Every venue running a perpetual computes its own funding, from its own book, against its own index, on its own schedule. There is no central rate. So the honest way to report funding on a coin is a set of numbers, one per venue.
An average of them is a number that no trader anywhere is actually paying.
The disagreement is the information
When every venue prints similar funding, positioning is uniform: the same lean everywhere, no particular venue carrying the crowd.
When one venue prints markedly different funding from the rest, that is a fact about where the leverage sits. It can mean a concentrated position on one exchange, a different participant mix, an arbitrage that has not been closed because moving collateral between venues is neither free nor instant.
Average those together and you get a middling number that describes none of them, and the venue that was worth looking at has been smoothed into invisibility. It is the same class of error as ranking a chain by turnover without excluding the plumbing: arithmetically fine, and about the wrong thing.
Why an average is worse than a single venue
At least one venue’s rate is a real price that somebody really paid. A weighted blend is a construct, and weighting it by open interest makes it worse rather than better, because it lets the largest venue’s conditions speak for markets with genuinely different participants.
There is also a practical asymmetry: you cannot trade the average. If your position is on one venue, that venue’s rate is your cost, and the composite is trivia.
The exception, and it is narrow
Comparing coins rather than venues sometimes needs a single figure per coin. That is legitimate as long as the reduction is stated — and better done by naming the venue used than by averaging, because a named venue is reproducible and a blend is not.
The NoVo Crypto Market Map keeps funding per venue for exactly this reason, and never collapses it into one rate.