The Viswanathan Rough-Hawkes (V-RH) Fractional Valuation Model
While traditional valuations rely on static spreadsheet projections, V Viswanathan Associates employs the V-RH Model — utilising Hawkes Processes and Fractional Stochastic Calculus to provide institutional investors with mathematically defensible enterprise valuations for hyper-growth startups with viral network effects.
The Critical Flaw in Current Startup Valuation
VCs routinely attempt to value hyper-growth startups using tools designed for mature, stable businesses. Discounted Cash Flow models assume smooth, predictable cash flows. Revenue multiples ignore unit economics entirely. Even the best Monte Carlo implementations assume that the startup's growth process is Markovian — meaning the market has no memory, and yesterday's events have no mathematical influence on today's growth rate.
This assumption is demonstrably false for startups. A viral product launch creates a self-exciting cascade of new users — each new signup increases the probability of the next. A bad press event or a buggy deployment suppresses growth for months, not days. Standard models cannot capture either of these dynamics.
What Standard Models Cannot Model
The V-RH model solves all four simultaneously by combining two of the most powerful tools in modern probability theory: Hawkes Processes (used by Renaissance Technologies and Two Sigma in high-frequency equity modelling) and Fractional Brownian Motion with rough volatility (the paradigm-shifting discovery of Jim Gatheral et al., 2018).
The Two Mathematical Pillars of the V-RH Model
Pillar 1: Hawkes Processes
A Hawkes Process (Hawkes, 1971) is a self-exciting point process in which the occurrence of each event temporarily increases the rate at which future events occur. Originally used to model earthquake aftershock sequences and adopted by elite quant funds for modelling order-book microstructure, it is the natural mathematical language of viral growth.
In the V-RH context: each new customer acquisition \((dN_t)\) causes an instantaneous jump \(\alpha \cdot dN_t\) in the future acquisition rate \(\lambda_t\). The viral coefficient \(\alpha\) directly quantifies the strength of the network effect: how much one new user accelerates the rate of the next.
Pillar 2: Fractional Brownian Motion (Rough Volatility)
Standard Brownian Motion assumes market volatility is memoryless. Fractional Brownian Motion \((W_t^H)\) introduces a Hurst parameter \(H \in (0,1)\). When \(H = 0.5\), we recover standard Brownian Motion. When \(H < 0.5\), paths become rough and exhibit anti-persistence — capturing the negative autocorrelation of early-stage market friction.
Jim Gatheral, Thibault Jaisson and Mathieu Rosenbaum (2018) demonstrated empirically that equity volatility is rough with \(H \approx 0.1\). For early-stage startups — subject to regulatory friction, founder risk, and market education costs — rough volatility is not a theoretical choice; it is the empirically correct assumption.
The V-RH Valuation Formula
V-RH Enterprise Value — Risk-Neutral Expectation
Viswanathan Rough-Hawkes (V-RH) Enterprise Value under Risk-Neutral Measure \(\mathbb{Q}\)
Customer Acquisition Rate — Stochastic Differential Equation
where \(W_t^H\) is Fractional Brownian Motion with Hurst parameter \(H\) and \(dN_t\) is a Hawkes jump
Risk-Neutral Enterprise Value
The fair enterprise value computed as the risk-neutral expectation \(\mathbb{E}^{\mathbb{Q}}\) of discounted future net customer value over the valuation horizon \([0, T]\). Taking the expectation under the risk-neutral measure \(\mathbb{Q}\) rather than the physical measure \(\mathbb{P}\) ensures the valuation is consistent with market pricing of risk — a requirement for IBBI-defensible fairness opinions.
Instantaneous Net Unit Economics
The integrand is the instantaneous net value created at time \(t\): the current acquisition rate \(\lambda_t\) multiplied by the Lifetime Value per customer \((LTV)\), minus the Customer Acquisition Cost \((CAC)\). This embeds unit economics inside the stochastic valuation rather than applying LTV/CAC as a static post-hoc ratio — a critical structural improvement over standard frameworks.
Mean Reversion — Natural Growth Ceiling
The Ornstein-Uhlenbeck drift term. Startups cannot acquire customers at an exponentially growing rate forever — the addressable market is finite. \(\theta\) is the long-run natural acquisition rate (market baseline), and \(\kappa\) is the speed at which viral growth reverts toward it. A high \(\kappa\) means viral spikes decay quickly; a low \(\kappa\) means viral momentum persists. This term prevents the model from producing unbounded, economically nonsensical valuations.
Hawkes Jump — Viral Coefficient
The self-exciting Hawkes term. When one customer joins (\(dN_t = 1\)), the acquisition rate jumps by \(\alpha\). This is the mathematical definition of a viral loop: each user directly increases the rate of further user acquisition. \(\alpha > 1\) indicates a supercritical regime (true viral explosion); \(\alpha < 1\) is subcritical (viral behaviour but naturally self-limiting). The Hawkes process also captures product launches, press events, and referral campaigns as exogenous excitations.
Fractional Brownian Motion — Rough Volatility & Long Memory
The diffusion term driven by Fractional Brownian Motion with Hurst parameter \(H\). \(\xi\) is the volatility of the acquisition rate. When \(H < 0.5\): growth paths are rough — highly erratic, with negative autocorrelation, reflecting the high friction of early-stage markets. When \(H > 0.5\): growth exhibits positive autocorrelation (momentum). Calibrating \(H\) to the startup's actual MCA-filed growth data provides the forensic CA's most important input: whether the company is operating in a momentum regime or a friction regime.
The Filtration — Forensic CA Audit Foundation
In probability theory, \(\mathcal{F}_t\) denotes the filtration — the complete sigma-algebra of all information known up to time \(t\). In practice, this is the forensic CA audit: the verified MCA filings, GST returns, bank statements, and cohort-level retention data that define the information set from which the model is calibrated. The V-RH model cannot be run on projected data; it requires the verified historical filtration that only an IBBI Registered Valuer with CFE forensic credentials can produce.
V-RH vs. Traditional Valuation Methods
| Capability | DCF | Revenue Multiple | Standard Monte Carlo | V-RH Model |
|---|---|---|---|---|
| Models viral self-excitation | ||||
| Captures long memory / market friction | ||||
| Integrates LTV/CAC into SDE | partial | |||
| Models natural growth ceiling | assumed | assumed | explicit | |
| Risk-neutral pricing under \(\mathbb{Q}\) | WACC proxy | partial | ||
| Requires forensic CA audit (\(\mathcal{F}_t\)) | mandatory |