Pareto to Creato
What if Hit Rates in the Creator Economy are a Harbinger for All Media?
I recently saw a LinkedIn post about YouTube becoming the biggest media company in the world by revenue, surpassing Disney. I was struck not by the post, but by one of the comments, which I’ll paraphrase:
YouTube became the largest media company by only paying a tiny percentage of its creators.
This statement is factually correct. A tiny percentage of creators make money on YouTube…and Spotify, Twitch, and Roblox, for that matter. But the implicit accusation—that this results from YouTube’s greed or unfairness—fundamentally misunderstands why the distribution of success in the creator economy is so wildly skewed.
Understanding why these power laws form and why they matter is critically important. As GenAI evolves, the power laws in the creator economy may be a harbinger for all media.
Tl;dr:
Power law distributions in media are becoming more extreme as more content is consumed on networks and the quantity of content increases.
Corporate media distributions approximate the 20/80 Pareto Principle: ~10% of titles represent ~80% of consumption. But creator media distributions are closer to <1/99: less than 1% of titles represent 99% of consumption or revenue.
These distributions matter. The more extreme, the higher the risk; the higher the risk, the worse the returns; and the more extreme, the more value shifts to top talent.
Corporate media and creator media distributions are different because consumers don’t consider them fungible. But as GenAI lowers the cost of high production values and the consumer definition of quality continues to evolve, this may change.
In other words, GenAI may not only change the supply/demand dynamic, it will probably change the distribution of success, too.
Incumbent media companies need to prepare for even more extreme distributions: fewer, bigger hits and more risk.
In the language of finance, this means operating like PE in the head; operating like VC in the tail; and walking away from the middle.
A Quick Refresher on Power Laws
I’ve written before about the increasing prevalence of power laws in media—a very narrow head of a few hits, a skinny middle, and a nearly infinite tail. (See Power Laws in Culture and Power Laws in Culture, Revisited.)
As a quick refresher, to get the basic idea behind power laws, it’s helpful to compare them with normal distributions.
Normal distributions. We all know the bell curve. Most observations cluster around an average and very few fall very far away from it. This is common in domains in which the observations are independent of each other and the range of outcomes is constrained. Height, test scores, salaries of municipal workers, rolls of the dice, etc. Your score on the chem test is independent of the score of that kid on the other side of the class.1
Power law distributions. Power law distributions are common in domains in which the observations are dependent on each other. They exhibit positive feedback loops—strong signals cascade throughout the system and gather steam.2 As a result, they are characterized by a few very large observations and many, many small observations. Power laws (or, strictly speaking, power-law like distributions) show up in a lot of places, especially complex systems: the magnitude of earthquakes, the occurrence of words in any given publication (called Zipf’s Law), the population of cities. For example: People are drawn to cities for economic, social, and cultural opportunities, so the number of people who live in a city is dependent on the number of other people who live in that city.
Figure 1. Normal Distributions and Power Laws
Source: The Mediator.
Normal distributions are easier to grasp intuitively and they’re more psychologically pleasing. Results are bounded and everything hangs out near the average, so they satisfy our innate desire for fairness. Power laws are defined by both massive and tiny observations, which are harder to get your head around. And as implied by the LinkedIn comment I mentioned at the beginning, they most definitely don’t feel “fair.”
Why Power Laws Form in Media
Power laws have become more common in media because people’s content choices aren’t independent. They’re increasingly influenced by other people’s content choices. This has happened for two reasons.
Power law-like distributions have become more prevalent in media because more content is consumed on networks and the more content choices, the more consumers rely on the signals coming from the network.
First, today most media is consumed on networks. Historically, media was one-way: a relatively linear chain from producer to distributor to consumer, in which consumers had limited feedback mechanisms. Today, consumers influence each other both directly (by liking, commenting, and sharing) and indirectly (by influencing recommendation algorithms, what’s trending, etc.).
Second, as the amount of content has increased, people have become more reliant on these network signals. The more choices, the harder it is to determine quality yourself (search costs) and the greater the consequences of making a bad choice (opportunity costs). The higher these costs, the more likely consumers are to rely on external signals, filters, or heuristics to choose their next movie, show, song, book, or game. These include:
Recommendation algorithms or algorithmic feeds
Recognizable brands (“a Pixar movie,” “a Random House book,” “a Rockstar game”)
Recognizable IP (“The Wizarding World of Harry Potter,” “MCU”)
Price (price is assumed to signal quality)
Effective marketing
Professional and crowd-sourced reviews
IRL word-of-mouth (“what are you guys watching these days?”)
Popularity (i.e., “what’s trending” or viral, whether in a large group or smaller communities)
Popularity is the most important factor because it influences all the others.
Of these, popularity is the most important, because it is a meta-filter that influences many of the others:
Recommendation engines typically rely heavily on collaborative filtering, i.e., what is popular among other people consuming similar stuff.
Recognizable brands and IP are recognizable because they’re popular.
Today, a big part of effective marketing is amplifying organic popularity.
Word-of-mouth is a form of popularity.
And reviews are partially influenced by what’s popular.
So, popularity—what other people choose—represents an increasingly powerful positive feedback loop. Popularity begets more popularity.
Pareto vs. Creato
In Power Laws in Culture, Revisited, I showed that while power law-like distributions are ubiquitous in media, they are much more extreme in the creator economy than the traditional or corporate media economy.
