Bitcoin Transaction Volume: What On-Chain Volume Actually Measures

There is a genre of crypto headline that runs "Bitcoin settled $200 billion on-chain last month." The number is real in the sense that somebody added it up correctly. It just does not mean what the sentence implies, and the gap is not small. Depending on the period you pick it is somewhere between three times and seventeen times too big.

Rather than repeat the usual explanation, we took a single ordinary block and measured it.

What we measured, and how

On 15 August 2026 we pulled every transaction in block 962,547, which was mined at 08:46:28 UTC and held 6,054 transactions. Not a sample: the full block, paged 25 transactions at a time out of a public Esplora index, with the previous output attached to every input so we could see which address funded each spend.

Then we applied the most conservative change rule there is. If an output pays an address that also appears as an input to the same transaction, that output is the sender paying themselves. No heuristic, no guessing, no clustering. The same address on both sides.

So the headline overstated the block by a factor of seventeen. Before you generalize that, read the next section, because the reason is more interesting than the ratio.

One transaction did most of it

The 3,036 BTC of same-address payback is not spread evenly across thousands of wallets. A single Taproot address accounts for 81 percent of it, and the top five addresses account for 95 percent. Here is the transaction, in full, so you can check it:

An address swept five dust outputs into a stack it already controlled, wrote a small OP_RETURN, and sent the whole balance straight back to itself. At the price mempool.space reported that morning, 62,959 US dollars, the transaction registered roughly 154.9 million dollars of "value transferred." The value that actually moved between two different parties was about 140 dollars of dust, and the network was paid 38 cents to do it.

That one transaction is 76 percent of the entire block's raw volume. This is the part that matters more than the seventeen times ratio: daily and hourly on-chain volume figures are hostage to a handful of large self-transfers. A custodian rotating storage on a Tuesday can add nine figures to a chart without a single coin changing owner.

Why is Bitcoin transaction volume calculated this way?

Because of how Bitcoin actually works. Bitcoin has no account balances. It has unspent outputs, and an output must be spent whole. If you hold a single 2 BTC output and want to pay 0.3 BTC, your wallet spends the entire 2 BTC and creates two new outputs: 0.3 to the recipient and roughly 1.7 back to you as change. The chain records 2 BTC of outputs. One person received 0.3.

Every explorer, every indexer and every "total value transferred" chart is summing outputs, because outputs are what the chain contains. Change is not labeled. There is no flag in the protocol that says "this one is mine coming back to me." Working out which output was change is an inference someone has to make after the fact, and that inference is exactly where honest providers differ. The same ambiguity is what makes address clustering a research problem rather than a lookup.

Why is adjusted volume lower than raw volume?

Adjusted volume removes the change. It is lower because it is trying to answer a different and more useful question: how much value actually changed hands, rather than how much value was rearranged. Providers use a few methods, roughly in order of confidence:

• Same-address payback. An output pays an address that funded the transaction. This is unambiguous and it is what we used above. It is also the weakest filter, because most modern wallets deliberately send change to a fresh address to protect privacy.
• Script type matching. In a two-output transaction, if exactly one output uses the same script type as the inputs, it is likely the change. We ran this too, and on block 962,547 it removed only a further 6.28 BTC, 0.19 percent of raw. Small, in this block.
• Entity clustering. Group addresses that appear to share an owner, then treat any transfer inside a cluster as internal. This catches the most and is the least certain, since CoinJoin and PayJoin are designed to break the assumption it rests on.

Only 5,752 of the 6,053 non-coinbase transactions in this block had a single input, and 90 percent had exactly two outputs, which is the textbook payment-plus-change shape. That is a normal-looking block. Its volume figure is still nearly all noise.

What is the difference between on-chain volume and trading volume?

They measure different things and people conflate them constantly. On-chain volume is value moved between addresses in blocks, settled by the network itself. Trading volume is buying and selling reported by exchanges, and the overwhelming majority of it never touches the chain at all, because it nets out inside an exchange's own ledger. An exchange can report a busy day with almost no on-chain footprint, and the chain can show a huge day driven entirely by one custodian reshuffling cold storage.

The confusion is baked into the search results. Type "bitcoin transaction volume api" into a search box and most of what comes back is price-and-volume charting, which is the trading number. If you need the settled, on-chain figure, say on-chain and check what the provider is actually summing.

So how wrong is the headline, in general?

Our seventeen times is one block, and it is skewed by one whale. Do not quote it as a long-run figure. Over long periods the published estimates converge on something much tamer. Fidelity Digital Assets, working from Coin Metrics data, put Bitcoin's cumulative adjusted transfer value at roughly 2.2 trillion dollars against about 7.5 trillion unadjusted, a factor of about 3.3. Glassnode's entity-adjusted work landed in the same neighborhood, concluding that under 25 percent of recorded on-chain volume represents real transfers between participants.

Three to four times over years, seventeen times in one block. Both are true, and the spread between them is the actual lesson. The shorter your window, the more a single self-transfer distorts it, which makes raw daily volume close to useless as a day-to-day indicator and makes any adjusted series worth the extra work.

Is Bitcoin transaction volume a good indicator?

Raw volume on a short window: no. It is dominated by internal shuffling you cannot see in the number. Adjusted volume over a longer window: reasonable, as one input among several, provided you know which adjustment was applied. What usually beats both is a metric that is harder to fake by moving your own coins. Exchange inflows and outflows tell you about intent because they involve a labeled counterparty. Coin days destroyed weights each spend by how long those coins sat still, so a custodian churning recent UTXOs barely registers while genuinely old supply waking up shows up loudly, which is why it pairs well with dormant wallet data.

None of this is unique to crypto. A headline number is only worth what its definition is worth, and if you cannot see the rows underneath it you are trusting whoever built the chart. It is the same reason nobody sensible buys a business off the seller's dashboard without metrics somebody has actually verified. On-chain data is unusual in that the underlying rows are genuinely public, so there is no excuse for not checking.

How to get the number you actually want

If you are computing this yourself from a public explorer API, the work is what we did above: page the block, attach previous outputs to every input so you can see funding addresses, then apply your change rule and document which one you used. It is roughly 120 paged requests per block, and at Bitcoin's block rate that is a background job rather than something you do on request.

If the chain is already indexed, it is a query. On on-chain metrics here, daily raw volume, adjusted volume and transaction counts are a dated series you pull over REST, and because the same indexed chain is available in SQL you can drill from any day straight to the transactions behind the figure, including the one whale transfer that moved it. You can also reproduce our block measurement directly against UTXO-level data, or pull individual transactions with their previous outputs already attached through the transaction API.

Whichever provider you use, the question to ask is not "do you have volume." Everyone has volume. Ask which outputs they exclude and by what rule, and whether you can see the transactions behind a day. If you are comparing Bitcoin-focused options, our TimechainIndex alternative page covers how that plays out in practice against another Bitcoin-only platform.

Every figure in this article was measured on 15 August 2026 from block 962,547, hash 000000000000000000021b1d45d411554e9e7d9078e982f2835a7a42fcc440eb, and the transaction ID is printed above so you can verify the whole thing against any explorer. This is informational on-chain data only, not financial, investment or trading advice.