My answer hopefully applies to beads but also ferrite clamps on cables.
If it's preventing susceptibility problems due to interference coming down a cable then its position is not that critical. However, if it's stopping emissions getting out from a device, a better position is as close to the interfering source as possible. There is every chance that interference getting onto a cable (such as from a switched mode power converter) can radiate from the cable so, best to keep the ferrite up as close to the source of noise as much as possible.
Noise coming into an enclosure on the conductors (conducted noise) should be filtered out closest to the source or right at the enclosure wall. Any circuit loop made by the filter components should be as short as possible. Feed-thru capacitors (shaped like short coaxial cables), inductors, and/or ferrites can help here. If there is one input with conducted noise then a single filter set might be enough for the whole system.
If the noise is being radiated into the enclosure (magnetically or electro-magnetically - eg. RF) then you might require filters at each regulator and on each sensitive component. An extra filter for each regulator output might not be needed unless the component being powered is some distance from the regulator, though standard by-pass caps and recommended regulator caps should still be used. (By-pass caps are also used to limit noise or pulse energy coming back from the component itself.)
Adding by-pass caps after an LC low-pass filter will further lower the noise and the frequency points, though this is usually a desirable effect especially on power supply lines. If for some reason you need to have a specific cut off frequency on the line you could add a parallel LC pair in the line. The caps to ground would still give the low-pass effect.
Don't go too far with too many caps or extremely large cap values. Remember that when power is switched on all those caps need to fill, this might put a big strain on the input power system or battery, (or system fuse). When power is switched off all those caps also need to drain.
To reduce RF noise from getting at sensitive circuits you could also consider shielding the whole electrical system within a metal enclosure. This is one alternative that might avoid using a large number of ferrites. Unfortunately if the internal components are producing the RF noise then there may still be the need to have several ferrites in the circuit.
For stubborn RF noise you might need to filter even the ground level power lines with a series inductor or ferrite, perhaps even with a cap connected to an earth ground or to another known quiet ground.
To filter common mode RF noise (equal noise on two opposing lines) a component such as a common mode choke can be used. This looks like a small transformer that tries to reduce RF on the two lines by winding them near each other to actively oppose the noise currents, sometimes with a ferrite core.
Using ferrites can be tricky too. These do not work the same as a standard inductor. They reduce RF by dissipating the energy into the ferrite material. Ferrite materials are also rated by frequency. You need to look over the manufacturer's spec. Ferrite material for 500 MHz might not help as much at 2GHz. A ferrite component also works best when there is actual noise current trying to flow through it.
Best Answer
I'm researching information on decoupling capacitors and came across some information about ferrite beads from TI:
I believe you should examine what your switching current spectrum looks like. If your digital circuits require large current transients, you should not use a ferrite bead on them.
I am currently of the mindset that the ferrite bead is useful in certain, very specific applications, but it is mostly used liberally as a band-aid when issues arise that should be solved by examining the power delivery network.
While it would be nice to see some graphs or other data, what I read here from TI sounds plausible. What do you guys think about it?