Keyword tools are notoriously unreliable at helping you predict the traffic you can achieve on a given website. Use them for relative
traffic judgements. For example, if they all agree that "widget learning classes" vastly outperforms "widget remedial learning", then I would go after the former before the latter (but I would still make at least one page targetting the latter, because the traffic estimators cannot be trusted).
It's not that unusual that they don't agree, in which case I tend to go with Google's guesstimate. Your ability to construct (and make appear!) SERP text that "sells" the searcher on clicking can make a big difference in the traffic you get for a given word.
Your website's relevance to the search term can make a big difference in the traffic you get for a given word.
As I look at my top 10 traffic keywords I see things like:
- One is a misspelling of a popular product name. I couldn't spell it right when making copy, and it turns out that many web searchers can't either.
- One is a very outdated, nearly antique technical product. No clue from keyword tools that it could offer decent traffic; just discovered that accidentally by studying my logs and then figuring out why people were searching for it and then making content to fill that obscure need.
- One is a modestly popular keyword. But I built all the content for it about a year before it became "hot". That "aged" content is the only reason I can rank well for it now. I figured it would become a a popular search term and it did -- eventually.
- One is an odd combination of two popular keywords. No keyword tool alerted me to this one. I never would have found it except that I regularly write short copy that is designed to be "test holes", drilling for keyword oil :-). The two keywords are, in some sense related, but one small page was the only place on my website that combined the two. It sat dormant for the biggest part of a year, and only started getting significant traffic a couple of months ago. Turns out that the two keywords represent products, and an incompatibility in using the two together has arisen. So now I have to write some content to satisify that need (and hopefully attract some relevant, profitable ads), and then add links to take visitors of that page to the new content.
I use keyword tools too, but I don't treat them as anything but very rough ideas of whether keyword A has more traffic to offer than keyword B. I find as much or more value in studying weblogs and in writing lots of "experimental" content -- short pages that each target a completely different term, to see what traffic is really out there.
It's surprising how often those experiments can sit there dormant for a long time, and then suddenly alert me to a new trend in search patterns.