Pull the same metric from two onchain platforms and the numbers will often differ. Not by a rounding error, sometimes materially. This is not a bug, and knowing why it happens is a prerequisite for using any of this data seriously.
Why the same metric differs
Almost every onchain metric depends on address labelling, and labelling is an estimate. To publish an exchange reserve figure, a provider must decide which addresses belong to which exchange. They do this with clustering heuristics, deposit-address tracing and manual research. Different methodologies produce different wallet sets, and therefore different reserves, different netflows and different whale balances.
Coverage also differs. One provider may track forty exchanges, another twenty-five. A metric summed across a larger set will read higher, and when a provider adds a venue the historical series can shift retroactively. Entity clustering, which groups addresses under one owner, varies in aggressiveness too. Any metric with the word entity in it is a model output rather than a direct measurement.
Glassnode
The deepest catalogue of derived Bitcoin metrics and the origin of much of the standard vocabulary, including SOPR variants, HODL waves and entity-adjusted series. Documentation is thorough, which matters more than it sounds: knowing exactly how a metric is constructed is what lets you tell a real signal from an artefact. The free tier gives a limited selection with delayed and lower-resolution data, enough to follow the widely cited charts but not to build on.
CryptoQuant
Stronger on exchange-level granularity, particularly flows broken out by individual venue and by miner wallets, with good coverage of Asian exchanges that others label less completely. If the question is which exchange something moved to, this is usually the better starting point. The free tier is comparatively generous on exchange flow charts.
Santiment
Different emphasis. Alongside onchain data it tracks social and development activity, which makes it more useful for altcoins and for questions about attention rather than settlement. For pure Bitcoin holder analysis it is the thinnest of the three. For asking whether a token’s developer activity matches its narrative, it is the only one of the three that tries.
The free options worth knowing
mempool.space for anything real time about transactions and fees, with no account needed. Block explorers such as Blockchair for verifying a specific transaction rather than a trend. Blockchain.com charts for long histories of basic series like active addresses and hash rate. For most readers these plus one paid-tier free plan covers everything. The free tools roundup goes through the full list, and the dashboard guide covers assembling them.
How to handle disagreement
When two providers disagree, neither is lying. Pick one source per metric and stay with it, because consistency in the series matters far more than which absolute number is correct. Switching providers mid-analysis introduces a step change that looks exactly like a market event. And treat any single-provider claim about an absolute threshold with care, since the threshold was calibrated on that provider’s labelling.
What to check in the documentation
Before relying on any metric from any provider, four things in the documentation are worth two minutes. Which exchanges are included, because the set determines the level. Whether the series is entity-adjusted, which tells you how much modelling sits between the chain and the chart. How the provider handles the ETF custodians, since treating them as exchange wallets or not changes reserve and flow figures materially. And when the methodology last changed, because a revision creates a step in the history that looks exactly like a market event.
Providers that publish this clearly are more trustworthy than providers with prettier charts. The ones that do not publish it are asking you to take the number on faith, and a number you cannot interrogate is not evidence.
Common questions
Do I need a paid subscription? For following the standard charts, no. For building a system on specific thresholds, yes, because free tiers usually delay and downsample exactly what you would need.
Can I compute these myself from a node? The raw ones yes, the derived ones effectively no. Realized cap and entity-adjusted metrics require years of labelled history that is expensive to reconstruct.
Which should a beginner start with? Whichever free tier covers the five metrics you actually check. The metrics guide can help narrow that down first.
Research and education only, not financial advice. Onchain Decoded has no commercial relationship with the providers named here.
Leave a Reply