Statistically speaking, you want a keyword which has the lowest possible competition and the highest possible amount of searches. The problems is that "competition" isn't really an exact measurement.
Chris talks about this in his keyword article. He calls it
Keyword Effectiveness Index. He uses the PageRank of the top search result for the competition variable. Other things you could use are: # of sites indexed for the top result, # of search results, # of backlinks for the top result or anything else like that. PageRank would probably be the most accurate of these, but if you could create a formula that uses all of these "variables of competition" to make one number to represent competition, it would increase the accuracy.
For comparing keywords, Chris' system plots the keywords on a pair of axis(pl?) and puts a gradient behind them so whichever is on the darkest spot on the gradient is the most effective. I would take the number of searches and divide it by the number you're using to represent competition to get an effectiveness number for each keyword. For me, it's better/more accurate to just compare 2 numbers than try to tell which spot is darker on a gradient. For a more visual people, the gradient might be better.
An example of how I would do it, if you have a keyword with 100 searches and a competition variable (possibly PR of the top result) of lets just say 5. You have an effectiveness of 20 for that keyword (100/5). If a keyword has 50 searches and the competition variable is 2, you have an effectiveness of 25... so that second keyword is 1.25 times more effective than the first.
This is how I would compare keywords... but as I said, the real problem is finding an accurate figure to represent competition. (Maybe someone should make that formula for getting an accurate competition number?)
Sorry if this is a little too "number crunchy", I've been thinking a lot about efficiency/productivity/statistics/systems so you caught me at a bad time.
Bookmarks