Coffee Shop Sales: When and Where Maven Roasters Sells Most

Exploratory data analysis of sales transactions across three coffee shop locations in NYC.

Repository: View on GitHub

Problem

Maven Roasters, a fictitious coffee shop chain operating three locations in New York City, needs to understand its sales patterns to make better operational decisions: when are the demand peaks?, which products generate the most revenue?, and on which days and hours should staffing be reinforced?

Methodology

• Loaded and cleaned transactional data with pandas (date, time, location, product, quantity, unit price)
• Feature engineering: total_amount (quantity × unit price), day of the week, month, and hour of each transaction
• Outlier handling: transactions above 300 were replaced with the median, as they were considered artificial/erroneous data points
• Monthly revenue trend analysis with seaborn
• Heatmap of transaction counts by day of the week and hour of the day
• Transaction breakdown by product category
• Ranking of products by total revenue and transaction count

df["total_amount"] = df["transaction_qty"] * df["unit_price"]
df["day_of_the_week"] = df["transaction_date"].dt.day_name()
df["month"] = df["transaction_date"].dt.month_name()
df["hour"] = pd.to_datetime(df["transaction_time"], format='%H:%M:%S').dt.hour

Results & Discussion

• Identified the monthly revenue trend, revealing seasons of higher and lower activity
• The day/hour heatmap revealed the busiest time windows, useful for planning staff shifts
• Determined which product categories had the most transactions and which individual products drove the most revenue — not always the same as the highest-volume products

Conclusions

1. Sales volume and revenue generated aren’t always aligned by product — some low-volume products drive more revenue per unit
2. Day/hour patterns are key for optimizing staff allocation and avoiding bottlenecks during peak hours