Fund & Stock Screening Platform

TL;DR

This platform brings professional-grade investment screening to advisors and self-directed investors: 50+ filter parameters, watchlists, saved screens, research and recommendations, and portfolio analytics, wrapped in a consumer-grade experience.

My product bet was that speed of exploration is the product. Every design and architecture decision, from client-side filtering budgets to additive-only schema evolution, protected the feeling of instant discovery while the business layer (leads, research distribution, engagement loops) grew around it.

---

1) The problem

Investment research firms typically publish reports. Users increasingly expect tools: screen, compare, save, track, and act.

The gap had three layers:

• User layer: advisors and advanced retail investors needed fast multi-parameter screening without spreadsheet exports and manual comparison.
• Business layer: the firm needed qualified lead capture, research distribution, and engagement loops in one governed surface.
• Operations layer: content teams needed publishing workflows with role-based control, not developer tickets for every recommendation update.

---

2) Product decisions that mattered

Speed as a requirement, not an aspiration

Performance budgets were written into the PRD as testable requirements: P95 cold render under 2.5 seconds, client-side filtering across 1,500+ funds under 150ms, CSV export under 2 seconds. That forced the architecture conversation early — client-side filtering with additive-only schema evolution — instead of after users complained.

Exploration converts better than forms

The advisor request flow auto-attaches the user's active filter context. Instead of a generic "contact us" lead, sales receives what the user was actually screening for. Analytics events (filter changes, watchlist adds, comparison opens, advisor requests) make the funnel measurable end to end.

Scope discipline

MVP shipped must-haves: screening, watchlists, saved screens, auth, and the advisor flow. Smart presets, ESG filters, and AI-assisted insights were explicitly deferred with reasoning documented, keeping the launch surface testable.

Refactor before scale

After MVP, the team completed a 10-ticket architecture program in one quarter: a service layer with repository contracts, decomposition of a monolithic state context into five focused contexts, unified research-domain logic, and build-time guardrails (forbidden imports, module size thresholds). Deliberately choosing stabilization over feature churn is a product decision, and it was the right one before regional expansion.

---

3) Governance and trust

• RBAC everywhere that matters: super-admin, admin, and viewer roles enforced at the route level for all admin APIs.
• Validation as a guardrail: every API write validated with schemas, so data integrity survives schema evolution.
• Privacy by design: minimal PII (email plus optional name), no PII in logs, and session purges after inactivity.
• Readiness evidence, not readiness claims: a 31-check smoke suite spanning code quality, public endpoints, authentication for all seeded roles, sessions, assets, admin APIs, and RBAC had to pass before the product was called production-ready.

---

4) Engagement and retention loops

Beyond screening, the platform builds habitual return paths:

• Watchlists and saved screens turn one-off research into recurring sessions.
• Performance dashboards (active, past, and today views with sector, theme, and risk filters) give users a reason to check back.
• Referral and rewards systems (tiered referral codes, daily rewards, achievements, leaderboard) add compounding engagement mechanics.
• Email notifications (welcome, portfolio alerts, recommendations, price alerts) with a user-facing preference center respect the opt-in relationship.

Target metrics were defined up front for the first 90 days: weekly active users, Day-7 retention, share of sessions using advanced filters, advisor request rate, and saved screens per active user, so success is judged against pre-committed numbers rather than retrofitted stories.

---

5) What I learned

• Non-functional requirements are product requirements. Filter latency shaped user trust more than any single feature.
• Deferring features is cheaper than deleting them. The explicitly-deferred list kept stakeholders aligned without inflating the launch surface.
• Architecture governance is a PM concern. Build-time guardrails meant velocity later did not depend on institutional memory.
• Lead quality beats lead volume. Context-rich advisor requests changed the sales conversation from cold to warm.

This case study presents a public-safe summary of the work.