The document analyzes trends in conversations on Twitter over 47 weeks for 11 topics varying in subject and geography. It finds that topic matters more than time for identifying trends, and that the most consistent keywords are for subjective topics. Trends are also more dependent on global rather than local background noise. These results provide suggestions for building tools to monitor trends, such as frequency of data collection, whether to use short-term or long-term data, and if local or global data is better.