I doubt solving a puzzle is a good idea for the conversation rate. It would be interesting to see some A/B testing results...