We just spent the past few weeks working specifically on tempo maps for our album, and almost every section has a tiny shift somewhere.
The way we did it was me (guitar) and the singer went through every song and created a draft map.
Then we'd go try it out with the drummer playing on an electric kit, and make adjustments as necessary. The only reason we did it in two sections not one was due to the drummer's work schedule, but I think it allowed me and the singer to discuss and notice things before we started the 'real' mapping.
We also referenced and analysed rehearsal recordings. It helps if you have a few versions of the same song to see if you do something every time that sounds good (or bad).
Two important things to note:
1. We're a hard rock band not a metal band, so YMMV with my own learnings here.
2. We don't write to a click, when we send each other song ideas it is not to a click, and we don't do 'high quality demos', we just record with a stereo mic in the practice room/bedroom and don't over analyse it performance wise.
When mapping:
We noticed that on big chorus fills we sometimes (not always) 'push' slightly, so sometimes need to start a hair early next bar compared to static click. Feels totally natural when playing as a band (no click) to push (tiny tiny amount) but if we tried to force it to static click the fill would end up needing to sound too robotic, or lose the energy from what we wrote. So sometimes pushing one or more beats up a tiny tiny bit on chorus bars with massive fills. There is also a song where some fills have a slight pull feel in the middle 8, so the tempo map got pulled back a tiny bit. We found that mapping beat to beat isn't totally necessary, it is more important that the next bar naturally starts on the 1 again. Do this however works best (it was sometimes whole fill bar up/down Xbpm, and sometimes beat 3+4, or beat 4 up/down Xbpm). Just try to have it that when the main beat comes back in right when it should musically.
You have to make a judgement call on if changing tempo sounds good or bad. Sometimes you do it live because you're speeding up/getting over excited, and the song would sound better if you held it steady. So don't change just because it is a new section, change if it needs to.
Four key things I learned specifically relating to tempo mapping -
1. The tempo map is done when the band can play along with it and it feels natural. If a section naturally speeds up or slows down the click should catch where you go. The drummer shouldn't feel like they have to change the way they play fills to work with the metronome, the metronome should follow them. This was pretty crucial to us. By the same token - if the drummer is playing it shit, tell them, don't account for it in the click. But if it sounds good to push/pull then make the click follow it.
2. The numbers mean nothing meaningful so don't get hung up on them. If you need to change bpm by 5 you need to do it, if it only needs 1 that is fine. On one song that is around 83bpm we even found out we had to do a half bpm increase for the chorus to get the feel right.
3. When you check changes check in context, don't just check a section at a time, or a bar at a time. It is time consuming but you really need to check against the sections either side to make sure it flows well. Also when it is done check it the whole way through more than one time and make sure everyone agrees, or at least accepts your final decision if there is disagreement (I'm assuming you're in charge).
4. Think of the big picture - for us we have 100% clean vocals, so if we were deciding between two BPM's for a section the most important is which makes the vocal delivery sound best.