Understanding the growth and death of online social networks: evidence and model

Bruno Ribeiro

Abstract

Driven by outstanding success stories of Internet startups such as Facebook and The Huffington Post, recent studies have thoroughly described their growth. These highly visible online success stories, however, overshadow an untold number of similar ventures that fail. The study of website popularity is ultimately incomplete without general mechanisms that can describe both successes and failures. In this work we present six years of the daily number of users (DAU) of twenty-two membership-based websites—encompassing online social networks, grassroots movements, online forums, and membership-only Internet stores—well balanced between successes and failures. We then propose a combination of reaction-diffusion-decay processes whose resulting equations seem not only to describe well the observed DAU time series but also provide means to roughly predict their evolution. This model allows an approximate automatic DAU-based classification of websites into self-sustainable vs. unsustainable and whether the startup growth is mostly driven by marketing & media campaigns or word-of-mouth adoptions.

Introduction: Beyond Survivorship Bias

In the startup ecosystem, analysis is frequently hampered by "survivorship bias," where researchers focus on giants like Facebook while neglecting the thousands of platforms that fail to gain traction. Furthermore, industry metrics often rely on total membership—a "vanity metric" that fails to account for inactive accounts. This paper argues that the true pulse of a website is its Daily Active Users (DAU). We address the lack of a unified model capable of describing the entire DAU trajectory, from initial viral explosion to eventual stabilization or decline, across diverse categories like social networks, forums, and online stores.

Methodology: A Reaction-Diffusion-Decay Framework

The proposed model utilizes Ordinary Differential Equations (ODEs) to treat users as particles in a chemical reaction-diffusion-decay process:

  1. Reaction (Word-of-Mouth): Active users ($A$) interact with inactive members ($I$) or non-members ($U$) to recruit them, representing network effects ($A + I \to 2A$).
  2. Decay (Attention Loss): Active users spontaneously become inactive over time ($A \to I$), capturing the natural loss of interest.
  3. Diffusion (Marketing): External forces, such as media campaigns or SEO, "diffuse" non-members into the active user pool ($U \to A$).

To fit this model to real-world data, the Levenberg-Marquardt algorithm was employed, supplemented by k-medoids clustering to identify representative parameter sets across different platforms.

Empirical Dataset

The study analyzes a high-resolution dataset from Alexa.com spanning June 2007 to June 2013. The data includes over 55,000 observations across 22 membership-based websites, including successes (e.g., Facebook, LinkedIn, Netflix) and failures (e.g., 12seconds.tv, Formspring). To account for general Internet growth, DAU is measured as a fraction of the total active Internet population (%DAU).

Key Insights and Classification

The Sustainability Tipping Point

A critical finding is the ratio $\beta/\alpha$, where $\beta$ represents the decay rate (attention loss) and $\alpha$ represents the reaction rate (user attraction). The model demonstrates that if $\beta/\alpha \ge 1$, the website is mathematically destined to decline, regardless of marketing expenditure. This allows for the early identification of unsustainable platforms.

Classification of websites into self-sustaining and unsustainable regions based on the ratio of attention decay (beta) to attraction rate (alpha).

Marketing vs. Virality Signatures

The model distinguishes between growth driven by "word-of-mouth" (characterized by an exponential ramp-up) and growth driven by "marketing" (characterized by a convex curve). By framing attention as a finite resource, the model explains why even successful sites eventually reach a plateau rather than growing indefinitely.

Mapping websites by their growth drivers: media and marketing intensive (lambda) vs. word-of-mouth intensive (gamma) membership growth.

Results and Predictive Power

The framework demonstrated significant predictive accuracy in holdout tests. For instance, using early-stage data, the model correctly predicted that Facebook would stabilize at approximately 45% of the active Internet population. It also successfully categorized websites into "self-sustaining" (e.g., The Huffington Post) and "unsustainable" (e.g., 12seconds.tv), identifying the death trajectories of sites like Flixster and True.com long before they reached their nadir.

Limitations and Future Directions

While robust, the model assumes constant underlying dynamics and may struggle with "black swan" events, such as major site redesigns or the sudden emergence of dominant competitors. Additionally, the Alexa dataset primarily reflects desktop traffic; future work should incorporate mobile app data to reflect modern usage patterns. Finally, as a "mean-field" model, it does not account for specific social graph topologies, which could offer more granular predictive capabilities.

Summary updated

Abstract

Driven by outstanding success stories of Internet startups such as Facebook and The Huffington Post, recent studies have thoroughly described their growth. These highly visible online success stories, however, overshadow an untold number of similar ventures that fail. The study of website popularity is ultimately incomplete without general mechanisms that can describe both successes and failures. In this work we present six years of the daily number of users (DAU) of twenty-two membership-based websites - encompassing online social networks, grassroots movements, online forums, and membership-only Internet stores - well balanced between successes and failures. We then propose a combination of reaction-diffusion-decay processes whose resulting equations seem not only to describe well the observed DAU time series but also provide means to roughly predict their evolution. This model allows an approximate automatic DAU-based classification of websites into self-sustainable v.s. unsustainable and whether the startup growth is mostly driven by marketing & media campaigns or word-of-mouth adoptions.
BibTeX
@inproceedings{ribeiro2013understanding,
  author = {Bruno Ribeiro},
  title = {{Understanding the growth and death of online social networks: evidence and model}},
  booktitle = {The Web Conference},
  year = {2013},
}
Figure