FIJI Insights

FIJI Insights required building a financial product where precision is not optional. A single rounding error in an EBITDA calculation or a misapplied discount rate would undermine the entire value proposition.

Here is what made this hard:

• •Proprietary valuation math. The engine computes adjusted EBITDA, applies industry-specific risk modifiers from weighted questionnaire answers, calculates discount rates using a multi-variable CAPM-derived formula, derives valuation multiples with floor/ceiling constraints, and produces min/mid/max market value ranges — all parameterized by admin-configurable constants stored in the database. Two distinct strategy engines handle brokerages and teams differently.
• •Platform migration. FIJI had existing user data and valuation records in Bubble. The new system needed to ingest legacy JSON structures, normalize financial arrays across schema versions, and maintain backward compatibility with older saved valuations while moving everything to Supabase on Postgres.
• •Two-sided marketplace with deal-room mechanics. Buyers create rich profiles with media, documents, and geographic targeting. Sellers browse and filter. But before any real conversation happens, NDA documents must be uploaded, sent, signed by both parties, or declined — with full status tracking and role-aware UI on both sides.
• •Real-time communication. Chat between buyers and sellers runs on Supabase Realtime subscriptions with optimistic updates, file attachments via cloud storage, read receipts, archive/delete per-user, and transactional email notifications through Loops.so.
• •Subscription and access gating. Valuation results, insights, and marketplace features are gated behind subscription status. The system needed clean upgrade flows, activation confirmation, and plan-tier differentiation for buyer profiles (basic, standard, premium).
• •Cross-application auth handoff. Users arriving from an external Bubble application needed seamless token exchange, session creation, user upsert, and secure redirect — with JWT issuer validation and open-redirect protection.

Without strong product leadership, this kind of system drifts into inconsistency fast. Financial logic gets buried in UI components. Marketplace and chat become disconnected. Migration becomes a permanent liability instead of a clean cut.

Prizmstack owned every layer of the product:

• •Technical Architecture
• •Product Management
• •User Stories
• •UX/UI Design
• •Full-Stack Development
• •Scrum Sprint Delivery
• •QA
• •API Design
• •Data Architecture & Migration Strategy
• •DevOps & Database Operations

1. Valuation Engine

A server-side calculation engine with Zod-validated payloads, strategy-pattern routing (brokerage vs. team), EBITDA computation with pro-forma adjustments, industry risk scoring from weighted questionnaire answers, CAPM-derived discount rates, and configurable system constants. Results include min/max/mid-market value ranges, supporting metrics, and automatic PDF report generation via React PDF.

2. Try Valuation Wizard

A multi-step guided flow that walks brokerage owners through financial inputs (revenue, GCI, company dollar, operating expenses, non-recurring items, W2/1099 agent splits, owner compensation) and qualitative risk questions. Includes real-time calculation feedback, form validation, and draft persistence to local storage.

3. Scenario Modeling & Insights

Interactive sliders for agent count changes and fee adjustments that recompute valuation metrics in real time using Valtio reactive state. Baseline vs. current comparison snapshots. Per-agent and per-closing derived metrics surfaced across dashboard cards and insight views.

4. Review Buyer Marketplace

Acquisition-ready buyers publish rich profiles with company details, geographic targeting by state, business type (Best Brokerage, teams, escrow, find), agent range, transaction volume, media galleries, video links, and document uploads. Sellers browse, filter, and favorite listings. Marketplace detail pages generate dynamic Open Graph metadata for sharing.

5. Buyer Profile Management

Multi-step profile creation with image and document uploads to Supabase Storage, publish/unpublish toggles for marketplace visibility, and tiered plan selection (Basic, standard, premium) with feature differentiation.

6. Secure Chat & NDA Workflow

Buyer-seller messaging with integrated Realtime Supabase subscriptions, file attachments, read receipts, and per-user rate limits. Complete NDA lifecycle management: upload, send, sign, decline, or cancel — with role-aware actions, signature tracking, and full audit chains on every conversation.

7. Onboarding & Close-App Handoff

Post-sign-up welcome flow with guided road maps: run valuation, complete buyer profile, review insights, and marketplace. Dashboard intake completion progress, surfaces valuation reduction summaries, and provides contextual CTAs. Marketplace onboarding wizard introduces the access model/selection concept to new users.

8. Authentication & Cross-App Handoff

Supabase Auth with email/password, Google OAuth, password reset, and email confirmation via Loops. Token-based JWT handshake for users arriving from an external Bubble app, with JWT header validation, user sync, access provisioning, and callback URL processing.

9. Admin Tools

Protected admin interface for managing valuation constants, risk weights, and answer weight tables that directly control the valuation engine output — giving product owners the ability to tune the model without code changes.

Every feature started as a user story with defined acceptance criteria, informed by the product roadmap and stakeholder input. Design preceded development. Architecture preceded implementation. We ran structured Scrum sprints with clear priorities.

The valuation engine was built and validated against Excel reference models before any UI touched it. Marketplace and chat were developed in parallel tracks with shared type contracts. Migration from Bubble was planned as a phased cutover with backward-compatible data normalization, not a big-bang rewrite.

Iteration was tight. Stakeholders reviewed working software every sprint. QA validated financial precision against known outputs. When the engine math spec changed (from 3.29b to 3.31a), the architecture absorbed it cleanly because the system was designed for exactly that kind of evolution.

FIJI Insights went from an ambitious concept with legacy Bubble constraints to a production platform with a proprietary valuation engine, a two-sided marketplace, and secure deal room messaging — all built on a modern, extensible stack.

Brokerage owners can now get a data-driven valuation in minutes, model scenarios interactively, and download a formatted PDF report. Buyers can publish acquisition profiles and reach sellers directly. Every deal conversation is NDA-gated and auditable.

The product has a clear onboarding funnel, subscription monetization, admin-controlled financial parameters, and an architecture designed to absorb new business types (property management, escrow) as the platform grows. The foundation is not fragile — it was built by a team that owned the full picture.

Complex products do not fail because the code is bad. They fail because no one owned the whole picture — the business logic, the user experience, the data model, the migration path, and the sprint cadence that ties it all together.

Prizmstack owned it. That is what we do.