The free tier goes a long way. In my experience, if you are thoughtful with the quantity of requests you send from the clients, it will take over 1000 users to push you into the paid tier. This really depends on the nature of your application though.
As for development, there are a few nuances that you have to get right to use Parse at full capacity. First, delegate as many operations to CloudCode as you can. I do this for separation of concerns (clients should focus on displaying data, not retrieving and formatting it) and more importantly, to make cross platform applications consistent and easier to write. Also, you can change behavior on the fly instead of creating new builds of your client when you just want to change a small detail, like the number of results to return. When using CloudCode, take advantage of Parss's implementation of JavaScript promises. And avoid mixing promises with the "backbone-like" callbacks. There's a bug I've found when mixing the two that will occasionally cause the operation to finish before executing all the callbacks.
And speaking of bugs- the biggest drawback of Parse is the documentation and support. Some entries in the docs will actually be a one sentence reiteration of the title of a method. It's mind blowing how shitty some parts of the docs are. The support is even worse. Looking through the help archives (they've now given up on internal support and moved their entire support platform to stack overflow) you will frequently find smug answers from the Parse team. It's common to see a Parse support teams response get tens of downvotes. It's bad. I once found a bug in Parse's method to save multiple objects at once (if the first request failed, all subsequent requests would fail too) and getting a fix implemented was a huge pain. I had to go through Facebooks bureaucratic big reporting process only to be told the bug was a 'feature' weeks later. I resubmitted the bug with all kinds of proof and demo projects, and finally after a few months I got the email saying the big was fixed. Huge pain.
Overall, if a prototype backend will suit your needs, I highly recommend Parse. It's certainly not the pie in the sky it could've been, but it gets the job done and lets you quickly get to market.
The free push notifications, data dashboard, and analytics are some nice sugar on top too.
I will probably continue to build my products on Parse, unless I know I will need immediate scale, or a client requests a specific backend configuration.
Been wanting to try Firebase too. It's a Parse like BaaS where everything is done in realtime. Very cool, but still somewhat unproven and new. Has anyone used Firebase?