Retail decisions were flying blind. The data lived in scattered CRM and ERP CSV files with no single source of truth. Reporting depended on ad-hoc SQL and spreadsheets, which made it slow and error prone. Basic questions went unanswered: when do we peak, which categories drive revenue, and who are our best customers. Non-technical teammates also lacked a simple way to see the answers.
gold.fact_sales, gold.dim_customers, and gold.dim_products. Reusable SQL covers trends, running totals, category share, top products, and customer lifetime value. Results show up as interactive charts with the query beside them. The site is hosted on GitHub Pages and drops into Squarespace. No BI license required.
With the data in one place, the story is clear. Sales peak every year in November and December, so promos, inventory, and staffing can be pulled forward. Revenue is highly concentrated. Bikes account for about 96.46 percent, which points to assortment diversification and accessories cross-sell. Customer value shows a long tail. The top 10 customers are about 0.45 percent of revenue and the top 20 are about 0.83 percent, so loyalty efforts should target the middle. From this run we have 29.36 million dollars in sales, 18,484 customers, and coverage from 2010-12-29 to 2014-01-28. Leaders can now self-serve answers in seconds and analysts reuse the same trusted queries.
Data is loaded from CSVs into SQL Server Express and authored in SSMS. Modeling follows Bronze, Silver, Gold and uses a star schema with window functions. Everything is versioned in Git and hosted on GitHub Pages. The front end is HTML, CSS, and JavaScript with Chart.js for visuals, Papa Parse for CSV, and Prism for code highlighting. The page embeds in Squarespace with cache-busted assets. Static hosting keeps cost and risk low.
