In my postal-mail days, there were major variables (offer, price, creative, list, etc.) and minor ones (font, colour, graphics, etc.). All affected response rates. The rule of thumb was to first test all the major variables, as they accounted for the greatest response increases. Then, test minor variables to squeeze out even more orders.

In the e-mail world, the postal-mail model must be updated, especially in light of recent stats. First, from an EmailLabs study:

• Subject lines of fewer than 50 characters performed better than subject lines of over 50 characters:
• Open rates were 12.5 percent higher.
• CTRs were 75 percent higher.
• E-mail messages with over 25 links fared better than those with fewer than 25 links. “The more links there are,” says EmailLabs Loren McDonald, “the greater the chance that one or more will resonate with the recipient and motivate them to click through.”

Though the following Advertising.com stats are based on the Web, not specifically e-mail, they offer clues about what variables to consider testing:

• A call to action (CTA) accounted for an 85 percent lift in revenue compared to no CTA (that’s a no-brainer, of course).
• Colour accounted for lifts of up to 50 percent.
• Disclaimers located at the bottom of a Web page accounted for a 14 percent lift.

Let’s create a comprehensive list of variables to test:

• Offer (e.g., 30-day free trial vs. free shipping)
• Price (e.g., $79.95 vs. four payments of $19.95)
• Copy
• Graphics (e.g., photo vs. drawing)
• CTA
• List
• Format (text vs. HTML)
• Copy length (long vs. short)
• Appeals to consumer (e.g., look beautiful, improve health, etc.)
• Colour
• Subject lines
• Number of links
• Disclaimer positioning

There may be other logical variables to test that are specific to your product or service.

The Right Way to Test

Testing numerous variables in postal-mail days was difficult because we had to create numerous variations of the mail piece and have enough names to test each one. With e-mail, things are a lot simpler.

The key to producing statistically significant test results is to isolate and test each variable independently. To do this, track your testing program on a spreadsheet. Say you want to test 10 variables. You want to use the same list for all tests, so your first test would look like this:

Test variable	Format	Text vs. HTML
Locked variables	Price	$19.95
?	Offer	Free shipping
?	CTA	Offer expires in 30 days
?	Colour	Red
?	Subject line	Creative A
?	Disclaimers	At bottom of e-mail
?	Number of links	10
?	Copy length	Short
?	Dominant consumer appeal	Beauty

You would create two messages, one in text and one in HTML, holding all other variables constant. Once you determine which works best, you move on to the next variable, so the next test would look like this:

Test variable	Price	$19.95 vs. $14.95
Locked variables	Format	HTML
?	Offer	Free shipping
?	CTA	Offer expires in 30 days
?	Colour	Red
?	Subject line	Creative A
?	Disclaimers	At bottom of e-mail
?	Number of links	10
?	Copy length	Short
?	Dominant consumer appeal	Beauty

And so on. Because you can create an e-mail message, send it, and read results within a day or so, you can easily test 10 variables in a couple weeks. Be sure to keep track of how each variable increases or decreases response. Because e-mail is so much easier to test and you can read results faster than postal mail, now’s the time to set up a formal testing program.

By Paul Soltoff