Earlier quoted context omitted.
How are you quantifying that the fills are better? I worked professionally on this problem a while ago and it was very hard to do even with full depth feeds direct from the exchanges.
price and execution time? how much slip or price improvement on limits? or god forbid the casual retail market orders fill vs nbbo and pools at exec
When I was doing this for work the issues we ran into came down to a) making the orders hit the tape close enough to ensure similar priority b) the size of the orders changing execution depending on venue c) differences in performance per symbol.
This was in a place that was sending a fair amount of orders in. Even then given the above finding statistical relevance was hard.