Evaluating keywords for relative opportunities is still very tough.
Without massive stores of data and detailed insights into the competitive landscape it is easy to throw a lot of time and money at targeting keywords that will never deliver a stable return.
In order to avoid this and instead focus only where there is rank to be had and money to be made, I believe it's best to use an evaluation method.
There are a number of methodologies ranging from complex formulas including many different heuristics to more simplified models designed just to give you a sense of the opportunity.
I have developed a simple keyword opportunity model to help with this process, but for this post I am going to use a simplified version.
This is meant to be more of a barometer than anything else and this data should not be used to make a business case, but more so just to give you a sense of opportunity.
I'm going to use keyword competitiveness scoring from SerpIQ, but you could also do this with SEOmoz's keyword difficulty score or something similar.
Scoring the opportunities takes some simple math. So for this version we are going to discount monthly search volume by multiplying is against the inverse discount rate of the competitive scores (1 minus the competitiveness index percentage).
For this example I'm going to stick with our root phrase, keyword research, and use 2 other closely related variations:
Using our quick and dirty evaluation model of (exact monthly search volume x (1 - CI%) we get:
keyword research opportunity score (4,400 x (1-.68)) = 1,408
keyword research tools opportunity score (720 x (1-.65)) = 252
keyword research software opportunity score (480 x (1-.56)) = 211.2
Showing (quickly remember) that even though keyword research is significantly harder in terms of competition, it is still the clear winner in terms of keywords to target for search traffic.
This is useful because it takes into account the opportunities as related to monthly search volume and then discounts the traffic potential based on the average competition.