My D3 strategy -

Find the type of visualization I want to make on http://bl.ocks.org/. Start with that code, understand the steps taken to build it, then tweak each step / the data source until I have what I want. Works fairly well and helps you learn the library, without having to roll everything from scratch.

There's a steep learning curve but a lot of that is to do with properly grasping the conceptual underpinnings of the library. It's not a charting library, thinking of it as a charting library and it'll make little sense.

As someone who'd been making data-viz on the web for a good 8 years or so before D3 came along I had a pretty good understanding of the domain and was immediately impressed with how well D3 fitted in with the way I wanted to work. A lot of its parts were better implemented better thought out versions of ad-hoc tools I'd built up over the years.

If you're really interested in making cool visualisations so the NY Times hires you then you have to put some work in to learning the tools and understanding the theory.

If you are skilled with ggplot2, this use case appears to abstract away the pain of d3.