Retail Sales Performance Dashboard

Why this project

This is the classic "which country, which month, which product" question every retail business asks — and it's a great first BI project because the answer isn't in one table, it's in how you rank and segment across several dimensions at once. You can build this exact project with the free, public UCI Online Retail dataset (a real 500,000+ row transaction export from a UK-based online retailer), so every query below is copy-paste runnable.

The data

One flat transaction table is all you need to start:

Step 1 — Clean it in SQL

Retail exports are messy by default — cancelled invoices, returns, and missing customer IDs will all silently wreck your totals if you don't filter them first:

create table stg_retail_sales as
select
  invoice_no,
  stock_code,
  description,
  quantity,
  invoice_date,
  unit_price,
  customer_id,
  country,
  quantity * unit_price as line_revenue
from raw_online_retail
where invoice_no not like 'C%'   -- exclude cancellations
  and quantity > 0                -- exclude returns
  and unit_price > 0
  and customer_id is not null;

Sanity-check it immediately:

select country, sum(line_revenue) as revenue
from stg_retail_sales
group by country
order by revenue desc
limit 10;

If the UK doesn't dominate this result for the UCI dataset, something upstream in your cleaning step is wrong — that's a useful self-check before you ever open Power BI.

Step 2 — Rank products with DAX

The naive approach is sorting a table visual by a column — it works until someone filters by country or date, and the ranking silently stops updating. Use RANKX instead, so the rank recalculates inside whatever filter context the report is in:

Total Revenue = SUM(Sales[LineRevenue])
 
Product Rank =
RANKX(
    ALL(Sales[Description]),
    [Total Revenue],
    ,
    DESC
)
 
Best Selling Month =
CALCULATE(
    SELECTEDVALUE('Date'[Month Name]),
    TOPN(1, ALL('Date'[Month Name]), [Total Revenue], DESC)
)

Step 3 — Segment customers

A simple order-count bucket is enough to make the "which customers matter most" question visual without building a full RFM model:

Customer Orders = DISTINCTCOUNT(Sales[InvoiceNo])
 
Customer Segment =
SWITCH(
    TRUE(),
    [Customer Orders] >= 10, "VIP",
    [Customer Orders] >= 3,  "Repeat",
    "New"
)
 
Segment Contribution % =
DIVIDE(
    [Total Revenue],
    CALCULATE([Total Revenue], ALL(Customers[Customer Segment]))
)

Step 4 — Lay out the report

Three pages cover the whole story without overwhelming a first-time viewer:

1. Overview — total revenue, revenue by country map, revenue by month line chart.
2. Products — top 20 products by Product Rank, with a slicer for country and month.
3. Customers — segment contribution as a donut, plus a table of top customers by revenue.

How to build your own version

1. Download the Online Retail dataset from the UCI Machine Learning Repository (or Kaggle's mirror of it).
2. Load it into Power BI directly, or stage it in Postgres/MySQL first if you want SQL practice.
3. Run the cleaning query above and validate the top-10-countries check.
4. Add the DAX measures for ranking, best month, and customer segments.
5. Build the three report pages.

Talking about it in an interview

The detail worth mentioning unprompted: why RANKX over a plain sort. A sorted column looks identical to a dynamic rank until someone applies a filter — showing you understand that distinction is a stronger signal than the dashboard itself.