The CTR Factor.
The CTR is all important. There is no Quality Score with Facebook as there is with say Google, so this makes things simpler. Generally you need your CTR to stay above 0.04% for your ads to keep showing. Go below that for too long and your traffic will peter out, even as you keep raising your bids. Obviously the sharper you are finding the best responding demographic for your ads the higher your CTR will be.

The Volume Factor.
Facebook being a relatively simple PPC ads system, creating more ads will get you more traffic. You can even do this simply by duplicating the exact same ad, with the exact same target demographic – this is not against any Facebook rules and Facebook will not detect it and try to even out the traffic.

The Facebook Randomness/Momentum Factor.
When you set your initial bids there is a point below which you will not be given any impressions at all. If you find that point and bid at it or just above it you’ll be given a certain number of test impressions – say 2000 – will be a make or break time for your ad. If its CTR is high enough it will keep showing. If not, it will fade away. But a quick experiment will show that you can submit the exact same ad 10 times at once (with bidz around the ‘break-point’ that is) and find that some of the ads take off and get lots of impressions and subsequently clicks, while other will get few impressions and die. An ad must get momentum in this initial test phase and whether it gets the momentum to keep going has an element of randomness to it. Sharper targetting and higher bids will raise and can ascertain that the ad will keep getting shown, but sometimes its hard to get a high enough CTR or a wide enough target to get the momentum. And why bid higher then necessary, anyhow?

Taking the above three factors into account, here are the two targeting strategies that work best:

Bulls Eye Targetting
Here you simply find the target demographic that give you the best CTR and aim for that. You can use keywords or other options, like sex/age/marital status. Then you scale your campaign around the target in wider and wider circles. Eg, say the ideal target is teenage females who like The Jonas Brothers (this is a real example offer that currently available). Your sharpest ads will target the Jonas Brothers as a keyword, but then you widen the net to fans of bands similar to the Jonas Brothers, and to males and to older demographics, as long as you’re still getting a positive ROI.
The main problem with Bulls Eye Targetting is that often the target is simply too small. This may mean it saturates fast or even does allow you to break through the momentum factor.
The main advantage is that you can get your CTR right up and thus bring your bids right down.

Sweep Targeting
The other targeting strategy is to go wide from the start. For some campaigns the appeal is a very wide demographic or it is simply too hard or impossible to pinpoint the target with keywords and other options. What people first try to do is to pitch the same ad or same variety of ads at different ages in succession. Eg. target 25 year old females, 26 year old females, 27 year old females etc. Or to target ranges: 25-27 year old females, 28-30 year old females, 31-33 year old females, etc However the Randomness Momentum Factor greatly reduces the effectiveness of that approach. One solution is to pitch the ad to each demographic multiple times. My preferred approach is what I call the Sweep. Say I find that an offer works well with older women – and this usually just comes down to common sense. Rather then target each age, or segment the age range I would target this way, as an example:
59-60 females, 58-60 females, 57-60 females, 56-60 females, 55-60 females, 54-60 females, 53-60 females, etc. If your whole target is women between 35 and 60 you can all the way down to 35 and then back up (35-36, 35-37 etc) and if you see a sweet spot, work around that as well.

One reason this works is because it eliminates some of the randomness factor. If you target each age you’ll find some ads simple wont work or seem inconsistent, and your test data will be off. Another reason is that generally the larger your target demographic, and the more general, the lower your minimum bids will be, which will help the ads get momentum.

For best results you can play around with combining the two strategies.