De-essing is also known as "frequency dependent compression," but whatever you call it, the idea is to control excessive sibilance (exaggerated “s” sounds), which can be particularly problematic on vocal tracks.
The multiband capability is a feature that puts the De-Esser ahead of the competition making it precise and versatile on top of good sounding.
When set as 2 band mode the De-Esser works in a broader fashion, effectively reducing the high end content of the area set by the frequency selector. Switching to 3 bands mode allows more precision and control, particularly useful in situations like vocal tracks where sibilance is in a narrow range, and where there is the need to leave the higher part “air” untouched. The dedicated control set also allows you to precisely tailor the operating mode, and audition the de-essing action or the part of the audio spectrum the processor is working on.
The primary use of this processor is obviously on vocal tracks: this is where the IK Multimedia De-esser shines thanks to its multiband capability which allows to zone in with extreme precision on the troublesome frequency and put it under control.
However, there are several other instances where a de-esser can be a secret weapon where an EQ would be too a drastic solution: hi-hat or crash cymbals can sometimes sound harsh or overly bright in a mix; using an EQ would fix this, but would probably affect also the rest of the high end content of the mix. That's where frequency conscious compression comes to the rescue: setting the De-Esser on the right frequency and threshold will allow to reduce the harshness of too bright cymbals only when they are hit, and with the precise setting of release time the action would be unobtrusive and transparent to the rest of the mix.