Doodle, the scheduling platform, brings people together to meet for a variety of reasons e.g. a quick one-on-one, board meetings, sales pitches, interviews, an evening of dungeons and dragons, etc. The more people participate in a meeting, the bigger the problem of finding a good time to meet becomes. We proposed to alleviate this by recommending advantageous time slots to the user to choose from. In this talk, we present the design, architecture and evaluation of such a suggestion service that proposes the best time to meet and how we learn from more than 100 million scheduled meetings.