More than 2% is excessive for motors; for other types of load, more than 4% is too much.
Suppose you are measuring the voltage for the nominal 480V supply to a motor. If you measured the same voltage on each phase (e.g., A-B, A-C, B-C), then clearly there’s no voltage imbalance.
Suppose you measure 481V, 480V, and 479V. That’s an imbalance, but is it significant? The general rule of thumb for motors is you want 2% or less imbalance. For other types of load, you’re looking for 4% or less.
That percentage isn’t a percentage of the nominal voltage. Nor is it the simple difference between the highest and lowest voltage; the actual calculation is a bit more complicated.
Here’s an approach for seeing if your circuit is within the desired limit.
Measure the voltage on each phase. Suppose you measure 469V, 474V, and 481V on a motor supply circuit.
Multiply your desired percentage by the lowest voltage. This gives you the maximum allowable voltage difference between the lowest measured voltage and the highest measured voltage.
In this case, the maximum difference is 9.38V. But the difference between 481V and 469V exceeds that, so you’ve got excessive voltage imbalance. If the circuit had been serving a non-motor load (other than a “sensitive electronics” load), voltage imbalance would be within normally acceptable limits.
The limitation of this approach is that it doesn’t tell you what your actual voltage imbalance is, only whether it’s within the desired limits.