Analytics · SQL · Python · Tableau

The Data
Index

Sell-through told me what had happened. It never told me what belonged.
Scroll ↓
Cases
Five
Worked through
D—01, end to end
Stack
SQL · Python · Tableau
Data
226,904 raw orders
Finding
€1.53 M given away
€1,53 M
The finding

Given away in discount — because 92.8 % of orderlines were discounted at all.

226,904 raw orders · SQL · Python · Tableau
One case worked through end to end

Five cases where a number was supposed to settle an argument. One is worked through end to end — from raw export to a pricing rule the board could adopt.

Case D—01 · Eniac · Spanish e-commerce · 2017–18 data

Are discounts actually good for business?

Marketing said discounts drive growth. The board saw orders rising while revenue fell. We rebuilt the true discount signal from 226,904 raw orders — cleaned to 40,985 completed ones — and classified 9,992 products into 27 categories.

The finding: discounting was not a strategy, it was the permanent default. Depth sat at ~20 % all year and barely moved for Black Friday — 23 %, while the Spanish market went to 47 %. Eniac still had its best week of the year on less than half the market's discount. The deepest cuts sat on the cheapest products, which earn the least.

The decision: stop blanket discounting, concentrate it on four events — Black Friday, Christmas, New Year, Valentine's — and cap it by price tier: low ≤ 10 %, medium ≤ 15 %, high ≤ 10 %, premium ≤ 5 %.

Team project · Aylin Yildiz, Khadija, Elena, Rene · WBS Data Analytics Notebooks & 15-slide deck ↗
What it cost
92.8 %of all orderlines discounted
€1.53 Mrevenue given away — 16.3 % of potential
−0.30correlation: deeper discount ≠ more revenue
PythonpandasNumPyMatplotlibSeaborn
Data pipeline · Azure Medallion architecture
BronzeRaw ingest · Parquet
SilverCleaned · conformed
GoldCurated · analytics-ready
Counted, then decided.
me@aylin-yildiz.com

Aylin Yildiz — Buying × Data Analytics, Premium & Luxury. Open from October 2026 for Buyer or Assistant Buyer roles · Hamburg.