PPC Testing Part 1: The Ground Rules

Repeat after me, then repeat again: success with paid search requires an always testing mindset. Why is this so important? I think that most of all, it is a continuous reminder that the difference between great search engine marketers and just good ones is that the great ones are always trying to get more efficiency from their accounts. They’re never satisfied because they know that even if they’re able to get an account humming through many optimizations, paid search is a constantly changing environment. Bids change. Competitors enter and leave the landscape. It’s never fixed, so how can a PPC account be set in stone? By keeping top of mind that all of your elements are always up for review, testing and optimizing, you’ll never get complacent.

Over the next few posts, I’m going to be talking about testing in-depth, so it’s important to set a few ground rules before we start.

Don’t disrupt normal account practices. Until you’re very confident with an optimization, don’t apply it to the entire account. The proper practice is to test a few keywords or small ad groups until you have sufficient data to justify larger changes. I’d say, never test on more than 10% of an account at any time. Of course, that’s really dictated by how well or poorly your account is performing at the time. “If it’s not broke, don’t fix it” is a good rule of thumb for accounts that are doing very well. But, if your account is really performing poorly, you may have to take bigger action sooner than later, and that’s understandable.

Don’t test too many variables at once. Frankly, if you test too much, the tests may start affecting each other and you won’t get clear results. Remember, think like a scientist. A lab worker won’t add chemicals A and B and C to chemical D to see how it affects things. They’ll add chemical A first and let that test run its course so they know what its affects are. Or they’ll run three different tests with single chemicals in each. The same applies for testing a paid search campaign. Don’t change too many bids, ads, keywords, match types, negatives, etc. at the same time or you’ll never know what worked. Try changing one or two things at a time and see what happens.

Give your tests enough time to get real results. Probably the biggest mistake I see with rookies is that they set up a test and if it doesn’t immediately yield the results they were looking for, they pull the plug. I can’t stress how important it is to give your tests enough time. It’s hard to say how long “enough” is, because I know many search marketers have their advertiser (or boss) looking over their shoulders waiting for results. But you need to set proper expectations when you start a test that it will take time to run its course. You’re learn from experience just how long you need to let each test run, but I’d say one or two weeks for small changes like mild bid adjustments and up to a month or more for larger scale tests. Frankly, if you have a nail-biting advertiser on your case, you may not even be able to attempt large tests. Stick to many smaller tests so you can make changes more often and appease their need for “action.”

Use the scientific method. You may remember from high school science class the four steps of the scientific method: construct a hypothesis, test, analyze and report your findings. That’s the basic framework for testing, but I think that the first task is the most important. I find that many search marketers start changing things around without first thinking about how those changes are going to affect the account. If you do that, how will you know whether it was your skills or luck that attributed to the end result? Before you set up a test, think to yourself, “If I make these changes, I think this result will occur.” That way, when the test works it can confirm your hypothesis and you can feel confident in your knowledge. If the test doesn’t work, it lets you know what you don’t know which can be just as valuable as getting confirmation so you can figure out what you need to learn.

Understand the context. Unfortunately, paid search doesn’t exist in a lab where the conditions are strictly controlled. There are many, many variables that affect your PPC accounts outside of what is in your control. The state of economy, the time of year/season and many other factors make worlds of difference. Using a silly example, if you sell Christmas decorations online and begin testing your paid search in January and February, there’s a good chance you’re going to see a drop in sales in your November and December data that had nothing to do with your paid search tests. The key here is to understand how the parts of your account would have normally performed before starting your test. That way you can compare your metrics against those benchmarks. Sometimes it’s even better to look at “last year at this time” numbers rather than data from the last few months.

Leverage the technology. This one probably appears in every list of best practices I have. But it’s just as true here as anywhere. There are testing tools available both within AdWords and from third party vendors that can really help you implement and analyze your tests. We’ll be getting into what some of these options are in the next few posts.

PPC Academy is a comprehensive, one-year search advertising course from beginning to end. Starting with the basics, PPC Academy progressively explores all of the varied facets of paid search, and the tactics needed to succeed and become an advanced paid search marketer.