http://www.mgchemicals.com/products/emi-and-rfi-shielding/acrylic-conductive-coatings-original-series/super-shield-nickel-841
This stuff is the business for EM shielding of plastic cases.
That, or you can get boxes already coated with it. I've used a Hoffman QLine (IP67) box like this. The coating looks like metal flake paint, and the resistance between most any points inside the enclosure is just a couple ohms. So the attenuation looking in from the outside, through a metal housing connector bolted into the box, is quite good (>60dB?).[1]
Keep in mind that exactly this is necessary. The metallic connectors, that is. If you aren't shunting outside noise, conducted along cables (similarly shielded), into and over the shielding, then you aren't doing much of anything with the shielded box. Or if your signals are unshielded but low bandwidth, so you can filter away the noise, then the EMI filter needs to be referenced to the shield, not necessarily circuit ground as such.
If you have good contact to the shield-ground, on board (such as by having many screw bosses, or using EMI spring tabs), then you can use that for the EMI filter or shield-to-ground connection, without too much lost performance. (Maybe you'll get 40 instead of 60dB attenuation, but maybe that's plenty for the application.) As a bonus, you then have tons of low impedance grounds around your board, which means you don't need to play any ground-splitting tricks whatsoever -- a single ground plane will do nicely.
If you still need separation between board-ground and shield-ground, then you have the perfect opportunity to add ferrite beads on the signals, improving CMRR even further. Do this cautiously, of course. Don't make CM-to-diff (mode) conversion happen. (Example: you have shielded cables, coming in on metallic connectors, bolted into the box. That's your main shield scheme. The connectors, inside, have wiring harnesses that carry the signals from the bulkhead up to the board. These cables may exhibit mode conversion, because they aren't shielded, or at least the signals aren't paired, symmetrically, with ground, and enough ground lines, for that matter. In such a case, ferrite beads help, by making the ground-loop voltage drop more even across all wires in the enclosed cable. Even better, is to continue the shielded cable on the inside as well, so that the shield is grounded in two places: the bulkhead and the PCB. A ferrite bead around that will have little effect on mode conversion -- because the shield is already quite good on CMRR -- but can further increase CMRR or immunity, especially against endogenous noise signals: that is, unbalanced emission within the enclosure itself, that can interfere with itself.)
And yes, foil is perfectly acceptable. Note that foil tape needs to overlap properly; not all foil tapes have conductive adhesive. It can help to curl the edge over, so it makes contact at least at a few points.
[1] Did you notice it's an IP67 (water tight) box? How does that work? Surely the cover won't connect. Aha, they thought of that: the o-ring is also coated, or made of some sort of conductive material anyway. I'm not entirely sure what material it is, but it seems to do the business!
Also, I'm not recommending QLine boxes, specifically -- they work, but they're rather expensive, and it looks like there are cheaper equivalent options out there.
Tim