Model Reef: A Financial Forecasting Platform Built for Real-World Chaos

Forecast on any business along with all its sub-branches.

Make it accessible for both financial and non-financial users.

Prepare dashboard charts for number-phobic users.

Establish a hierarchy of operations based on categories.

Replace “variance analysis” with “here’s why you’re over budget” alerts.

Enable real-time visualisation of reports on the same screen with updated data.

Display the entire business structure on a single page.

Allow multiple users to work on the same model with role-based access restrictions.

Simplify and reduce the cost of subscription purchases.

Develop an LMS (Learning Management System) for non-financial users to learn and earn certification from ModelReef.

Designing the hierarchy of models and operations.

Handling large-scale numerical data in reports, including millions or trillions amounts for operations.

Loading reports for models with multiple sub-branches and more than 5 years.

Aggregation calculations causing delays in report loading.

Designing a user-friendly interface based on feedback to display all operations and reports on a single page.

Structuring the model hierarchy on a single page while allowing users to switch between models.

Implementing model sharing based on roles and subscription plans for multiple users.

Research and review multiple platforms to design an optimised hierarchy for models and operations.

Enhance Node.js numerical processing capabilities to handle large-scale calculations efficiently.

Develop a PoC for storing reports in JSON using Node.js and integrating Python with cutting edge database technologies.

Implement cutting edge database technologies for faster aggregation execution compared to traditional relational databases.

Transition report generation to Python, leveraging its support for MATLAB and cutting edge database technologies for improved performance.

• Studied finance teams for weeks, learning that 73% of forecasting time was spent cleaning data, not analyzing it.
• Discovered most failed forecasts stemmed from outdated assumptions buried in spreadsheet tabs.

• Prototyped in Figma, testing with number-phobic users.
• Prioritized glanceable dashboards with critical metrics that quickly update.

• Used Python and Apache Spark to build a data engine that consumes 100M rows in seconds.
• Deployed on Google Cloud with auto-scaling, that quietly adds server muscle during quarterly reporting crushes.

• Developed a web app using Angular, optimized for fat fingers on touchscreens during budget meetings.
• Added CEO Mode, a one-screen summary that hides the complexity but shows cash flow, risks, and growth levers.

• Stress-tested with data from a Fortune 500 retailer, 20TB of sales records, 15 currencies, 3 fiscal calendars.
• Hired actual accountants to break the system.

• Created onboarding videos narrated by a former finance manager.
• Launched with a 24/7 Forecast Hotline staffed by ex-bankers.

Adjust pricing strategy before the competition finishes their PowerPoint.

Reduced data cleanup time by 99%.

Spot cash flow gaps early enough to fix them with a polite email, not emergency loans.