Pizza Sales Analysis & Revenue Forecasting
Final Report
Jonathan Katz
2025-12-08
Executive Summary
This report analyzes one year of pizza restaurant sales data (2015) to understand customer behavior and forecast future revenue.
Key Findings:
- Annual Revenue: $817,860 from 21,350 orders
- Peak Hours: 12pm (lunch) and 5-7pm (dinner)
- Best Day: Friday generates the highest revenue
- Top Seller: The Classic Deluxe Pizza
- Most Popular Size: Large (46% of all sales)
- Average Order Value: $38.31
Forecast Performance:
The 7-day revenue forecast model performed well during typical weeks but struggled during the Christmas holiday period. This highlights the importance of incorporating seasonality and holiday effects into forecasting models.
Recommendations:
- Staff fully during peak hours (12pm and 5-7pm)
- Run promotions on slow days (Sunday)
- Ensure Classic Deluxe ingredients are always stocked
- Incorporate holiday calendar into future forecasting models
Business Questions
This analysis answers six key questions:
- What will total sales revenue be for the next 7 days?
- What are the peak hours for orders?
- Which day of the week has the highest sales?
- What is the best-selling pizza?
- What is the most popular pizza size?
- What is the average order value?
Dataset Overview
Source: Kaggle Pizza Restaurant Sales Dataset
Period: January 1, 2015 - December 31, 2015
Size: 48,620 order line items
| Metric | Value |
|---|---|
| Total Orders | 21,350 |
| Total Pizzas Sold | 49,574 |
| Total Revenue | $817,860 |
| Unique Pizza Types | 32 |
Key Findings
What are the peak hours for orders?
Answer: Peak hours are 12pm (lunch) and 5-7pm (dinner).
Recommendation: Ensure full staffing during these windows.
Which day of the week has the highest sales?
Answer: Friday has the highest sales. Sunday has the lowest.
Recommendation: Run promotions on slow days (Sunday) to increase traffic.
What is the best-selling pizza?
Answer: The Classic Deluxe Pizza is the best-seller.
Recommendation: Ensure ingredients for this pizza are always in stock.
What is the most popular pizza size?
Answer: Large (L) is the most popular size, accounting for 46% of all sales.
Recommendation: Optimize pricing and inventory around Large pizzas.
What is the average order value?
Answer: The average order value is $38.31.
Recommendation: Set delivery minimum at $35-40 to optimize operations.
Revenue Forecast
Daily Revenue Trend
7-Day Forecast Model
We used an ARIMA model trained on 358 days of data to predict the final 7 days of the year.
| Date | Forecast (\() </th> <th style="text-align:right;"> Actual (\)) | Error ($) | |
|---|---|---|---|
| 2015-12-24 | 2112 | 2138 | 26 |
| 2015-12-26 | 2519 | 1643 | -876 |
| 2015-12-27 | 2224 | 1419 | -805 |
| 2015-12-28 | 2235 | 1637 | -597 |
| 2015-12-29 | 2218 | 1353 | -865 |
| 2015-12-30 | 2208 | 1338 | -870 |
| 2015-12-31 | 2268 | 2916 | 648 |
Forecast Accuracy
| Metric | Value |
|---|---|
| MAE (Mean Absolute Error) | $670 |
| MAPE (Mean Absolute % Error) | 42.7% |
Lessons Learned
Why the Forecast Missed
The model significantly underperformed during the test period. Upon investigation, the reason is clear: the test period was Christmas week (December 25-31).
What happened:
- Christmas Day closure — Revenue dropped to zero when the restaurant was closed
- Post-holiday behavior — Customers ate homemade food with family instead of ordering pizza
- Atypical demand — The week between Christmas and New Year is not representative of normal business
Why the model couldn’t predict this:
- The model learned patterns from “typical” weeks throughout the year
- It had no information about holidays or special events
- Christmas only occurs once in the training data, so the model couldn’t learn this pattern
How to Improve
For a production forecasting system, I would:
- Add a holiday calendar — Flag holidays as a feature in the model
- Exclude anomalies — Train on typical weeks, handle holidays separately
- Use a model designed for seasonality — Facebook Prophet handles holidays explicitly
- Incorporate external data — Weather, local events, promotions
This experience demonstrates that understanding your data’s context is as important as the technical modeling.
Summary of Recommendations
| Finding | Recommended Action |
|---|---|
| Peak hours: 12pm, 5-7pm | Full staffing during peaks |
| Friday is best day | Maximize Friday capacity |
| Sunday is slowest day | Run Sunday promotions |
| Classic Deluxe is #1 | Stock Classic Deluxe ingredients |
| Large is most popular | Optimize Large pizza pricing |
| AOV is $38.31 | Set $35-40 delivery minimum |
| Forecast missed holidays | Add holiday features to model |
Conclusion
This analysis provided actionable insights into pizza restaurant operations:
- Clear identification of peak hours and days for staffing optimization
- Product performance data for inventory management
- Average order value for pricing and promotion strategy
- A forecasting model with documented limitations and improvement path
The forecast’s failure during Christmas week turned into a valuable learning opportunity, demonstrating the importance of domain knowledge and contextual understanding in data analysis.
Report generated on 2025-12-08