UNIT IV
Social Media Analytics
Affiliation · Identity · Social Capital · Predictive & Descriptive Modeling
Facebook & Google Analytics · Outlier Detection · A/B Testing
12 Comprehensive Topics
UNIT IV — TOPICS OVERVIEW
Affiliation Identity
01 02
Homophily · Group Membership · Social Capital · Structural Balance Social Identity · Role-Based · Online vs Offline · Identity Performance
Social Capital Structural Holes
03 04
Bonding · Bridging · Linking No Structural Hole · With Structural Hole
Predictive Modeling Descriptive Modeling
05 06
Regression · Decision Trees · SVM · Neural Networks · Clustering Data Collection · Metrics · Visualization · Reporting
Facebook Analytics Google Analytics (GA4)
07 08
Intro · Parameters · Demographics · Audience · Reach & Engagement Implementation · Limitations · Performance · Privacy · GWO
Outlier Detection Google Website Optimizer
09 10
Global · Contextual · Collective · Detection Methods A/B Testing · Multivariate · Redirect · Personalization
AFFILIATION & IDENTITY
Affiliation Identity
Connection of individuals to social groups. How individuals define themselves in a network.
Homophily Social Identity Theory
People associate with similar others Categorizing into in-groups and out-groups
Group Membership Role-Based Identity
Joining groups based on shared goals Identity changes based on social role
Social Capital Online vs Offline Identity
Resources gained through connections Multiple identities across platforms
Structural Balance Identity Performance
Adjusting connections for harmony Behavior modified by social context
SOCIAL CAPITAL & STRUCTURAL HOLES
Social Capital
Advantages gained through social connections.
Bonding Bridging Linking
Strong ties among close-knit groups (family, close Weak ties connecting different groups. Enables Ties to people in power or authority. Provides
friends). Creates trust and reciprocity. access to new information and opportunities. access to resources and influence.
Structural Holes
Absent ties giving one actor positional advantage in a network.
No Structural Hole With Structural Hole
All actors are connected to each other. No single actor holds an A tie is absent between two groups. The third actor who bridges the gap
informational advantage. Network is balanced with equal access. gains bargaining power and controls information flow.
PREDICTIVE MODELING
Forecasting future outcomes using machine learning algorithms.
Linear Regression Logistic Regression Decision Trees
Predicts continuous values (e.g., number of likes) Predicts binary outcomes (e.g., ad click: yes/no) Categorizes users into groups via branching rules
SVM Naive Bayes ARIMA / SARIMA
Classification in high-dimensional data spaces Probabilistic classifier; used in sentiment analysis Time series forecasting for trends and seasonality
Neural Networks Clustering Ensemble Methods
Deep learning for complex pattern recognition Groups similar users or content unsupervised Combines multiple models for improved accuracy
DESCRIPTIVE MODELING
Summarizing historical data to understand what happened — and why it happened.
Data Collection
01
Gather data from social media platforms via APIs, web scraping, or platform export tools.
Key Metrics
02
Track engagement rate, reach, impressions, click-through rate (CTR), and conversion rates.
Data Visualization
03
Present insights using charts, graphs, heatmaps, and dashboards for clear communication.
Reporting
04
Summarize KPIs for stakeholders in actionable reports to guide future strategy.
FACEBOOK ANALYTICS
A tool for tracking audience behavior and content performance on Facebook Pages.
Page Views Post Reach ❤ Engagement
👁 📡
Total number of times the page was viewed Number of unique users who saw at least one Interactions including likes, comments,
by visitors in a given period. of the page's posts. shares, and reactions on posts.
CTR Retention LTV
🖱 🔄 💰
Click-Through Rate — percentage of users Rate at which users continue to interact with Lifetime Value — total revenue generated
who clicked a link or call-to-action. the page over time. per user across their lifetime.
GOOGLE ANALYTICS (GA4)
Web analytics tool tracking user behavior across websites, apps, and platforms using an event-based model.
Data Collection
Tracks every user action as an 'event' — page views, scrolls, clicks, and custom events across web and app.
Reports
Pre-built dashboards for acquisition, engagement, monetization, and retention metrics.
Explore
Custom analysis using funnel exploration, path exploration, segment overlap, and user explorer.
Attribution
Identifies which marketing channels and touchpoints drive conversions using different attribution models.
Audiences
Groups users by behavior, demographics, or conditions for targeted re-engagement campaigns.
OUTLIER DETECTION
Finding data objects that deviate significantly from the expected pattern.
TYPES OF OUTLIERS
Global Outlier Contextual Outlier Collective Outlier
Deviates significantly from the entire dataset as a Normal globally but deviates within a specific A group of objects that together deviate from the
whole. context or condition. expected.
DETECTION METHODS
Supervised Unsupervised Semi-Supervised
Labeled data to train a classifier to identify outliers. No labels; identifies points far from cluster centers. Mix of labeled and unlabeled data for detection.
Statistical Proximity-Based Clustering-Based
Flags points with low probability in a distribution. Flags points far from their k nearest neighbors. Points not belonging to any cluster are outliers.
GOOGLE WEBSITE OPTIMIZER & UNIT SUMMARY
Google Website Optimizer (Discontinued)
A/B and multivariate testing tool to optimize website conversion. Now integrated into Google Optimize / GA4.
A/B Testing Multivariate Testing Redirect Test Personalization
Compares two versions of a webpage Tests multiple element combinations Tests two completely different page Permanently implements the winning
to determine which performs better. simultaneously on the same page. URLs against each other. experiment for all users.
UNIT IV — KEY TAKEAWAYS
🔗 Network structure shapes behavior — Affiliation, Identity & Social Capital explain how we connect and what we gain.
🤖 Predictive & Descriptive Modeling enable both forecasting future trends and understanding past patterns in social data.
📊 Platform Analytics (Facebook & GA4) provide the raw data and dashboards needed to measure real-world performance.
🎯 Outlier Detection & A/B Testing help detect anomalies and scientifically optimize digital experiences.
FACEBOOK ANALYTICS — DEEP DIVE
Parameters, Demographics, Page Audience Analysis, Reach and Engagement
Key Parameters Demographics
Page Likes, Followers, Post Reach, Impressions, Engagement Rate, Age groups, Gender distribution, Geographic location, Language,
CTR, Video Views, Story Views, Ad metrics (CPC, CPM, ROAS) Device type (mobile/desktop), Education level, Job title and
relationship status
Analyzing Page Audience Reach and Engagement Analysis
Audience Insights shows fan demographics, page section views, Organic vs Paid Reach, post-level engagement breakdown
actions on page, where fans see posts, and competitor benchmarking (likes/comments/shares), best posting time analysis, content type
via Pages to Watch performance comparison
GOOGLE ANALYTICS — IMPLEMENTATION, LIMITATIONS &
PRIVACY
Implementation Technology Google Website Optimizer
01 [Link] / Google Tag Manager snippet added to every page; event- Introduction and Working (now Google Optimize / integrated into GA4)
based data model in GA4 tracks user interactions automatically
Performance Concerns
02 External JavaScript loading adds page load time; blocking scripts How It Works:
Serves different page variants to user segments, tracks conversion goals,
affect Core Web Vitals; high-traffic sites may hit data sampling
thresholds and statistically determines the best-performing version.
Test Types:
Limitations
03 Cannot track cross-device without login; ad blockers prevent
A/B Test: compare 2 versions of a page. Multivariate: test multiple
element combos. Redirect: test entirely different URLs.
tracking; no 100% real-time data; limited raw access in free tier Outcome:
Winning variant is permanently deployed. Improved conversion rate and
Privacy Issues better user experience with data-driven decisions.
04 GDPR/CCPA requires consent banners; IP anonymization mandated
in EU; Google server storage raises cross-border data transfer
concerns
Community Detection & Anomaly Detection in Social Networks
Community Detection: Identifies tightly-knit clusters (Louvain, Girvan-Newman algorithms). Anomaly Detection: Flags unusual user behavior (sudden spike in posts,
bot-like patterns, coordinated inauthentic behavior) using statistical or ML-based methods.