I'm making a software for a very small and specific market. Worldwide, the numbers seem good. That is, there is a possibility for business. Nevertheless, I wanted to post my procedure. The question is simple: is this right? would you consider this valid?
I first looked for data in the US census information. There I learned that the sector I'm interested in makes X amount of money per year. But I'm not aiming to the whole sector, just a small portion of it. To calculate the portion, I went to google analytics and found the relationship R between searches looking for information related to the whole market, and searches for keywords related to my more specific market share. I then assumed that the market share for my product is X*R in the US.
Then, I would have to estimate a realistic share of that market that I can get (100% is not realistic, ever), let's call it S, so that my final magic number (in the US) is X*R*S.
I'm aware that I may be dead wrong. But is that probability high? would you count that as valid? would you bet on it?
I don't think this is a great method of measuring market size. Factors I don't see accounted for in your method:
That said, while I don't think niche to broad search traffic ratios are a good method for determining market size, niche market search traffic alone isn't a bad indicator of market opportunity. Think about how you could validate the opportunity through ads leading to a pre-sales page.
Also, make sure you're looking at exact-match search counts in the AdWords keyword tool, not broad-match, otherwise your estimates of how much search-traffic these keywords are getting will be vastly inflated.