Two tracks at the same BPM can still refuse to sit together

You line up two records that both read 124. The beats land on top of each other. The blend still sounds wrong, and not in a way that nudging the pitch fader fixes.

The number was never lying. It was just answering a narrower question than the one you were asking.

A tempo is an average over the whole file

Beat detection finds the recurring pulse across a track and reports a single figure for it. That figure is exactly right for a record programmed to a grid and quantized to it, which is most club music and is why the technique works at all.

It is a summary for everything else. A live recording, a band playing to a click that somebody drifted off, a record cut before grids existed: all of these have a tempo that moves, and the reported number is the middle of that movement rather than a property the whole file holds. Two tracks that agree on the average can disagree by a beat over four bars.

Four ways the number hides something

Octave. A detector deciding between 63 and 126 is choosing which layer of the rhythm counts as the pulse, and both answers are defensible. Halving or doubling a figure is the single most common tempo error, and it is easy to spot: if the number is roughly half or double what your ears say, that is what has happened. Nothing else is wrong with the track.

Groove. Two tracks at the same tempo can place their offbeats in different positions. Swing pushes every second sixteenth late by an amount that a straight programmed track does not have at all. Both are 124. Layer them and every hi-hat flams, because the downbeats agree and nothing between them does.

Phrase. Beatmatching aligns beats. It does not align bars, and it certainly does not align the sixteen or thirty two bar structure that both records are built out of. A blend that is beat perfect but phrase shifted puts one track’s breakdown against the other’s drop, which reads as a mistake even to a listener who could not name what happened.

Drift. Where tempo genuinely moves, alignment decays. The mix holds for eight bars and comes apart by the sixteenth, which is the version of this problem that is hardest to diagnose in a booth, because it sounded fine when you started it.

What to do about it

Read the number as a filter rather than a verdict. It narrows a library of thousands down to a shortlist that could work, which is the expensive part of the job and the part worth automating.

The shortlist is where listening starts. Groove and phrase are decisions about where two records sit against each other, and they are made with the same ear that decides whether the transition should happen at all. A tempo figure gets you to that decision quickly. It was never going to make it for you.