The Question

Revenue grew steadily in the first half of the year, then dropped in Q3. The business question: what caused the dip, and where should the team focus next?

The Data

One year of order-level data with these columns: date, category, region, customer (new or returning) and revenue.

How I Cleaned the Data

• Removed duplicates: repeated order IDs

• Checked missing values: filled or dropped depending on the column

• Fixed formats: standardised dates and category names

• Flagged outliers: used the IQR method on order values

import pandas as pd

df = pd.read_csv("retail_sales_sample.csv", parse_dates=["date"])

df = df.drop_duplicates()

df["category"] = df["category"].str.strip().str.title()

df["month"] = df["date"].dt.to_period("M")

monthly = df.groupby("month")["revenue"].sum()

Finding 1: Revenue Fell in July to September

Monthly revenue held at roughly \(52-56K from January to June. It then dropped to about \)44K in July, $41K in August and $39K in September, before recovering in Q4.

Finding 2: Electronics Drove Most of the Drop

Comparing Q3's monthly average to the rest of the year, Electronics fell about 31%, the biggest drop. Beauty fell 20%, Fashion 14% and Home 12%. Grocery stayed flat (+3%), so everyday demand held up.

Finding 3: The South Region Was Hit Hardest

Every region declined in Q3, but the South fell about 33%, compared with 15-22% elsewhere. That points to a regional problem on top of the category issue.

One More Insight

Returning customers spent about $140 per order versus $113 for new customers, roughly 24% more. Keeping repeat buyers matters.

Recommendations

Investigate Electronics in Q3: check pricing, stock and competitor promotions.

Review the South region: look at delivery times, local campaigns and stock levels.

Invest in retention: loyalty rewards and win-back emails for returning customers.

Track weekly KPIs: build a Power BI or Tableau dashboard so dips are caught early.

Tools used: Python (pandas, matplotlib), Excel, SQL

Contact with me on LinkedIn

Please click

Click to view the dataset