It never ceases to amaze me how many power searchers and search engine optimization consultants still do not understand basic statistics and data mining. Using Google for SEO research can barely be considered data mining, since the information is so easy to acquire.
When an SEO or anyone else tells you that the number of competitors for a particular keyword search are in the upper right hand corner of a Google search, politely inform them of the misinformation.
Below are 2 examples of competition using results and the differences of finding the exact number of competitors versus the number of word references on a page.
screenshot for the keyword phrase: los+angeles+florist
this only shows the number of times the words appear on a site and is not accurate. There are not 414,000 florists in Los Angeles California.
When evaluating the potential of a keyword or keyword phrase, do an allintitle search before beginning any search engine optimization campaign.
screenshot for the keyword phrase searching only the title string: los+angeles+florist
This will show you the true number of competing websites that are trying to rank for the keyword by only the phrase as it appears in the title tag.
This targeted search shows there are only about 12,000 florists and competing floral websites in Los Angeles California. This makes a lot more sense and is a logical number. I estimate there are only several thousand physical flower stores in the area, and the rest are floral websites, distributors, online order takers, flower directories and national florist chains with no local presence.
Here is an interesting piece of trivia: there are less than 100,000 fast food restaurants in the entire United States. It seems like there are millions of McDonalds, Burger Kings, Taco Bells, Subways, Jack in the Boxes etc, but this is just an illusion. The same is true for broad keyword searches, there are less true competitors than you realize.
Rand wrote more about search parameters, how this relates to keyword research and how to make meaning out of generic search data