Webanalyticsbook usually has stuff in his blog that I don't get to read (except in his blog - or because he told me about it - that's Sebastian Wenzel who's also a friend of mine and on the Social Media Committee of the WAA).
In this case, segmentation by three level deep is just another way of saying segmentation using 3 dimensions (makes sense, doesn't it?)/
"..Segmenting Three Deep
It’s a good suggestion to segment any key performance indicator to a level three deep in order to understand the meaning of the statistic in greater detail. We’ve already segmented by search engine referrer and by keyword but what if your ad group has several different ads, each with different ad copy and a slightly different appeal? Wouldn’t it be helpful to know which ad is driving sales and which is simply costing you money?
It’s almost a trivial matter for you to add another adCenter parameter to your destination URLs which now looks like this:
www.example.com/LandingPage.html?keyword={Keyword}&adid={AdID}
Now the same keyword can be tracked separately for each ad, providing insight into the effectiveness of your ad copy. What could be more actionable than knowing which ad is generating revenue-and how much?”
Entire segmentation post here"
For example, if you have referrers by keyword to a page, that's one dimension (or two, it depends on the analytics package) if you add if it's a new visitor vs a repeat visitor, that's 2 dimensions - and if you add geolocation (ie: New York State) that's 3 dimensions.
Trick is, to do meaningful marketing analysis you need at least 3 dimensions, perhaps 4 or 5 would be even better but most analytics packages, except the high end one's only are 2 dimensional - you can get one thing charted aginst something else, but not the third thing.
Comment Preview