance team went on leave during what used to be the most stressful week of their month. The system processed ₹840 crore in transactions in its first quarter with a match rate of 99.7%.
- →Monthly close: 4 days → 4 hours (96% reduction)
- →Transaction match rate: 99.7% automated in first quarter
- →₹840 crore processed in first 3 months
- →Exception rate: < 0.3% of transactions require human review
- →Zero reconciliation errors in first 90 days of production
What they walked away with.
- →Reconciliation engine (Node.js + Python)
- →Transaction normalisation layer (6 payment rails)
- →Finance team dashboard (Next.js)
- →Legacy ERP middleware integration
- →ML exception classifier
- →Automated month-end report system
- →GitHub repo (theirs), full documentation
- →Notion handover + system architecture
- →30-day post-launch support
The tools we used.
We spent 9 months with another agency and got nothing. Microsive shipped a working product in 11 weeks.
Ravi V.
CEO, Coast FinTech