Often, the features we launch seem so simple that you might think they're the result of blatantly obvious design decisions. In fact, every feature is subjected to a healthy dose of scrutiny within the Gmail team, and usually that includes rapidly iterating on designs by collecting user feedback, learning what works and what doesn't, and improving on our work based on this knowledge.

Gmail chat is a great example of this: we went through handfuls of design ideas, from the tame to the insane, before coming up with the pop-up chat window (which is also called a chat "mole") that exists today. Here are a couple of the early ideas we tried out:
Chat within threads, making chats feel strongly related to the emails that may have spurred them:

Separate tabs for each chat conversation, which provided lots of space for chatting and kept these conversations separate from the inbox:

We took each of these ideas into the usability lab where users were asked to try them out and tell us what they thought. From these studies we learned that chatting was one of many multi-tasking activities that users performed, and that chats needed to be quickly available at any time, not just within a single email or on a separate page. So we tried a design reminiscent of a sticky note that stayed up on the page at all times:

Once again, we collected feedback on this design, leading to even more refinements. One of the biggest findings was that no one knew how to send a message because there was no Send button. Adding instructions (initially with a bright yellow background, and later with a more subtle design) made all the difference:

With chat and other features, collecting live feedback from real users helps evolve feature ideas into simple and more elegant experiences. These extra steps may add a little time to our development cycle, but they allow us to create a design that works well for the tens of millions of Gmail users.