Three trackers, three answers about what bitcoin ETFs did on Tuesday
SoSoValue and Farside both put 8 September at a $46.6 million net outflow, while Lookonchain published a $14.4 million net inflow for the same session.
Three trackers looked at the same funds on the same day and published three different answers.
US spot bitcoin ETFs took $46.6 million of net outflows on 8 September 2026, according to SoSoValue, ending three straight sessions of inflows. Farside Investors puts the same day at the same figure, to the decimal. Lookonchain published $14.42 million going in.
One of those is wrong. Possibly two.
The gap is about $61 million and, more awkwardly, a sign. Ether tells the same story from the other end. SoSoValue had ether funds shedding $24.3 million that day while Fidelity's FETH pulled money in. Lookonchain had ether ETFs adding 3,844 ETH. Same funds, same Tuesday, opposite directions.
"Flows" is not one number, which is the whole explanation
Farside and SoSoValue publish creations and redemptions as the funds report them, which is presumably why they agree to a rounding error. They are almost certainly reading the same filings (agreement to the decimal is not corroboration).
Lookonchain counts coins moving in and out of custodial wallets on chain and multiplies by a reference price. That lags settlement, and it misses in-kind moves that never touch the addresses it watches. Divide its dollars by its coins and the implied bitcoin price on Tuesday comes to about $79,231, against roughly $78,065 in the post it published the day before.
Neither method is the closing print.
None of which makes anyone here a villain. The on-chain series answers a real question, which is where the coins actually are, and it answers it faster than the funds do. It just does not answer the question the headline is asking, which is whether anybody bought.
So when a headline tells you the ETFs bought, ask which ledger. You are being shown either a change in share count or a change in wallet balance, and on a quiet day those two can point straight at each other.
We pulled the full session table to see how much the quiet days matter
The first week of September nets out to $723.4 million of inflow across five sessions. And $730.8 million of that arrived on Thursday 3 September alone.
Take Thursday out and the week is slightly negative, which is roughly the pattern all year. (There was no session on Monday 7 September, because US markets were shut for Labor Day, which is why the three-day streak covers the 2nd, 3rd and 4th.)
That is the shape of the whole year in miniature. One session does the work. Four sessions do the noise. And the daily flow post gets written anyway, because the slot exists.
A small number on a new product reads as a trend
Elsewhere in the same day's tape, CoinDesk reported that XRP ETFs took in nearly $2 million on Tuesday while bitcoin, ether, solana and HYPE funds all saw outflows. Two million dollars is about four percent of what left bitcoin, and it got a markets bulletin.
Which is how the format works. A small number attached to a new product reads as a trend, and a large number attached to an old one reads as weather.
Our read is that daily ETF flow has stopped carrying information and is now mostly a content format. The useful series is weekly, and even weekly it is dominated by single blocks trading at the close.
So we would rather watch one thing. Does any week in the fourth quarter post net inflows without a single session above $400 million doing the lifting? Broad demand looks like that. Nothing we can see this year has.
If you quote a flow number in the next month, quote the source alongside it. The trackers are not going to reconcile, and the funds are not going to report any faster.
