There are many definitions for KEI, but Wordtracker has the best definition for it:

The KEI compares the Count result (number of times a keyword has appeared in our data) with the number of competing web pages to pinpoint exactly which keywords are most effective for your campaign.

Credit to WordTracker for providing KEI for their keywords, but everybody does that now. KEI does look effective on the surface as you can find terms that are less competitive and easy to rank for. The problem is that KEI does not factor in the quality of each search term. For instance, KEI cannot predict if your keyword attracts shoppers, information seekers, or buyers. If you are in business to make a living, you want to convert folks to buyers, but KEI has no way of determining that.

KEI also does not take into account the quality of competition. You might find a keyword that looks effective according to KEI, but if you have 30 Yahoo’s or Google’s in that niche, you pretty much have not much of a chance for getting ranked in the top 30.

In essence, KEI is not a credible metric on its own. It can be helpful if you do your research to figure out the quality of keywords but by its own KEI is just a big empty word. Now, I am not suggesting that you shouldn’t use it. It’s just that it is not accurate enough to build an entire SEO campaign based on KEI numbers, unless you don’t like your job.