Like a chemistry experiment, every test needs a hypothesis. Planning what you want to test lets you understand why you are testing it and what you hope to prove. There is no point testing if there isn’t a clear goal.

Try and get a range of tests planned in advance so that you can easily switch from one test to another. Ask yourself what features and benefits of your product or service you can include in the tests; for example you might want to test ‘free delivery’ against ‘next day delivery’ to gauge which one your customers value more.

Once you’ve established which of the above makes a better ad, you can move on to your next test, which could be the position of the winning delivery message – does it perform better on the first or second line of the ad?

What else to test

Once you’ve established your best performing sales messages, you can delve into the micro levels of ad testing.

Test punctuation – you’ll be surprised how people react to the most minute features of your ad copy. Even changing numbers for letters (e.g. ‘12 month warranty’ vs ‘twelve month warranty’) can have a substantial impact on your CTR.

Another feature I like to test is the call to action. ‘Buy now’ may be more or less popular than ‘shop now’, while different features are more appealing for different products – durability might be a key feature in mattress sales, but when you only have room for one selling point in your ads you might want to test this against an ad that pushes price or comfort. 

Keep changes small

Make subtle changes to each ad so you can be sure that the differences in performance are down to the one element you changed. Run the change against a control ad for a period of time and then pick the winner.

Again, when you reach this stage you’re not finished. Make another change and test again using the successful ad from the previous test as the control.

Pro tip: It’s vital to check your rotation settings. Set to ‘indefinitely’, as there is no point running a test if you’ve asked Google to prioritise the best performing ad! 

Running multiple ads alongside each other

I always run a test with two adverts, one control and one with the changes. The control should be the ‘winner’ of the previous test and by running just two ads you guarantee a clear result.

Having more than two ads running concurrently will always make decision making more difficult as you are invariably testing multiple elements.

When is the test over?

This is dependent on the traffic your ads see. You must ensure that you allow the ads to run for a feasible amount of time to ensure you have sufficient data to enable you to draw conclusions.

An account with high traffic can provide conclusive data within a week, but to be safe I suggest a minimum of two weeks to eliminate differences in performance based on day of the week. A less active account may take months to give you the answers you’re looking for.

Remember, remember:

Don’t forget what you’re testing, otherwise all of that hard work invested in the planning and prep will be wasted! Finding a way to log your tests is useful, even if it’s just in the form of a simple spreadsheet.

If you’re struggling for test ideas, I suggest reading PPC Hero’s guide to the top 20 ad copy tests for improved PPC performance.