Unsure I understand your questions. You were looking to replace the new entity's 3500s with 3750s, or were hoping, by looking at some stats on your 3750s, to estimate the performance requirements of the new entity's switches?
I will say, first, comparing 3500s to 3750s isn't as straight forward as it otherwise might be. The latter offers more fabric bandwidth and higher Mpps (especially the 3750E/3750-X, effectively "wire-speed" capable), but the 3750 series' QoS, when enabled, allocates buffers, by default, much differently. So much so, the (3560/)3750 series (and I believe the [3650/]3850 series) tend to have QoS "issues", not seen on the 3500 series. (NB: many of these QoS issues can be "corrected" by configuration changes.)
As to looking at device stats, well the new entity might have totally different usage patterns then yours, and again, how the device's are configured (optimally, or possible not), could skew your analysis just looking at device stats. Further, truth be told, many of the "older" non-wire speed (for all ports) switches are more than adequate for user edge usage, assuming they are not still running at 10 Mbps.
(Many years ago, I was in a fortune 100 company "upgrading" from 4 Mbps token-ring to 10 Mbps Ethernet. I asked what "kind" of Ethernet, and was told shared hub. Well, I said, we'll be the last to convert - thank you very much. Which was good, because other departments "noticed" the network difference moving from 4 Mbps TR to 10 Mbps shared Ethernet. Yes, those "upgraded" saw a huge drop in network performance! Before getting back to us, they replaced the 10 Mbps hubs with 10 Mbps switches. After that, I "approved" our upgrade, which from the user perspective, users saw no network performance difference [i.e. from TR]. I mention this, because moving from 3500 to 3850, may not be seen, by users, as an improvement, and in some cases, might actually be seen as a loss.)
