Unbundling the Hype (Cycle)
Despite being an idealization of reality, the Gartner Hype Cycle is intuitive and digestible, practical and ubiquitous. At times, it is possible to unpack our intuition and shed light on a latent processes which describe some phenomena.
“Whenever we look at life, we look at networks.”― Fritjof Capra
As venture investors, we review founders and products in the foreground, while, in the background, we estimate market breadth and timing. For deep technology and infrastructure, assessing timetables can be a treacherous exercise, distorted by social and economic influences under which decisions are made. The yin-yang of estimation and influence catalyzes human biases like herding and FOMO. And when it comes to our technological future, we tend to be overly optimistic.
In my last company, we explored complex network theory and social influence. We concluded that human opinion formation, especially in the face of high uncertainty, is inherently socialized —that belief in a network is embedded in the network of social relationships (we filed a similar patent in this area). In decision making under uncertainty, information cascades propagate with more fluidity, and we discard endogenous conclusions. This process folds on itself and reinforces social biases— especially when economics and financial rewards are at risk.
Value Estimates vs. Value Realization
Late one night, I asked myself the question: “if we are to believe that the Hype Cycle is a good representative model for investment and innovation cycles, how could we unpack it?”
I came up with the following: we can think of the hype cycle as a superposition of two regimes — a value estimation regime and a value realization regime.
During the value estimation phase, the market is determining the potential future value of a particular investable theme. Given the level of uncertainty, the value estimation phase is noisy and is subject to behavioral effects and social dynamics (i.e. herding, overestimation, regret bias). The belief outputs of the value estimation phase are then expressed as capital (investments into companies).
Breaking it Down
It is known from academic literature that complex communication patterns in humans may possibly be expressed with the Log Normal distribution. For instance, viral spread can be explained by the log-normally distributed observation and reaction times of individuals. In his seminal work on modeling adoption cascades, Duncan Watts modeled the distribution of adoption thresholds (fraction of exposed neighbors that triggers adoption) using a Log Normal distribution. Viral meme propagation has also been modeled with a Log Normal distribution:
Where μ expresses the quickness of the viral spike, sigma represents the tail effects (how long it lasts), a represents the asymptotic estimate of full adoption (i.e. % of total addressable market), b is the lag of the adoption, and c is the growth rate of adoption.
When we “estimate”, in an investment setting, we are estimating the value of a technology (as a function of time). The function which are are attempting to estimate is that of behavioral adoption (consumer and/or enterprise). The value of a technology is a function of its adoption or penetration in a network.
It is known from academic literature, that the cumulative sum of the adoption curve can be expressed with the family of sigmoid functions, of which Gompertz function is one:
Where a is the asymptote (asymptotic population size), b is the displacement on the x-axis, and c is the growth rate.
The intersection of these two functions represent the shape of the Hype Cycle.
As rational investors, we should ideally stage capital deployment with proper adoption estimates, while also de-risking R&D as we reach incremental milestones. The Log Normally-distributed value estimation phase for thematic technology systematically overly-estimates adoption curves.
Capital deployment is often driven by social propagation and consensus dynamics, and a large supply of capital chasing a limited supply of assets (i.e. investable opportunities). This of course drives up valuations and subsequently more capital deployment.
Sometimes it is better to wait (Crypto…).
