The only way to properly control for the “interrogative” variable is to have two articles that are identical in everything (content, image, sub-headline) except for the headline, and then compare the views each receive.
For example, you would write an article about, let’s say, how to get over someone you never dated.
Now, you make two copies. You give one the headline: “How to Get Over Someone you Never Dated”
And you give the second, identical copy the headline: “Things that Help you Get Over Someone you Never Dated”
In this scenario, we have isolated the ‘interrogative’ variable, the “How-To” everything else is the same. Now we can attempt to make a correlation between how-to and higher views.
In order to draw empirical conclusions about headlines, you have to control for more variables. That means setting up an experiment.
What you’ve done in this article is a good way for you to review your own work, but because an experiment wasn’t set up at the beginning that controls for different variables, we can say nothing conclusive about the results.
And just because I am criticizing the methods in this article does not mean I find this piece worthless. In fact in science, we highly value frequent and detailed criticism.
I am criticizing the methods because I believe the initial work was so well done, that it merits a closer look.