Every session starts the same way. You've got a part. You open a new project, you play the part a couple of times against the click, you nudge the number until it feels close, and then you never touch it again.
That number goes on to decide the phrasing of every vocal line, the length of every reverb tail, how much air the drums get, and whether the chorus feels like it opens up or just gets louder.
And it came from you, alone, playing eight bars of a riff before the song existed.
The riff is a bad witness
When you're playing a part by yourself, there's nothing else in the room competing for time. No snare taking up space on the backbeat. No vocal that has to fit six syllables into a bar. No low end blurring the transients.
So you play it a little quick. Almost everybody does. There's adrenaline in a new idea, and adrenaline lives about 3 to 6 BPM north of where the song actually wants to sit.
Then the arrangement fills in. Drums arrive, bass arrives, a pad or a second guitar arrives, and the same tempo that felt urgent when it was one guitar now feels rushed. Not obviously rushed. Just busy in a way you can't locate.
That's the moment most people start EQing. It isn't an EQ problem.
The vocal is the authority
Here's the rule I've landed on: the riff proposes the tempo, the vocal decides it.
Instruments are flexible. You can play almost any guitar part 5 BPM slower and it still works, it just breathes differently. A voice is not flexible. A voice has a fixed amount of time it needs to get through a phrase and still land the consonants, and it needs somewhere to breathe before the next one.
If your chorus lyric is dense, if there are a lot of words crammed against the front of the bar, the song is probably too fast and you're compensating by singing flat, clipped, unmusical lines.
If your chorus lyric is sparse and you find yourself adding melisma or holding notes just to fill the space, it's too slow. That one is rarer but it happens, usually on ballads people slowed down because slow felt more serious.
The test takes ten minutes
Bounce a rough of the section you're least sure about. Then make three more versions of it at minus 4, minus 2, and plus 2 BPM. Any DAW will do this without repitching anything.
Sing the chorus over all four. Not in your head. Out loud, full voice, the way you'd actually track it.
You will find one where you stop working. The phrase ends and there's room, the breath happens on its own, you're not chasing the front of the bar. That's the tempo.
I've had songs move 6 BPM off this test. Nothing else changed. Same parts, same key, same arrangement, and the chorus suddenly had a floor under it.
The reason this works is that singing recruits your body. You can talk yourself into a tempo while you're listening. You can't talk your lungs into one.
Sometimes it isn't the tempo
Worth saying, because I've chased this the wrong direction plenty of times.
If a section feels frantic and every tempo you try still feels frantic, the problem is density, not speed. Too many events per bar. Sixteenth hats plus a busy bass plus a strummed acoustic plus a counter-melody means the bar is full regardless of what the number says, and slowing it down just makes it a slower traffic jam.
The diagnostic is quick. Mute everything except drums and the lead vocal. If that duo feels right at the current tempo, your tempo was fine and you have an arrangement problem. If the duo still feels wrong, it's tempo.
Run that check first. It'll save you from re-tracking a whole song to fix something a muted shaker would have solved.
Feel is not the same as BPM
One more thing, because it's the part people skip.
Finding the right number doesn't mean the song has to sit rigidly on it. Real records push and pull. The verse sits back, the pre-chorus leans forward, the last chorus is a hair on top of the beat because the drummer got excited.
You can build that in deliberately. Most DAWs let you draw a tempo map, and moving the last chorus up by 1 BPM does something a fader can't. It's below the threshold where anyone consciously notices, and it reads as lift.
You can also just play it. If you're tracking real instruments, let the take be slightly ahead in the chorus instead of quantizing that feel out of existence. The grid is a reference, not a verdict.
Do it early
The reason this fix gets skipped is timing. By the time the tempo problem is obvious, you've tracked twelve parts and comped a vocal, and changing the number means the whole thing goes out of alignment.
So do it before you commit. Get a rough arrangement up, get drums and a scratch vocal in, and spend the ten minutes on the four bounces before you record anything you care about keeping.
The tempo is a decision. Right now, for most of us, it's a leftover from the first four minutes of the session, made by a person who hadn't heard the song yet.