Datadriven on corrupt data
Every time you analyze data, it’s useful to ask yourself what conclusions you can actually draw from that data. However, it is often important to take a step back and check the data itself.
Here is an example of a somewhat anecdotal, but well illustrating the problem:
Imagine making some conclusions from the data obtained in this way. It does not matter what results we get at the end of the experiment — the data is doomed anyway. We don’t know what users think by clicking on this button.
But if you think about it, even if you remove all the ambiguity from the interface of this questionnaire, we still won’t get accurate data. Now it will be the answers of only people who decided to press the button, which also does not reflect reality.
In this case, to get the real data is very simple — just stand up with the counter, and calculate manually. Of course, this happens rarely, and we have to make all sorts of assumptions from the data. However, before you do them, it is important to make sure first that the data is generally suitable to do so.