You can test just about anything: text ads, banner ads, web pages, product colors, and so on. Testing lets you see what happens when you make changes. Which ad works better? What if you change something? You can find what works better. You can also delete whatever doesn't work. Always use data to make decisions.
There are two kinds of tests:
A/B Split Testing: This is a basic test and easy to carry out. Which works better? An ad with a summer or a winter background? A family or a couple? Instead of using opinions (or generally deferring to the person with the most seniority), you can test. Compare the results.
Multivariate Testing: This allows you to test many versions at the same time to find the best combination. Let's say you want to make the best-possible landing page so you test four headlines, three photos, five body texts, and three call-to-action buttons. That's 180 combinations (4x3x5x3 = 180). With a multivariate testing tool, you can easily create all of these combinations and test them to find the best combination. As the test is going on, you'll quickly see that some combinations are weak and can be turned off. Google offers a free multivariate testing tool as part of Google Analytics.
When the test ends, you're ready to start the next round. Look at the best results from your test and make new versions of ads or landing pages based on those results. If you find that a photo of a couple works better than one of a family, try new ads with the couple on the right side or the left side. Try different clothes on the people.
When I first started testing, I found that switching two words could result in a 2% difference in conversions. I tested all sorts of variations to end up with ads and landing pages with very high conversion rates.
All of your campaigns must include testing. This lets you be certain that you're using the best ads and landing pages.