How Webskyne Transformed Finova's Digital Banking Platform: A 340% Increase in User Engagement

## Overview In early 2024, Finova—a digital-first bank serving 180,000 retail customers across Southeast Asia—was at a critical inflection point. Their mobile app, once a competitive advantage, had become a liability. Daily active users (DAU) had declined 28% year-over-year, session duration had dropped below 3 minutes, and app store reviews were trending sharply negative. Internal metrics showed that 62% of users abandoned onboarding within the first 30 seconds, and support ticket volume related to "can't find feature" complaints had tripled in twelve months. Finova's CTO approached Webskyne with a clear mandate: rebuild the digital banking experience to match the standards users had come to expect from leading fintech platforms, without disrupting the financial operations that millions depended on. ## The Challenge The problems were both visible and systemic. On the surface, the app's interface felt dated, with confusing navigation and inconsistent design patterns. But beneath the UI, deeper architectural issues were constraining the product team's ability to iterate. The monolithic backend API had grown organically over seven years, embedding business logic into tangled endpoint chains. Adding a new feature—say, a spending insights dashboard—required cross-team coordination across four squads and a deployment cycle that averaged eleven weeks. Customer complaints revealed a pattern. Users couldn't find bill pay. They struggled to categorize expenses manually. Push notifications arrived hours late or not at all. The biometric login flow failed on 15% of attempts, forcing users back to SMS OTPs. Each of these issues incrementally eroded trust. The constraint that made this project different from a standard redesign was the zero-downtime requirement. Finova operated under a financial regulatory mandate that prohibited service interruptions exceeding 15 minutes per quarter. Any migration or deployment strategy had to honor that boundary. ## Goals We established five measurable goals at project kickoff: 1. **Increase daily active users by 200%** within 12 months of launch. 2. **Reduce onboarding dropout from 62% to under 20%**. 3. **Achieve a 4.5-plus star average app rating** within 9 months. 4. **Cut feature deployment time from 11 weeks to under 2 weeks**. 5. **Maintain zero unplanned downtime** throughout the transformation. Every design decision, technology choice, and sprint plan was traced back to these north-star metrics. When disagreements arose, the goals served as the tiebreaker. ## Approach Webskyne assembled a cross-functional pod: two UX designers, three backend engineers, one iOS developer, one Android developer, one QA automation specialist, and a dedicated product manager embedded with Finova's team. We operated in two-week sprints, with a hard rule that no sprint would modify both the UI and the backend data layer simultaneously. Our methodology centered on **strangler fig migration**—a pattern in which new services gradually wrap around legacy ones, extracting functionality endpoint by endpoint until the old monolith can be retired. This allowed us to improve the system incrementally while keeping the existing platform fully operational. We also adopted **observability-first engineering**. Before writing a single line of new code, we instrumented the entire legacy API with distributed tracing, structured logging, and custom business metrics. This baseline gave us the ability to detect regressions within minutes rather than days.  ## Implementation ### Phase 1: Diagnostic and Design System (Months 1-3) The first phase focused on understanding exactly where users were leaving and why. We deployed session replay on 10% of traffic, analyzed 4,200 customer support tickets, and conducted 28 user interviews spanning power users, recently churned customers, and prospects who had downloaded the app but never completed onboarding. From this research, we built a new design system—**Finova Design System (FDS)**—containing 68 reusable components, 24 interaction patterns, and a documented accessibility standard achieving WCAG 2.1 AA compliance. The design system was not merely a style guide; it included code implementations for React Native, pushing consistency to the implementation layer. ### Phase 2: API Decomposition and Service Extraction (Months 4-9) Working from the strangler fig strategy, our backend engineers decomposed the monolith into six bounded contexts: **Account & Identity**, **Transactions**, **Notifications**, **Analytics**, **Bill Pay**, and **Support**. Each context became an independently deployable service with its own repository, CI/CD pipeline, and schema. The highest-risk extraction was the authentication service. We ran a parallel shadow deployment for six weeks, mirroring real authentication traffic to the new service without switching production traffic. When error rates between the old and new path showed less than 0.01% divergence, we performed a gradual traffic shift—1% on day one, 25% by week two, 100% by week four. ### Phase 3: Frontend Rewrite and Onboarding Redesign (Months 10-14) The mobile team rebuilt the app using React Native, adopting a modular navigation structure that allowed the team to A/B test new flows without shipping full app updates. The onboarding sequence was completely reimagined: from a 12-screen linear flow to a progressive disclosure model that presented only three initial choices, reducing perceived complexity. We introduced an interactive balance preview immediately after signup—a small visual showing a sample account balance and transaction timeline—so users understood the app's value before being asked for KYC documents. This single change reduced onboarding dropout by an estimated 11 percentage points. ### Phase 4: Real-Time Infrastructure and Intelligence Layer (Months 15-18) The final phase introduced a real-time notification engine using WebSockets and a background processing layer that reduced notification latency from an average of 4.2 hours to under 90 seconds. We also shipped an AI-assisted spending categorizer that learned from user corrections, achieving 94% categorization accuracy by month 18 without requiring manual rule maintenance. ## Results The transformed Finova app launched to 100% of users on a Tuesday morning, with the rollout monitored live by both Webskyne and Finova engineering teams. The first-hour metrics were promising: crash-free users held at 99.94%, and API latency remained within the baseline. Within 30 days, DAU had risen 140%. By month 9, it had crossed the 200% target, reaching 340% above pre-launch levels. Session duration, previously under 3 minutes, grew to an average of 11 minutes. Users were not just opening the app more—they were staying inside it longer and exploring features they had previously been unaware of. Customer satisfaction, measured through in-app NPS surveys, rose from a baseline of 32 to 71. App store ratings climbed from 3.8 to 4.8 within seven months, with the volume of monthly reviews increasing by 220%—indicating that more users were actively engaged enough to leave feedback. The deployment velocity metric—feature lead time from commit to production—shrank from 11 weeks to 9 days, well exceeding the original 2-week target once teams became familiar with the new CI/CD pipelines. ## Metrics | Metric | Before | After | Change | |---|---|---|---| | Daily Active Users | 42,000 | 185,000 | +340% | | Onboarding Dropout Rate | 62% | 17% | -45 pts | | App Store Rating | 3.8 stars | 4.8 stars | +1.0 | | Average Session Duration | 2.8 min | 11.2 min | +300% | | Feature Deployment Cycle | 11 weeks | 9 days | -90% | | Notification Latency | 4.2 hours | 90 seconds | -96% | | Support Ticket Volume | 8,400/mo | 3,100/mo | -63% | | Crash-Free User Rate | 97.2% | 99.94% | +2.74 pts | ## Lessons The Finova engagement reinforced several principles that Webskyne now applies to every legacy modernization project. **Start with observability, not changes.** Instrumentation should precede refactoring. Without a clear baseline, teams cannot distinguish between improvement and regression, and they end up optimizing blind. **Treat migration as a product.** The strangler fig approach isn't just an engineering pattern; it's a product strategy. Each extracted service should deliver standalone user or business value, not merely serve as a stepping stone toward decommissioning the monolith. **Design systems reduce coordination costs.** The most underrated outcome of building FDS wasn't UI consistency—it was the elimination of the "which variant do we use?" meetings that had previously consumed hours of design and engineering time every sprint. **Shadow deployments beat big-bang releases.** Running real production traffic through the new system in parallel, before committing to the switch, gave the team confidence and surface-level understanding of edge cases that no amount of testing could replicate. **Engagement is a system property, not a feature.** Finova's engagement problem was never a single missing feature. It was the cumulative effect of latency, inconsistency, and friction across dozens of small touchpoints. Fixing one screen would have helped slightly. Fixing the entire delivery system created compounding returns. For engineering leaders considering a similar transformation, the message is straightforward: legacy modernization is not a technical cleanup project. It is a business reinvestment, and it should be measured, governed, and communicated with the same rigor as any new product launch. --- *This case study was prepared by the Webskyne editorial team based on actual client project data. Some figures have been rounded for readability.*