On-chain data has an unusual property for financial information: it is complete, public, and verifiable by anyone. That is genuinely valuable, and it is also the reason it is so frequently misread. Data that is precise feels authoritative, and authority is easy to extend past where the data actually reaches.
Addresses are not people
This is the single most consequential caveat. An address is a destination, not an identity. One person can control thousands; one address can be shared by millions of customers of a single service.
Every metric built on counting addresses inherits this problem. “Active addresses” is a real measurement of something, but it is not a headcount, and it moves when a large service changes how it batches or rotates addresses — which has nothing to do with adoption.
Volume includes movement that is not economic
The total value transferred on a chain includes internal transfers, consolidation, change returning to its sender, and automated activity that no human initiated. None of that is fraudulent; it is just not the thing most people think they are measuring.
Adjusted volume estimates exist and are more useful, but each involves judgement about what to exclude. Two providers making different reasonable choices will produce different numbers, and neither is lying. If a figure matters to you, find out what it excludes before relying on it.
Labels are inference
Dashboards that attribute addresses to exchanges, funds or categories are doing informed detective work, not reading a register. Clusters are inferred from behaviour, and inference has an error rate that is rarely published alongside the chart.
Treat labelled flows as a hypothesis with evidence behind it rather than as a fact. They are often right and occasionally confidently wrong, which is the more dangerous combination.
What on-chain data is genuinely good for
It is excellent for anything that is a direct property of the protocol: supply, issuance, block production, fee levels, and the mechanical facts of what the network did. Those are not estimates. It is also good for detecting that something unusual happened, even when it cannot tell you why.
It is weakest exactly where people most want to use it — inferring intent, predicting price, or counting participants. The chain records what happened. It does not record why, and no amount of resolution turns one into the other.
A workable discipline
Prefer protocol-level facts over derived aggregates. When you use a derived metric, learn its construction. Be suspicious of any single number offered without a definition, and be more suspicious when it confirms what you already believed.
Our methodology page sets out where the data on this site comes from and how each figure is calculated. For foundations, see what a blockchain is. Nothing here is financial advice.