It's probably not that they're hard, but rather that they might require some boilerplate code and bloat that somebody might want to offload to an external service like this one.
The idea was to make something easier than booting up a database. Personally, I just needed a semi-persistent count that I could discard later. This fit the bill for me, and being able to retrieve the counts via JSON, I can easily add them to my Geckoboard dashboard.