5 Rules of Thumb for Testing

by ·1 Comment 

We’ve given some great ideas about testing when it comes to offers, timing and calls-to-action in the past which is great. But there are so many things to think about when you’re doing any kind of testing with your business. So we thought about it, I mean you don’t put both salsa and ketchup on your eggs at the same time without trying each one first right?

Here are 5 rules of testing you can put to work today!

  • Test one variable at a time: No need getting confusing results. Again, if you are a pro at this and your site gets lots of traffic, you can of course test A-B-C-Z variables. But if you’re a little more wet behind the ears as they say, try to choose 1 thing to test. Meaning, test your button CTA vs. a sign-up form, but typically don’t test radically different copy at the same time…
  • But…Sometimes Subtle Sucks: At the get-go, test variables that you think will show big variance. Spending time on tests that only provide a 5% variance are for folks that have tested and tested and…tested some more and are now at the point of eeking out whatever they can. But sometimes it’s Go Big or Go Home time. So, design 2 completely different landing pages – copy, design, format – and see what happens. You WILL get high variance. Then you just have to break things down and start testing in different variables. Exciting!
  • Run it through the RANDOMIZER!: What? This just means, make sure that you aren’t getting skewed results because, for instance, you had returning visitors only see one page, and new visitors see the other test page. That’ll really screw things up!
  • Timing is everything: Meaning, test simultaneously at the same time! People behave differently on a Wednesday than a Saturday. So if you test one variable on Wednesday and the other on Saturday, your results will be kaput!
  • Get a good sample size: We’re talking Statistically Significant results here people, standard deviations and all that stuff you slept through in calculus class. So, don’t pull the trigger too fast and declare a winner of your test after a handful of responses. Most commercial testing apps out there will tell you when a test result IS statistical. But if you’re going old school and looking at log files or web analytics or your shopping cart software, try to get 50-100+ responses in for a valid winner. And try to test for several days to normalize your test and minimize any extraneous circumstances that may skew results.

So go forth and test and get as much out of your testing as possible!