Netflix Recommendation Correlation - Excel
Project links
Skills
About this project
Netflix Content Recommendation Using Excel
Objective
To explore how viewer preferences across different TV shows are connected and simulate a basic recommendation logic using Excel-based tools like correlation matrices, scatter plots, and pivot tables.
Problem Statement
Streaming platforms such as Netflix rely heavily on user behavior data to recommend content. This project aimed to simulate that logic in a simple, transparent way using Excel, by analyzing how viewers’ interests in one show may indicate interest in another.
Tools and Techniques Used
- Microsoft Excel
- Correlation Matrix using CORREL function
- Pivot Tables for organizing and summarizing user data
- Scatter Plots to visualize relationships
- Conditional Formatting for highlighting trends
Methodology
1. Define the Business Goal
The central question was: Can we identify patterns in show preferences to recommend similar content?
2. Data Preparation and Structuring
- Collected sample data on user engagement for multiple shows such as viewing time and completion rate
- Structured the data in Excel, using user IDs as rows and show titles as columns
3. Correlation Analysis
- Used the CORREL function to calculate pairwise correlations between shows
- Identified strong correlations such as Dahmer and Dexter, implying that viewers who watched one were likely to enjoy the other
4. Scatter Plots
- Created scatter plots to visually explore the relationship between pairs of shows
- Helped confirm which shows had consistent viewing patterns among users
5. Pivot Tables for Summary Metrics
- Created pivot tables to analyse average watch time, viewer count, and completion percentages per show (Top 10 as shown below)
- These metrics added deeper context to the correlation results
6. Recommendation Simulation
- Created a simple logic: If Show A has a correlation above 0.8 with Show B, recommend Show B to users who liked Show A
- This created a basic framework of how streaming platforms might leverage viewing behavior for content suggestions
Key Learnings
- Excel provides a strong foundation for beginner-level recommendation system modeling
- Correlation analysis is a valuable method to uncover content similarity based on viewer data
- Scatter plots and pivot tables enhance data storytelling and interpretation
Future Scope
- Segment users by genre preference or demographics
- Include additional variables such as user ratings or viewing frequency
- Recreate the model in Power BI for enhanced interactivity and visuals