The downside of this kind of project is that it can often be quite time-consuming, and often-times you wonder if failures are caused by the parts you can't see; the owners not soldering/wiring things properly.
I've had interactions where my "code was broken", which mysteriously got fixed without me making any changes. Throw in timezones into the mix, and you can waste a day looking for a bug that just isn't there.
But I far prefer working with actual businesses because they just want an end result and won't try to dicker with me by pointing out how high my prices are and how much less "this guy on Guru.com" can do it for.
The thing I truly love about the Arduino ecosystem is how fast it lets me put together those one-off projects. Things that would take weeks in "normal engineering" can be out the door in an evening. What I don't like about Arduino is that the focus on simplicity often comes with a huge cost in performance. Luckily, it's rarely been an issue.
Dataloggers are a dime a dozen, but yet someone still wanted to pay more than the price of buying three off the shelf units simply because what he wanted was slightly different from everything he could find.
The point is that it doesn't matter how many times something has been done; there's always something different you can do that a niche of customers need.