Lyris' Wendy Roth explains that email split tests are a great way to learn how to optimize your email marketing campaign.

Studies show that email campaigns sent on Monday get the best results. Or was that Sunday? And text links get more clicks than images-- or is it the other way around? 

Email marketing studies can tell you what works on average for others, but as they say, your mileage may vary. You've probably drawn some conclusions about the relative success of different campaigns, but it may be difficult to know what exactly about those campaigns made them better. Was it the creative, the offer, or just the time of the year? How can you know for sure? 

Email split tests let you conduct meaningful research on your own email marketing campaigns. To run an email split test, you create two or more messages and send them to randomized, equal portions of the larger list. You can then deliver the message with the best results to the remainder of the list. You'll have learned a little more about what works best with your list audience, and you'll be able to act on that information immediately by sending the best performer to the remainder. Best of all, you won't need to speculate or rely on hunches-- you can point to the statistics.

Careful preparation and planning will make your split tests a success. Here's how:

Identify your test goals
Consider first what you are hoping to optimize. Do you want to improve your open rate? Clickthroughs? Conversions? Reduce the number of opt-outs and spam complaints? Improve deliverability?

If your answer is "all of the above," you may need to choose a primary goal, but select a test that could potentially affect multiple metrics.

What to test
Anything in your campaign can be tested. But keep it simple-- if you test more than one element, you may not necessarily know which one made the difference. That said, don't be afraid to make the tests radically different. If the change is too timid or subtle, it may not make enough of a difference in response.

Here are a few ideas:

• Subject Line-- Often the first thing people test, because it's easy and low-risk to create several subjects and compare open rates between them.
• Layout, colors or format-- You may need big changes to see a difference in the response rate. If your campaigns are usually a single image, a text message might compel those who normally suppress images to open or click them. Or, if your messages are text-heavy, more images may improve readability, clickthroughs, and ultimately, conversion rates.
• Offer-- You may get a great response when you offer 25 percent off. But what if you got the same great results if you offered 10 percent off? Or no discount at all? Multiple splits for a campaign let you identify if you can get away with less. 
• Call to action-- Discover if you get more transactions with a to-the-point "Buy Now" link or lure them in with a lower-obligation "Learn More" link.
• Landing page-- Often neglected, it can make or break your campaign. Make the messages identical except for the page that clickers are taken to.
• Scheduling-- The conventional wisdom changes as to what time of day or which day of the week is best. Send an identical message at different times to know for sure. Just check that if you are scheduling a message, it really is sending it at that time; it's not much of a test if you are trying to compare mornings to afternoons and half of your morning send doesn't get out by noon.

Big enough… but not too big
What you test will determine how many recipients you'll need for your test sample. If you're testing open rates, a thousand recipients for each test should be sufficient. But if you're comparing conversion rates, a thousand recipients may not be enough for meaningful results. Check your baseline statistics for the list to be sure the test sample is big enough to return actionable results.

On the other hand, you don't want to make the tests too big. If you send a flop to 20,000 people, you've missed an opportunity to send those recipients the more effective message.

And the winner is…
You don't necessarily need to know what all the results will be before you decide which message won your test. Although you may see results trickle in for weeks, it's likely you'll see enough activity within 48 hours of the initial send to determine which message better achieved your goals. 

Some tests might take a bit more time-say, if you're evaluating whether a two- or three-day sales event triggers more transactions. In this case, you may want to wait four days to give procrastinating recipients enough time to act.

And don't focus on one metric to the exception all others. Sure, a test might have had higher sales, but the additional revenue might not have been worth it if it came at the cost of higher opt outs or spam complaints. Similarly, don't be depressed by fewer clicks if it turns out the right people were doing the clicking.

Split tests take much of the guesswork out of your email campiagns, so you can be bolder and more creative. When you try something out on just a small part of your list, you can risk trying something new. And if the risk pays off small in your test, it can pay off big for your entire list.

Wendy Roth is the strategic account manager for Lyris Technologies, a pioneer in email marketing solutions since 1994. She works closely with enterprise-level marketing and advertising professionals to help them achieve their email-related objectives, and collaborates with engineering teams to ensure Lyris' products continue to be based on marketers' changing needs.