There is no precise rule nor recipe for what you should do. It is definitely not "free for all" and "everything will do" situation, but to some extent what you do with your results is up to you.
You should definitely try to describe the trend in a numerical way - find and fit a line/curve that nicely follows the data. You can try to guess what is the reason behind the trend, and possibly propose ways of checking whether the suggested mechanism is correct, or not.
You should try to search the literature and see if someone has not tried to research the effect before, to be able to quote their findings in your report. It might happen that the effect they report is a different from yours, which should lead to discussion of whys.