UMG YouTube Engagement & TikTok Policy Change Sentiment Analysis

This project explores how social media analytics can support business decisions for Universal Music Group (UMG). I combined Twitter sentiment analysis about UMG’s contract clash with TikTok with an analysis of YouTube engagement metrics for UMG, its artists and its subsidiary Republic Records.

The work blends NLP, sentiment lexicons and dashboarding: tweets scraped via Apify from January–April 2024 were cleaned and scored using both TextBlob and VADER; YouTube metrics were collected using the YouTube Data API and visualised in Tableau to uncover performance patterns across channels.

The analysis focused on two core questions:

• Sentiment: How does the public perceive UMG’s policy change and dispute with TikTok, and how does that sentiment evolve over time?
• Engagement: How does UMG’s own YouTube channel perform compared to its superstar artists and Republic Records, and what publishing patterns drive engagement?

The goal was to turn raw social data into practical recommendations on content, partnerships and platform strategy for UMG’s marketing and leadership teams.

• Data sources: Used the YouTube Data API to gather channel and video statistics for UMG, Taylor Swift, Rihanna, Ariana Grande, Billie Eilish, Nicki Minaj and Republic Records, stored in UMG_artist_YT.csv and UMG_YT.csv.
• Tweet scraping: Used Apify’s “Tweet Scraper V2” actor to collect tweets mentioning “TikTok AND UMG” between January and April 2024, producing month-wise CSV files.
• Cleaning & preprocessing: Applied the emoji library to convert emojis (e.g. 👍 → :thumbs_up:), used re to strip usernames, hashtags, links and punctuation, tokenised with NLTK, removed English stopwords and lemmatised tokens to prepare text for sentiment scoring.
• Sentiment analysis:
  • TextBlob: polarity and subjectivity to label tweets as positive, negative or neutral based on polarity (>0, <0, =0).
  • VADER: lexicon-based scores (neg, neu, pos, compound) to categorise sentiment; chosen for its suitability to informal social media text.
• Visualisation & dashboards: Used Python (Matplotlib, Seaborn, WordCloud) and Tableau to build dashboards for: UMG vs TikTok Twitter sentiment and Universal Music Group: YouTube Engagement.

• Tweet volume & period: 321 tweets were collected between 1 January and 23 April 2024, with a clear spike in conversation and views in February following the public escalation of the dispute.
• Sentiment distribution: VADER showed ~44% positive, 38% negative and 18% neutral tweets, suggesting that while many users criticised impacts of song removals, a slightly larger portion supported UMG’s stance on artist rights.
• Model comparison: VADER outperformed TextBlob by capturing informal tweet context more accurately (44% positive, 38% negative, 18% neutral).
• Temporal sentiment: Month-wise sentiment charts showed negative tweets decreasing and positive sentiment slightly increasing across January–April, indicating some stabilisation of perception over time.
• Word clouds: Sentiment-specific word clouds emphasised terms like “TikTok”, “muted”, “artist”, “UMG”, “failed” and “song”, capturing the core narrative around content removal and artist support.
• YouTube engagement patterns: UMG’s own channel underperforms relative to artists like Taylor Swift and Ariana Grande in total views; Tuesdays and October are dominant upload times, and top-performing videos often feature high-profile artists or viral-friendly topics.
• Business story: The analysis shows UMG must balance protecting IP and artist rights with maintaining a positive public image and strong platform presence, especially on TikTok, YouTube Shorts and Reels.

This project deepened my practical skills in social media analytics: from ethically scraping data and cleaning noisy tweets with emojis and slang, to applying multiple sentiment tools, validating them, and presenting findings in interactive dashboards.

It also highlighted the limitations of platform data and sentiment tools - demographic biases, API restrictions and disagreements between TextBlob and VADER - and the need to communicate these caveats clearly when advising decision-makers. Finally, turning charts into concrete recommendations for UMG’s YouTube and TikTok strategy sharpened my ability to bridge technical analysis with marketing and strategy.