All PPC advertisers, no matter how small, should be testing ad copy on a regular basis. Much has been written on this topic, yet I continue to see a surprising number of advertisers who aren’t doing any testing at all. Here are the steps I take when setting up a testing matrix for our clients.
Step 1: Turn off auto-optimization. Auto-optimization is based on click-through rate, which may or may not be the best success measure for your PPC ads. Additionally, auto-optimization skews the number of impressions for each ad variation, so you could have one ad generating 90% of the total impressions. That’s not the best way to conduct a statistically valid test.
In Google, go to Campaign Settings and find Ad Delivery Options. Expand that section of Settings, where you will see the option to Optimize or Rotate. Choose Rotate. Even though Google warns you that this is not a “recommended” setting, do it. Google doesn’t recommend it because it may not make them as much money if you rotate ads evenly. Selecting Rotate will ensure that each ad variation gets approximately the same number of impressions.
In Yahoo, you’ll also need to go to Campaign Settings, under Optimize Ad Delivery. Turn this off, and you’ll get a message saying “Ads will display in turn.” This is what you want.
Unfortunately, MSN/Bing doesn’t offer the option to turn off ad optimization. So, what I usually do is to test ad copy in Google and Yahoo, and then roll out the winner to MSN. Not optimal, but necessary.
Step 2: Create at least 2 different ads for each ad group. If you have a high-traffic campaign, you may want to test 3 or more ads, but 2 is the bare minimum.
Step 3: Let the test run until you have a statistically significant number of clicks and conversions. There are lots of statistical programs and applications out there that will quickly tell you whether you have enough data for statistical significance. I like Super SplitTester, a free tool from Perry Marshall. Super SplitTester tells you which ad variation gives you the best cost per impression – in other words, which ad will make you the most money – by factoring in both click-through and conversion rate. It takes seconds to key in your data and get the answer. Bookmark it and use it!
Step 4: Start a new test. Many advertisers make the mistake of taking the winning ad and pitting a new ad against it. Mistake, you ask? Indeed. It’s a huge risk to expose 50% of your PPC traffic to a brand new, untested ad that may or may not convert. Instead, I use the method outlined by Dan Thies in his SEMMY award winning post, Split Testing Adwords: You’re Doing It Wrong. The method is spelled out step by step in that article, so I won’t repeat it here. But do it – your bottom line will thank you.
Step 5: Rinse and repeat. There’s a good chance you’ll hit on a “strong hero” ad that’s hard to beat. Keep testing. Create an ad that you think will never work, and test it (just make sure to use Dan Thies’s method above to minimize your risk!). You may be surprised – I know I have been on more occasions than I’d like to admit.
Follow these steps, and you’ll be well on your way to improved PPC performance!