Coffee Sales Performance Dashboard

Problem

A coffee chain in New York needed a clear way to monitor business performance, compare results across stores and categories, and anticipate future sales from historical data. Transactional data was available, but it required cleaning, modeling, and visualization to become actionable indicators.

Methodology

Transactional sales data was processed with Python to clean, aggregate, and prepare key metrics such as total sales, average ticket, and ticket count. Time series models were then built with Prophet to generate sales forecasts, and the results were stored in PostgreSQL through Supabase, using views to combine actual sales and predictions. Finally, an interactive Looker Studio dashboard was designed with executive views, store comparisons, product analysis, and accuracy metrics such as MAPE.

Results

The project delivers a clear visualization of current performance and future sales forecasts, enabling comparisons between stores such as Astoria, Hell’s Kitchen, and Lower Manhattan. The dashboard makes it easier to read trends, evaluate model accuracy, and analyze customer behavior through average ticket, product detail, and exportable tables.