I didn't spend time doing an analysis beforehand. I did IE8 support which was impossible to ignore given its ~30% market share at the time. A friend in our early alpha testing said she couldn't use our app unless it supported IE6 for her site. Since I valued her feedback I started hacking to see how difficult it would be, and got most features working that day. From our [controlled output] perspective, IE6 really isn't that much different than IE8. The end of the story is that she filed several critical bugs dealing with how we insulate our JavaScript against other scripts, so it was worth it from that perspective alone!