Marketing’s Weekly Dose of the Truth

Ken Magill

About Us

Five Email Testing Pitfalls to Avoid

11/13/12

By Julia Peavy

With the holidays fast approaching, many email marketers are gearing up for the busiest time of year. Even though time and budget may be tight, email testing needs to be a part of your email marketing plans. Without testing, marketers can be leaving a great deal of money on the table. However, a poorly implemented testing program can actually be detrimental to your program. During the busy holiday season, watch out for these five testing pitfalls.

Timing

While potentially obvious, it is worth pointing out that email test cells need to be deployed at the same time (unless timing is the element that is being tested). If the test cells are sent at different times, you can’t be sure the element being tested was the reason for the difference between the cells.

Interpreting Results Too Soon

Email marketers are extremely busy (especially during the holiday season). This hurried environment can lead to interpreting results too soon in the effort to get the “winning” campaign out the door. If tests aren’t given enough time to run, you might deploy “the winner” which ends up actually being the loser. If possible, wait 48-72 hours before analyzing the results and deploying the winner.

Not Statistically Relevant

There’s nothing more frustrating than taking the time to deploy a test and not having the test cells be statistically relevant. If you don’t have statistically relevant cells, you can’t be confident your test is a likely indication of how your overall audience would behave. As a general rule, 2,000 to 3,000 individuals per test cell is a good guideline. Also, be sure your sample is randomized across your audience.

Unintended Consequences

Make sure to evaluate all of your campaign metrics to completely understand the test results. While you might have increased clicks with a special discount trying to win-back lapsed subscribers, you might have also increased your complaints to an unacceptable level. Your initial analysis might lead you to believe that one campaign was the winner, but a closer look at all the metrics provides a more complete picture.

Hidden Variables

Are there things outside the test environment that could be affecting the results to your email test? Direct mail, in-store offers, or special promotions on your social media sites are all things that could skew results. Do the holidays or seasonality make certain behavior more likely? Pay attention to what is going on outside of the email program to make sure you understand all the factors that could be affecting your test.

Avoiding these testing pitfalls during this holiday season can help ensure your email holidays are happy.

Julia Peavy is director, response consulting for email intelligence firm Return Path.

Comments

Show: Newest | Oldest

Post a Comment
Your Name:
Subject:
Comments:
Verification:
Please type the letters in the image above

Terms: Feel free to be as big a jerk as you want, but don't attack anyone other than me personally. And don't criticize people or companies other than me anonymously. Got something crappy to say? Say it under your real name. Anonymous potshots and personal attacks aimed at me, however, are fine.

Posted by: Tim Watson
Date: 2012-11-15 05:47:46
Subject: Statistcal significance calculator

All good points. To check statistical significance use a calculator such as this one http://www.emarketingdynamics.com/resultcalc.asp As well as sample size, the absolute value and relative difference also play a part in whether a result is valid.

Xverify