Sigma Predictor

Get one standard deviation above the average!

Comment 3

Problem We Are Solving

Managing finances can be hard and, at times, overwhelming. Many personal finance tools focus mostly on budgeting, which only shows where your money went, not what it means or what to do next. It’s easy to lose track of interest, underestimate how small transfers impact future balances, or fall behind on savings goals without realizing it.

We wanted to solve the problem of uncertainty and guesswork in everyday money decisions and help people feel confident about managing their finances instead of stressed by them.

What It Does

  • Checking + HYSA overview with balances, interest earned since last login, and trend/forecast charts
  • What-if simulator to preview how transfers impact future interest
  • Goal tracking (e.g., “LA Plane Tickets”) with progress and remaining amount to reach the goal
  • Personalized Financial Insights tips engine (12 deterministic tip types like overdraft risk, subscription creep, safe transfer amount, best/worst week, goal shortfall, etc.)
  • Transfers API to move money between checking and HYSA and reflect it in analytics

How We Built It

  • Frontend: Next.js + TypeScript, Tailwind + shadcn/ui, Recharts for visualizations
  • Backend/Auth: Express + Passport Google OAuth2 + sessions + CORS
  • Data: MongoDB with users, accounts, and transactions collections + indexes
  • Hosting: nginx reverse proxy on Vultr Ubuntu server
  • Analytics: Server-side computation + forecasting
    • Linear regression for checking balance trends
    • GBDT-based forecasting for HYSA interest series
  • Tips Engine: Deterministic, explainable rules engine with full test suite and documentation
  • Tooling: Scripts for DB setup and realistic data seeding so the UI has meaningful activity immediately

Challenges We Ran Into

  • OAuth configuration (redirect URIs, consent screen, cookies/sessions)
  • Designing MongoDB schemas that support both analytics speed and product features
  • Generating realistic financial data (2 accounts per user + hundreds of transactions) without breaking balance logic
  • Building forecasting that is stable, explainable, and fast enough to run on-demand
  • Keeping financial tips actionable without feeling spammy or random

Accomplishments We’re Proud Of

  • End-to-end pipeline from transactions → analytics → forecasts → actionable tips
  • A 12-tip deterministic insights engine that is documented and unit-tested
  • A polished dashboard with what-if simulations and editable savings goals

What We Learned

  • How to integrate OAuth safely and debug common pitfalls
  • Why data quality is critical for financial UX (schema design, indexing, realistic seeding)
  • Turning analytics into product: insights must be explainable and actionable
  • Practical time-series handling (daily/weekly grouping, edge cases, estimating today’s interest)

What’s Next for Sigma Predictor

  • Real account linking (e.g., Plaid) with live transaction ingestion
  • Tip personalization (dismiss/snooze, sensitivity controls, feedback loops like “helpful / not helpful”)
  • Multi-account and category-based rules with improved merchant normalization

Built With

+ 5 more
Share this project:

Updates