I've been working on an article for a few months now. I had an idea of an interesting benchmark to do - but to be fair I had little idea on how to actually conduct it.
I started little by little: I knew I needed to write software to execute certain algorithms, and I did that.
I knew I needed to store certain results, and I did that.
I knew I needed to use appropriate statistical methodologies to analyse the results - I started reading articles and chose appropriate ones.
Now everything fits together, I just have to put together all the pieces and I'll get the results needed to write the article.

Now I'll have to see what the results are and what to write in the article, but what seemed a daunting task at first is now just a logical sequence of steps.