Abstract
The rapid development of the internet has led to the rise of multinational internet unicorn companies, attracting significant attention from both domestic and international markets. Identifying an appropriate valuation method for these entities is a critical challenge, as traditional methods often fail due to short operational histories, limited financial disclosures, and the dominance of intangible assets. This study applies an improved Discounted Equity Value Analysis (DEVA) valuation model, incorporating market penetration rates and industry-related coefficients. Using ByteDance as a representative case, the study estimates its 2023 valuation at $225.3 billion, closely aligning with market buyback valuations.
Introduction
Valuing "unicorns"—startups valued at over $1 billion—is notoriously difficult. Traditional financial metrics like Price-to-Earnings (P/E) ratios require stable profitability, which high-growth startups often lack. Furthermore, the standard Discounted Cash Flow (DCF) model relies on predictable future earnings, whereas unicorns frequently reinvest aggressively, leading to volatile or negative cash flows. Most of a tech unicorn's value resides in its user base, brand, and proprietary algorithms—intangible assets that are poorly captured by cost-based accounting. This research seeks a model that prioritizes "user value" over "book value" to better reflect the reality of the modern internet economy.
Methodology: The Improved DEVA Model
The proposed approach builds upon the Discounted Equity Value Analysis (DEVA) model, rooted in Metcalfe’s Law, which posits that a network's value is proportional to the square of its users. However, this study introduces two critical refinements to address the law of diminishing marginal returns:
- Zipf’s Law Adjustment: Rather than assuming every new user adds equal value, the model applies Zipf’s Law. This assumes that as a network grows, the additional utility provided by the $n$-th user decreases, leading to a logarithmic growth curve ($\ln(n)$).
- Correction Factors ($P$ and $I$): To ground theoretical network value in market reality, the model introduces:
- $P$ (Market Penetration): Adjusts for narrowing growth potential as a company saturates its market.
- $I$ (Industry-Related Coefficient): A calibration constant derived by comparing the subject to a mature, publicly-traded peer.
The final modified formula is:
$E = M \times (ARPU \times MAU) \times \ln(ARPU \times MAU) \times P \times I$
Where $M$ represents the initial capital invested per user, $ARPU$ is average revenue per user, and $MAU$ is monthly active users.
Case Study: ByteDance
ByteDance, the parent company of TikTok and Douyin, serves as the primary research subject due to its global scale and relatively comprehensive disclosure.
Datasets and Parameters
The study utilizes 2023 financial reports and market analysis:
- ByteDance Data: Total revenue of $120 billion and 1.8 billion Monthly Active Users (MAU).
- Meta (Facebook) Data: Used as the benchmark to solve for the industry coefficient ($I = 0.283$), given its similar "attention economy" advertising model.
- Global Ad Market Data: Sourced from MAGNA forecasts to calculate market penetration rates.
A key parameter, $M$ (initial capital per user), is calculated by dividing the total capital raised in all financing rounds by the total number of users. This treats venture capital as the "cost" of building the network.
Results and Discussion
The improved DEVA model estimated ByteDance’s valuation at $225.3 billion USD**. When compared against the $268 billion buyback valuation reported by Sina Finance, the result shows a 16% discrepancy**. This margin is considered acceptable within the high-volatility context of unicorn valuations, where a 20% error is often the benchmark for accuracy.
The inclusion of the natural log ($\ln$) was a vital adjustment, acknowledging that while a network of 1 billion users is vastly more valuable than one of 1 million, the marginal value of the billionth user is less than that of the hundredth.
Conclusion
While the model provides a new perspective, several limitations remain. Financing data for ByteDance was only available up to 2018, potentially leading to an undervalued $M$ parameter. Additionally, the industry coefficient was derived solely from Meta; a basket of peers would be more robust. Future iterations should incorporate "user lifecycle" and "brand value" to capture long-term retention rather than just current active user snapshots.
Summary updated