Use of TF vs odometry information

edit

Hello, I started using Ros a while ago to acquire information from some sensors for robot control and I ran into the following doubt.

I understand how TF works, although i'm still figuring out transforms, but I don't fully get why is it better to use the TFs instead of the odometry information directly from the sensors. I have made two codes that are very much identical, but one reads the TF information, while the other reads the sensors' odometry/pose topics and they both seem to work.

Every package seems to go through the trouble of building a tf tree and that adds latency. Isn't it easier to work with the raw sensor outputs?

Thank you in advance.