MIGROS SI
A Case Study · 2021 – 2023

Migros
Sales
Intelligence

Three years of B2B sales — analyzed live in your browser. No Power BI, no installs, no logins. Just SQL running in WebAssembly against a single parquet file.

Rows
1.78M
Stores
2,097
Engine
DuckDB
Scroll
The Brief

What was actually asked.

Whichever date the user selects from the filter, YTD and MTD are computed against that date — not the data's max date. If the user picks 2022-08-16, YTD = [2022-01-01 … 2022-08-16]. And I want the same windows compared to the previous year as a percentage growth.

— Mete Adıyaman, BI-Boost · May 2026
Mete's evaluation criterion

“Visuals aren't important to me — what matters is that you build the structure and can explain it.”
That sentence is the brief behind this brief. The site is built for it.

Six pages

The whole thing, top to bottom.

Each page does one job. The Dashboard answers the case; the rest shows the work behind it.

01

Dashboard

The case answered. Pick a date — YTD, MTD, and PY% compute live via DuckDB-WASM SQL in your browser.

Start here
02

Method

Architecture diagram, ETL pipeline (Excel → Parquet → WASM), and the actual SQL window functions explained.

Open
03

Segments

Beyond the brief. Store leaderboard, product top-N, seasonality heatmap, anomaly flags, data-quality report.

Open
04

Recommendations

Three concrete actions extracted from real query results — each with the SQL that surfaced it.

Open
05

vs Power BI

Honest side-by-side. When Power BI wins. When custom web wins. This isn't a takedown — it's about fit.

Open
06

About

Who built this, why it represents how I approach data problems, and how to reach me.

Open
Where the case is answered

Start with the
Dashboard.

Pick a date. Watch YTD, MTD, and prior-year comparisons recompute from a SQL query running against sales.parquet inside your browser.

Open the dashboardSee how it works
Build time
~7 days
Data weight
5.8 MB
Backend
None