Data Projects
analytics & data science work
A collection of analytics projects where I turn raw data into clear, visual answers to real business questions.
Business Intelligence & Analytics · UMass Amherst
Viewer engagement & growth trends in YouTube Shorts
- Role
- Data Analyst, team of 6
- Tools
- Tableau, Excel
- Year
- Spring 2026
- Category
- Analytics, Data Visualization
Challenge
Creators post constantly, but what actually drives views, likes and comments on YouTube Shorts? We wanted to help creators make data-informed decisions about their content.
Action
Using a Kaggle dataset of 799 Shorts, we cleaned the data, removed extreme outliers, built calculated fields like title length and engagement rate, and tested six research questions with scatter plots, trend lines, regression and keyword analysis in Tableau.
Results
Engagement is driven more by audience connection than reach: comments were the strongest predictor of engagement, shorter titles performed better, and the most-viewed videos tended to have the lowest engagement rates.
799
YouTube Shorts analyzed
6
research questions tested
0
null values after cleaning
24-35
character title sweet spot
Research questions
Does a video's age affect its views per day?
Key findings
- Videos get most of their views right after upload, then daily views decline.
- The trend is real but weak: age explains very little of the variation.
- p < 0.0001
- R² = 0.018
Lessons learned
Statistical significance isn't the same as predictive power: always check R² alongside the p-value.
Comments signal true audience connection better than views do.
Keywords only become meaningful when paired with a performance metric like views per day.
Scatter plots and trend lines can reveal tradeoffs that aren't obvious from the raw data.
Team project with Neeva Tuladhar, Carter Bouchard, Aiden Roy, Bobby Johnson, Jacob Lucas and Ken Huang.