However, if you search for "dog house" using quotes the page must contain the exact phrase.

This is why it's a more accurate measure of true competition to use the quotes when researching, as you want to find sites that focus on "dog house", not just "dog" or "house".

Here's a Google hack to give you even more accurate results...Use the three "in" attributes when researching:

intitle:"dog house" inanchor:"dog house" intext:"dog house"

(paste the above into google search)

What this does is find pages that must contain "dog house" in all of the the following:

In the page title

In the body text

In a link anchor text

These are the three basics of SEO and will give you a good estimate of the competing pages that have done basic SEO, which after all, is really what you are researching. It isn't perfect, but it is the most accurate measure of your true competition.

Do the 3 following searches and see the difference in number of pages returned, as well as the results themselves:

dog house

"dog house"

intitle:"dog house" inanchor:"dog house" intext:"dog house"

A search without quotes should be considered the macro competition, whereas finding the micro competition would be better.

I think what most people miss though, is that the competition numbers are more of an "index" than an actual number and neither tell you exactly how much competition there is. Instead, these numbers are relative and should be compared to other keyword numbers. That's the value.

Also, if you are really looking for long tail keywords, which "rock guitar lesson" probably is borderline, authority will mean less.

In my opinion, the best option, one that's both accurate,andeasier:

Use the 3 "in" hacks at the same time:

intitle:"rock guitar lesson" inanchor:"rock guitar lesson" intext:"rock guitar lesson"

Paste the entire line above into a google search...

(Note: I don't think you can use the "allin" hacks at the same time, which is why this is easier)

I get 42 results for this search:

http://www.google.com/search?hl=en&rlz=1T4GGIH_enUS274US274&num=100&q=intitle:"rock guitar lesson" inanchor:"rock guitar lesson" intext:"rock guitar lesson"&btnG=Search

Now, the "42" results is NOT the true number of competitng sites, as Rob pointed out. It is the number of sites that have basic "on page" SEO.

However, I believe this "42" is the best number, based on accuracy and ease. And this is the number one should use to compare to other keyword phrases.

For example, the "3 ins" search for "guitar lessons" returns 327,000 results.

3 ins for "beginner guitar lesson" returns 96 results.

Here's the numbers we should compare:

42
96
327,000

Now, having said all of this...I only use this when I am limited to a single page, like a squido lens, which isn't often.

The truth is, if you already picked a niche, you want to target as many keyword phrases as you can, focusing on the ones with the "best" numbers, be it low supply, high demand and/or high value. The only way to get serious traffic is to rank well for dozens, if not hundreds, of decent keyword phrases.

IMO, keyword research shouldn't be done to find a good keyword, but instead to find the BEST keywords in a particular niche.

The term "best" is also relative...The less experienced you are, the more you should focus on competition (supply). Whereas highly skilled SEOers will consider high-paying, high-traffic numbers as "best".