Features
Part of Guide to creator net worth: what is genuinely checkable
Sorting out creator net worth methodology: the chain, link by link
Creator net worth methodology traced through its conversion chain: six multiplications between a public view count and a person, each one a guess.
Every published creator figure is the output of a chain of multiplications that starts with a public number and ends with a private one. The chain is short enough to write down. Once it is written down, the reason the output means nothing becomes arithmetic rather than opinion.
What to take away
- The method multiplies one known number by a series of invented ratios.
- Multiplicative error does not average out; it widens the range at every step.
- The one input that is measured is the one that matters least.
The chain, link by link
Here is the full sequence, in the order an estimator walks it.
One measurement, six assumptions
- Public count: views, subscribers, plays
- Multiply by assumed rate per unit
- Multiply by assumed platform share
- Add assumed brand-deal component
- Multiply by assumed margin
- Multiply by assumed retention rate
- Compound over assumed years and return
One measurement, six assumptions. And they are chained rather than added, which is the technical heart of the problem.
Why chaining is worse than it sounds
If you add uncertain quantities, some errors cancel; if you multiply them, they do not.
A method generous at step two and again at step five does not give a slightly high answer. It is high by the product of both mistakes. By step seven, the spread between a pessimistic and an optimistic pass through the same chain can cover a range so wide the midpoint carries no information.
This is not a controversial point in measurement. It is why serious work reports a value with its uncertainty attached rather than a bare number, and why the National Institute of Standards and Technology publishes guidance on expressing measurement uncertainty. A creator estimate published as a single value with no interval is claiming a precision that its own method forbids.
The steps ranked by how much damage they do
| Step | How wrong it can be | Effect on the output |
|---|---|---|
| The public count | Small; it is measured, though it can be inflated or bought | Minor |
| Rate per unit | Large; real rates vary by a wide multiple across countries and categories | Major |
| Platform share | Moderate; the rules are published even if the applicable terms are not | Moderate |
| Brand-deal component | Very large; both the count and the fee are invented | Major |
| Margin | Very large; a team-run channel and a one-person channel differ completely | Major |
| Retention | Unbounded; there is no evidence of anyone's spending | Decisive |
| Return and years | Unbounded; compounding magnifies the assumption | Decisive |
The first row is the only one with a document behind it, and it is the row that barely matters. Everything that decides the answer is in rows four through seven, where the method has nothing at all.
Damage by step
Step
- Public count
- Small, measured
- Rate per unit
- Large
- Platform share
- Moderate
- Brand-deal component
- Very large
- Margin
- Very large
- Retention
- Unbounded
- Return and years
- Unbounded
How wrong
- Public count
- Minor
- Rate per unit
- Major
- Platform share
- Moderate
- Brand-deal component
- Major
- Margin
- Major
- Retention
- Decisive
- Return and years
- Decisive
Effect
- Public count
- Rate per unit
- Platform share
- Brand-deal component
- Margin
- Retention
- Return and years
What would have to change
A defensible method would publish, per entry, the rate and its source, the platform terms believed to apply, and the number of commercial posts and how they were counted. It would also give the fee assumed and why, a production cost estimate, and a stated interval.
Nobody publishes this, and not from laziness. Publishing it would show the interval is wider than the estimate.
Disclosure obligations do give a researcher one honest foothold. Commercial relationships have to be made clear to an audience, and the Federal Trade Commission's answers on the endorsement guides explain what that covers. Counting disclosed posts is legitimate. Attaching a price to each one is not.
The alternative that works
Ask a question the chain need not answer.
How does platform revenue sharing work as a mechanism, what does a brand actually buy in a sponsorship, and what does owning a copyright let a creator do that a license does not? Which of a creator's revenue lines survive a policy change? Each of those has an answer that does not require knowing anyone's balance.
The structure of the revenue lines themselves is on creator net worth. Where the same reasoning goes in other trades is on estimate methodology. The commercial side, where the fees are equally private, is on endorsement income, and why the outputs then get sorted into an order is on richest rankings.
Common questions
Could the chain work if the first number were bigger and better?
No, and this is the counterintuitive part. Improving the measured input does nothing, because the measured input is not where the error lives. Perfect view data multiplied by six guesses is still six guesses.
Some estimators publish a range rather than a point. Is that better?
Much better, if the range is derived rather than decorative. A range built by running the chain at pessimistic and optimistic settings is honest work. A range that is the point estimate plus or minus a tidy percentage is a point estimate with a decoration.
Why do these figures cluster on round numbers?
Because they are the product of round assumptions. A method with seven guessed inputs cannot produce anything else, and a real balance never lands there.
