Grabbed a pizza sales dataset off Kaggle, mostly to see if I could turn raw order data into something a stakeholder would actually want to read. Four CSVs, orders, order details, pizza types, pizzas, none of it particularly clean, all of it waiting to be asked the right questions.
Started with the basics: total revenue, average order value, total pizzas sold, total orders, average pizzas per order. Sounds obvious, but half the work in any analysis is just making sure your foundational numbers aren't quietly wrong before you build on top of them.
Wrote SQL queries to pull sales trends by pizza size and category, trying to figure out which items were carrying the business and which were dead weight. Nothing exotic, mostly joins and aggregations, but the kind of grunt work that actually tells you something true about customer behavior instead of what you assumed going in.
Numbers in a query result don't convince anyone of anything, so I built an Excel dashboard on top, daily and hourly trends, top and worst-selling pizzas, size and category breakdowns. The worst-sellers list was oddly satisfying, it's easy to talk about what's working, less common to point clearly at what isn't.
Turned out sales weren't evenly spread across the day or the week, and a handful of pizza types were doing most of the heavy lifting while others barely moved. Nothing shocking in hindsight, but that's usually how good analysis feels, obvious after the fact, invisible before it.
Made with Bullet