1. Be sure to send all e-mail at the same time. You don’t want the day or time of the send to influence results (unless that’s what you’re testing).

2. Make clear notes about what you’re testing and which group is getting the "business as usual" version. This makes it easy to quickly determine a winner and implement what you’ve learned.

3. Wait at least 48 hours before declaring a winner. Early results aren’t always accurate. Often, what’s ahead after an hour doesn’t maintain its momentum.

4. Make sure test groups are of significant quantity. You can do an analysis to determine statistically significant sample size and margin of error. Or you can use these rules of thumb:

• Have at least 1,000 e-mail addresses per group; 5,000 is better if your list is over 10,000.
• If your total list is smaller than 2,000, repeat the same test over a few sends until you have aggregate results for 2,000 e-mail addresses (1,000 per each group).

Continue reading here and learn more about the most common things to test in email marketing: sender lines, subject lines, and the body of the email and the landing page.