An analysis should be done for a 24 hour period, since tunes are no doubt purchased in higher numbers during some hours (based on time zones for the target markets). This should be repeated for each day of the week, since that produces its own cycle. Once you had a one-week hour-by-hour measurement of the number of songs sold, you could assume the same volume for matching hours the next week and predict the day/time of the big moment. No, I haven't been gathering such data myself.