Sorry but obviously the best is not to have a poor keyword for test because contrary to your perception it will have the maximum chances of interference.
Let me give you an example. Do a google search for seo case study chrome bookmarks, you will find this thread ranking on the first page. Why?
Because no body else is competing for this keyword and its more than enough for the thread to have a title that matches the search query for it to rank.
You make a little change to the search query and google seo case study chrome and you wont find the thead on the first page.
Now, apparently an exact title match and few phrases of keywords in content are enough to influence the rankings of a very poor keyword.
When we do a case study to find out whether something affects the search rankings, we assume (or rather make sure) that other factors remain constant or rather too unimportant to affect the rankings.
In this case even a slight mention of the website in question can cause it to jump a few places. Even more traction to a thread with the website mentioned can cause rankings to jump a few places. It renders the case study completely useless. You can't ever be sure what really caused the ranking to jump.
Sorry for my Egnilsh.
