TL;DR
This paper provides a comprehensive, year-long analysis of Google+'s evolution, revealing that while G+ experienced rapid user growth, its core "connected component" shrunk relative to the total user base. The study highlights that G+ users are significantly less active than those on Facebook and Twitter, suggesting G+ is primarily used for broadcasting rather than deep interaction.
What Problem Does This Paper Address?
The social media landscape is dominated by giants like Facebook and Twitter. Google's entry with Google+ (G+) sparked debate: was it a genuine competitor experiencing organic growth, or a "ghost town" propped up by Google's ecosystem? This paper aims to settle this by dissecting G+'s first year, specifically asking: can G+ attract and retain engaged users, and how does its growth and user behavior compare to established platforms? The core challenge is understanding if G+ is a vibrant social network or merely a passive broadcasting platform, especially given users' increased sophistication with online social networks.
Methods
The researchers employed a large-scale measurement approach, collecting data over a year to track G+'s evolution. Key methods include:
- Breadth-First Search (BFS) Crawling: To capture the connectivity structure of G+'s Largest Connected Component (LCC), they used high-degree users as seeds and systematically crawled the network.
- Efficient ID Space Partitioning: To speed up crawling, they divided G+'s vast user ID space into zones and assigned dedicated crawlers to each, enabling parallel data collection.
- Random User Sampling: Due to the sparse nature of G+'s ID space, they developed a novel method using the G+ API's search function with carefully selected, low-popularity American surnames to obtain representative random user samples.
- Activity Data Collection: They leveraged the G+ API to collect all public posts and associated reactions (plusones, comments, reshares) for LCC users over a 15-month period.
- Comparative Analysis: They compared G+'s metrics with existing datasets from Twitter and Facebook to contextualize their findings.
Datasets
The study utilized several datasets:
- G+ LCC Snapshots: Multiple snapshots of the Largest Connected Component of Google+ were collected at different times (e.g., LCC-Dec, LCC-Apr, LCC-Aug, LCC-Sep, LCC-Oct, LCC-Nov). These provided insights into the network's structure and growth.
- G+ Random Datasets: Samples of random G+ users were collected (e.g., Rand-Apr, Rand-Oct, Rand-Nov) to analyze user characteristics and component distribution.
- G+ Activity Data: A comprehensive dataset of public posts and reactions from LCC users over 15 months.
- Comparative Datasets: Existing datasets for Twitter (TW-Pro, TW-Con, TW-Act) and newly collected datasets for Facebook (FB-Pro, FB-Con, FB-Act) were used for comparison.
Key Insights and Ideas
- The "Singleton" Surge: G+'s rapid growth is largely driven by "singletons" – users who are part of the network but not connected to the main component. This is attributed to Google's integrated registration process, where G+ accounts are automatically created for users of other Google services, many of whom may not actively use G+.
- LCC Shrinking Paradox: Despite the overall user base growing dramatically, the relative size of the Largest Connected Component (LCC) has been decreasing. This means the core, interconnected part of G+ is becoming a smaller fraction of the whole.
- Activity Skewness: A tiny fraction of G+ users are responsible for the vast majority of posts and, even more so, for receiving reactions. This suggests a highly centralized activity pattern, with a few "super-users" dominating engagement.
- Broadcasting, Not Bantering: G+'s connectivity features (like unidirectional follows and less clustered networks) are strikingly similar to Twitter's, and very different from Facebook's. This points towards G+ being primarily used for broadcasting information rather than fostering deep, reciprocal user interactions.
- User Engagement Lag: The rate of new active users joining G+ is significantly slower than the rate of new users joining the LCC, indicating that simply joining G+ doesn't translate to active participation.
Key Results
- LCC Growth: The LCC grew at an impressive rate of 150-200K users per day.
- LCC Relative Size: The LCC's relative size dropped from 43% in April 2012 to 32% by November 2012, while singletons grew from 55% to 66%.
- User Activity: Less than 9% of G+ users generated public posts, and less than 25% provided public attributes.
- Activity Concentration: The top 10% of users contributed 80% of posts, and the top 1% received 80-90% of comments and reshares.
- Activity Rate Comparison: G+ users exhibited a significantly lower typical activity rate (0.08 posts/day) compared to Facebook (0.19 posts/day) and Twitter.
- Connectivity Maturity: G+'s connectivity features, like degree distribution and bidirectional relationships, have stabilized in recent months, suggesting the network has reached a level of maturity.
Limitations and Future Work
- Private Data: The study acknowledges the inability to analyze private posts due to obvious technical and ethical reasons, focusing solely on publicly visible activity.
- Implicitly Added Users: The significant number of singletons is linked to Google's integrated registration, which implicitly adds users without their explicit intent or awareness, making it hard to gauge true user engagement.
- Future Work: The authors suggest that while G+ can attract a large user base, the challenge lies in meaningfully engaging them. Future research could explore how to foster deeper user interaction and what incentives might encourage more active participation. The paper also hints at the need for further investigation into the "broadcasting" nature of G+ and its implications.
Summary updated