Your latest online promotion is going gangbusters. Lots of clicks, lots of calls, lots of purchases. Everybody’s happy. But could it perform even better?

A/B testing can answer that question. Though often dismissed as only useful if you need to improve underperforming campaigns, A/B testing is always applicable, and you should incorporate it into all your online marketing efforts. Testing a single element such as an image, a headline or your call to action —the list is limitless — can shine a light on what’s working, as well as what could be better.

So, what should you test? EVERYTHING! That might sound ridiculous, but it’s true. How do you know that a larger image on your landing page won’t perform better if you don’t try it? Does a shorter signup form result in more leads? Will an image of a human perform better than an illustration? The possibilities are endless, and the only way to find those answers is to test, test, test.

OK, Fine. So, How Do I Go About This?

 Start with a plan. Not just a list of elements you want to test but also how you’ll manage the process. A clear plan of action ensures that you and your team stay on track.

  • Take notes. Keep track of every variation in your testing and their performance. You’ll thank yourself later, as this will be invaluable when you begin to refine your campaign.
  • Go beyond the obvious. Yes, it’s a good idea to test image size and button colors. But have you tried different testimonials? Does shorter copy make a difference? Does a visual cue, like an arrow pointing to the CTA result in more conversions?
  • Allow yourself to fail. Some of your efforts are going to be duds. That’s just the way it goes. Instead of moving on, view these situations as opportunities to identify why something failed and how it could be better next time. Apply lessons you’ve learned from other tests to the underperforming variant.
  • Give it time. In most cases, a “one and done” approach to A/B testing won’t give you reliable, actionable data. No matter what you’re testing — copy, headlines, email subject lines — you should run at least two tests to make sure your results are reliable and weren’t affected by elements outside of your control.
  • Never stop. It’s tempting to declare a test “done” when one variation blows the other one out of the water. But what if that version could be made even better? Keep testing.