Actually, and this sucks but it's probably what is occurring: it's simply likely the utilization breakpoints are slightly different on TU 04 vs. TU 8. They aren't the same necessarily between bureaus either.
Paying down utilization is a good thing on any FICO score, but that doesn't mean that the same paydown will produce the same magnitude result, or even produce a result at all on different models... it's awkward unfortunately, and TU 04 is probably the least well characterized score out of any of the major scores from a data perspective.
Basically revolving utilization when we're talking individual or aggregate (or actually numbers of tradelines with balances too) is all about percentages and breakpoints... if you don't reduce your percentage on whatever the metric is to something under a new breakpoint in the given FICO score, there will be no change. Conversely similar if you don't increase it over a breakpoint, no change.
To give an older example from my own file when I was playing with it a while back on EQ Beacon 5
1% utilization = 660
13% utilization and a maxxed tradeline = 646
27% utilization and a maxxed tradeline = 646
The first two utilization aggregate breakpoints are held to be 10% and 30% there, but when we start talking 50% and higher our data isn't as good TBH.
If you post your balances and limits for each card, and how much money you have to work with, we can optimize a strategy for paying things off to try to maximize scoring benefit.
