Back to portfolio

TheLook eCommerce Analysis

An end-to-end SQL audit of a fictional online store: where revenue is growing, who comes back, which categories actually make money after returns, and whether the tracking data can be trusted.

BigQueryStandard SQLLooker Studio2025 – 2026

TheLook is a fictional online fashion store, but the dataset behind it is large and realistic. I treated it like a real business and asked the questions a company actually cares about: is revenue growing, are customers coming back, which products make money after returns, and can we even trust what the tracking is telling us? The goal was a single end-to-end read of the store's health, backed by SQL I could defend line by line.

Source
bigquery-public-data.thelook_ecommerce
Span
2019 to 2026
Scope
Completed orders

A public, multi-year synthetic dataset of a fictional fashion store: customers, orders, order items, products, and web events. I worked across these tables with joins and held every revenue query to the same completed-order scope so the figures reconcile across sections.

  1. 01

    Frame the questions

    Decided up front what a healthy store looks like: steady growth, repeat customers, profitable categories, and tracking data that holds together.

  2. 02

    Model the revenue trend

    Aggregated completed orders by month and computed month-over-month growth with the LAG() window function.

  3. 03

    Profile customers and categories

    Split new versus returning buyers using a percent-of-total window, then joined order items to products to rank categories by profit and by return rate with COUNTIF.

  4. 04

    Audit the funnel

    Counted unique users at each event stage to check the journey held together before trusting any of the downstream numbers.

  5. 05

    Visualize in Looker Studio

    Surfaced the headline trends in an interactive Looker Studio report so the findings are easy to explore.

₹1.56L
latest monthly revenue
June 2026, up from under ₹1K in early 2019
27,470
customers analyzed
across completed-order history
88%
buy only once
heavy reliance on new acquisition
13%
top return rate
Clothing Sets, far above the rest
Monthly revenue, completed orders
₹0₹40K₹80K₹120K₹160K20192020202120222023202420252026

Completed-order revenue by month, 2019 to 2026. Hover any point for that month's orders and average order value. The latest month is the dataset's in-progress period.

Growth is real, but only in the mature phase

Monthly revenue climbs from a few hundred rupees in early 2019 to ₹1.56L by mid-2026. The early months are noisy: an eye-catching +125% jump in April 2019 is really just 7 orders becoming 13. I read the trend from the mature period instead, where 2026 grows a steady 0.5% to 16% month over month, which is a far more reliable base for forecasting and stock planning.

Most customers buy once and leave

Of 27,470 customers with a completed order, 88% placed exactly one. Leaning that hard on first-time acquisition keeps customer-acquisition cost under constant pressure, so even a small lift in the repeat rate is worth real money.

Revenue and profit don't rank the same

Outerwear & Coats and Jeans lead on profit (₹1.84L and ₹1.46L all-time), but the more useful find was on the cost side. Clothing Sets are returned 13% of the time and Plus sizes 10.5%, well above everything else. That points at a sizing-and-fit problem, not a demand problem.

The funnel didn't add up, so I stopped and checked

More users 'purchased' (80,151) than ever visited the home page (63,206), and the cart and purchase counts matched exactly. In a real store that is impossible. It turned out to be an artifact of how this synthetic dataset generates events, but flagging it mattered more than glossing over it: numbers get validated before they get reported.

  • Stand up a post-purchase email flow (a check-in plus a tailored recommendation within 30 days) to convert one-time buyers into repeat customers.
  • Audit sizing on Clothing Sets and Plus ranges. Clearer size guides and real fit feedback on the product page should pull those return rates down.
  • Plan off the mature-phase trend, not the sparse launch months, so forecasts aren't thrown off by early small-sample swings.

See the full dashboard

The headline trends are live in an interactive Looker Studio report you can explore yourself.

Open the Looker Studio dashboard