The inductor and capacitors form a 2nd order low pass filter and one thing you have to be careful about is the Q of that filter. For instance, if you have a repetitive noise on the 3V3 that matches the resonance of the filter you could end up amplifying that frequency (or a harmonic) and making the analogue supply even noisier. Worst case scenario for circuits that don't have "damping" (R damps that resonance), is that applying power rapidly can damage some devices on the analogue rail.
So, in short you don't want Q to be that great.
I have, in the past on some designs, put small series resistors with the inductor to avoid this when I'm unsure about how things will pan-out in a design and it's no great deal if the analogue current is about 10mA milliamps or less - maybe 10 ohms will "lose" 100mV and if you can live with this I'd recommend going down this route. The graph below shows an LCR 2nd order LPF - the less peaky frequency response is when Q is lower and this is when series R is greater.
Clearly as the analogue load current increases this also applies dampening to the possible resonance of the tuned circuit so, for greater loads the requirement for a series resistor also diminishes BUT, watch out for loads that are intermittant because this might also cause ringing on the analogue rail despite the 3V3 rail looking steady.
Other things to avoid are saturating the core of the inductor because it's inductance rapidly reduces and it becomes a less-effective filter as load current increases.
Ironically, having lowish properties of self resonant frequency can be a benefit to blocking a repetitive 3V3 noise so this sometimes is worth factoring in.
I would also consider using a third capacitor to reduce really high frequency noise - the two you have already might not be that great above several tens of MHz - check the capacitors SRF too. Above SRF they turn into an inductive reactance and may cause problems.