Why Decentralized Is a Spectrum, Not a Checkbox

The word gets used as a binary property. It is not one. There are at least five separate axes, each independently measurable, and a network can be strong on one and weak on another.

Axis one: who validates

How many independent parties verify the chain, and what does it cost to become one?

Measure: node count, and the hardware and bandwidth needed to run one. A network where validation requires enterprise infrastructure has few independent validators regardless of how many machines exist, because they belong to few operators.

Axis two: who produces blocks

Distinct from validation. A chain can have thousands of verifying nodes and a handful of entities producing nearly all blocks.

Measure: the share of blocks produced by the largest few entities. On proof-of-work chains this is mining pool concentration; on proof-of-stake it is staking provider concentration. Both have historically been high enough to be uncomfortable.

Axis three: who writes the software

How many independent implementations exist, and how concentrated is development?

Measure: client diversity, and the distribution of commits across organisations. A network with one implementation has a single point of failure that no amount of node distribution addresses.

Axis four: who can change the rules

The most consequential and least discussed.

Measure: whether upgrade keys exist, who holds them, whether there is a timelock, and what the governance process actually requires. Many networks marketed as decentralised have a multi-signature wallet controlled by a small team that can change the contract logic.

Axis five: where the assets are

If most of the supply sits with a few custodians, the practical control over the asset is concentrated even if the protocol is not.

Measure: share of supply held by the largest custodial entities.

Why this matters

Because comparisons are usually made on the axis that favours the speaker.

A high-throughput chain will quote node count. A chain with an active development community will quote client diversity. A chain with wide token distribution will quote holder counts. None of these is dishonest and none of them is sufficient.

The composite question worth asking

What would it take for a small group to censor a transaction, reverse a transaction, or change the rules?

Answer that for each network and the differences become concrete. On some, it would require coordinating a majority of globally distributed capital or hash power. On others, it would require three people to sign a transaction.

Both descriptions currently apply to networks describing themselves the same way.

How to check

All five axes are measurable from public data. Node counts and client distribution are published by community trackers. Block production concentration is computable from the chain. Upgrade key arrangements are in the contract code and are documented by independent watchdog sites for major networks.

Axis five is the easiest to check: exchange holdings appear on-chain as a small number of very large addresses, and venues such as a platform with real on-chain withdrawals publish their own reserve figures, which is a primary source rather than a clustering estimate.

It takes an afternoon per network, and it produces a more useful picture than any amount of reading about philosophy.

The conclusion I have reached after doing it repeatedly: almost every network is more concentrated than its advocates believe and less concentrated than its critics claim, and the specific axis on which each is weak differs enough that generalisations are useless.