Actually, it is pretty important in research first, because it determines whether a large enough sample can be generated to gather robust findings and second because the bigger your pool to easier it is to reach low incidence targets (for example purchasers of a product with a 1% market share).

That said, as ever, size is not everything (thankfully). 

Larger panels can theoretically generate large enough samples for low incidence groups, but only if they already know this piece of information.  If not, they still have to ask and screen out thousands to find a low incidence group (expensive).
Also, larger panels, inevitably start treating their members as numbers and not as individuals.  We’re always very grateful when people complete our surveys and we want to keep on thinking of them as people and not sample!  As a trade off, 100,000 feels about right to us therefore.

Going by stated numbers, MyVoice’s panels appear to some of the larger ones in the UK and Japan.  Companies like ResearchNow and Ciao have larger UK panels, whilst YouGov and Brainjuicer and many of the “traditional” research agencies like Mori and Ipsos have smaller or no panels at all.  In the US, Greenfield (the owner of Ciao) is a major player running a couple of panels with several millions in each of them (including Gozing.com).  In Europe we favour Bloomerce – they’re a really nice, quality-conscious company and were recently bought by SSI (the other major US sample provider).  As a good failsafe, Global Market Insite also offer excellent coverage.  (I think that’s enough plugging the competition – Ed).