Safiul Azam Data & Information Management
Power BIDAXProfitability

Global Superstore — Performance & Profit Leakage

A four-page Power BI report over 51,290 order lines that answers a question the sales chart hides: revenue nearly doubled in four years, so why did margin go backwards?

RoleAnalyst & dashboard developer
OrganisationIndependent project
Year2026
ToolsPower BI · DAX · Power Query

This project can be shown in full. It uses a public dataset with no sensitive content, so every figure, method and result on this page is the real thing — the visuals are screenshots of the working output, not rebuilds.

Context

Global Superstore is a four-year transactional file — 51,290 order lines across seven markets. The obvious dashboard is a sales trend, and the obvious conclusion is that things are going well: sales grew 26.3% and orders 28.7% year on year.

That reading is wrong, or at least incomplete. Over the same period the profit margin moved down 0.3 points. A business can grow itself into a worse position, and the report’s job was to make that visible rather than let the revenue line tell a flattering story on its own.

Approach

  • Lead with the contradiction. The subtitle states it outright: “sales nearly doubled in four years while margin moved 0.7 points”. A dashboard that buries its own headline is a report nobody reads twice.
  • Make the loss a first-class KPI. Profit leakage — the value given away in discounts on loss-making lines — sits in the KPI row beside sales and profit, not in a footnote. It is the only red tile on the page, deliberately.
  • One axis, always. Sales and profit are plotted together in dollars on a single scale rather than on a dual axis. Profit being a thin sliver against revenue is the finding; a second axis would have hidden it by rescaling.
  • Margin at the bar end. Category and market bars are sized by revenue but labelled with margin, so a big-but-thin segment is impossible to miss.
  • Four pages: Overview, Where the Profit Goes, Geography, Product.

The four pages

Built on the public Kaggle Global Superstore dataset — 51,290 order lines, FY2011–FY2014. Real screenshots of the working report.

1 Performance overview

Real output · public dataset Power BI overview page showing sales, profit, margin, orders and profit leakage KPIs with a monthly trend
$12.64M sales, $1.47M profit, 11.6% margin, −$920K leakage. Sales and profit share one dollar axis, so the gap between the two lines is the story. Technology carries a 14.0% margin against Furniture’s 7.0% on almost identical revenue.

2 Where the profit goes

Real output · public dataset Power BI page decomposing profit leakage by discount band and sub-category
The leakage decomposition — which discount bands and sub-categories are destroying margin, and how much each one costs.

3 Geography

Real output · public dataset Power BI geographic page showing sales and margin by market and country
Market-level performance. APAC leads on revenue at $3.59M, but Canada returns a 26.6% margin on just $67K — volume and profitability are not the same league table.

4 Product

Real output · public dataset Power BI product page ranking sub-categories by revenue and margin contribution
Sub-category detail, ranked so the profitable long tail is as visible as the loss-making headline items.

Outcome

−$920K profit leakage identified

The report isolates $920K of profit leakage — and shows it growing at 26.2%, almost exactly in step with sales. That is the crux: the discounting behaviour driving growth is scaling the loss alongside it.

Two further findings have direct commercial consequences. Furniture returns 7.0% margin against Technology’s 14.0% on comparable revenue, and Q4 carries 34% of the year with February as the trough — so a discount policy set as an annual average is wrong for both ends of the calendar.

What I’d do differently

The dataset has no cost-to-serve, so “profit” here is gross of logistics. Some of what looks like a margin problem in the bulky Furniture category is probably a shipping problem, and I cannot separate the two from this file alone.

I would also want the discount approval chain. Knowing that $920K leaked is useful; knowing which desk authorised it is what actually changes behaviour.